Campbell et al v. Facebook Inc.
Filing
183
EXHIBITS re 181 Administrative Motion to Seal Documents Accompanying Class Certification Briefs and Evidentiary Objections filed by Facebook Inc.. (Attachments: # 1 Exhibit 29 (Unredacted), # 2 Exhibit 30 (Redacted), # 3 Exhibit 31 (Unredacted), # 4 Exhibit 32 (Redacted), # 5 Exhibit 33 (Unredacted), # 6 Exhibit 34 (Redacted), # 7 Exhibit 35 (Unredacted), # 8 Exhibit 36 (Redacted), # 9 Exhibit 37 (Unredacted), # 10 Exhibit 38 (Redacted), # 11 Exhibit 39 (Unredacted), # 12 Exhibit 40 (Redacted), # 13 Exhibit 41 (Unredacted), # 14 Exhibit 42 (Unredacted), # 15 Exhibit 43 (Unredacted), # 16 Exhibit 44 (Unredacted), # 17 Exhibit 45 (Unredacted), # 18 Exhibit 46 (Unredacted), # 19 Exhibit 47 (Unredacted), # 20 Exhibit 48 (Unredacted), # 21 Exhibit 49 (Unredacted), # 22 Exhibit 50 (Unredacted), # 23 Exhibit 51 (Unredacted), # 24 Exhibit 52 (Unredacted))(Related document(s) 181 ) (Chorba, Christopher) (Filed on 3/28/2016) Modified on 3/29/2016 (kcS, COURT STAFF).
Expert Witness Report: Campbell vs. Facebook, Inc.
Catherine Tucker
January 15, 2016
TABLE OF CONTENTS
Page
I.
INTRODUCTION .............................................................................................................. 1
II.
ASSIGNMENT ................................................................................................................... 2
III.
SUMMARY OF CONCLUSIONS ..................................................................................... 3
IV.
BACKGROUND TO THE SPECIFIC FACTS IN THIS CASE ....................................... 5
V.
THE USE OF BEHAVIORAL INFORMATION TO ORGANIZE THE
INTERNET ......................................................................................................................... 7
VI.
MANY PROPOSED CLASS MEMBERS WERE UNAFFECTED, SOME
BENEFITED, AND QUANTIFYING BENEFITS OR LACK THEREOF
REQUIRES INDIVIDUALIZED INQUIRY ................................................................... 11
A.
Some potential class members were unaffected by the challenged practices ....... 11
1.
2.
B.
Potential class members were unaffected if the website did not
have a relevant social plugin ..................................................................... 11
Potential class members were unaffected due to the aggregate and
anonymous nature of the data used ........................................................... 16
Some potential class members benefited from the challenged practices .............. 18
1.
2.
C.
Some proposed class members benefited directly from incremental
publicity .................................................................................................... 18
Some proposed class members benefited indirectly from
incremental publicity ................................................................................ 21
It is difficult to determine the effect of the at-issue practices on some
potential class members ........................................................................................ 23
1.
2.
Due to the use of aggregate counts it is very unlikely any single
increment of the social plugin counter had a negative effect for that
individual .................................................................................................. 26
3.
VII.
A “Like” button does not necessarily imply endorsement........................ 25
It is difficult to determine potential negative effects of any sharing
of a URL without intrusive inquiry ........................................................... 27
REBUTTAL TO MR. TORRES’S REPORT ................................................................... 32
A.
Mr. Torres estimated “benefits” to Facebook, not “damages” suffered by
putative class members ......................................................................................... 33
B.
It is not clear what the proposed methodology relating to the Social Graph
is or why the alleged practices are being related to advertising ........................... 34
1.
Summary of Mr. Torres’s method for estimating the alleged benefit
to Facebook of enhancing the “Social Graph” .......................................... 34
2.
Mr. Torres’s method is based on a false assumption ................................ 35
i
TABLE OF CONTENTS
(continued)
Page
3.
C.
The parts of the proposed methodology where Mr. Torres does
give details have several flaws .................................................................. 38
It is not clear how the proposed methodology related to allegedly inflated
social plugin counters is linked to the disputed practice....................................... 43
1.
Summary of Mr. Torres’s method for estimating the alleged benefit
to Facebook related to allegedly “inflated” social plugin counters .......... 43
2.
The analysis focuses on the value of “Likes” to website owners,
which has no reliable link to Plaintiffs’ allegations of harm .................... 43
3.
The analysis fundamentally misunderstands or distorts why
website owners value “Likes” ................................................................... 46
D.
Mr. Torres’s two potential methodologies cannot be reconciled with each
other ...................................................................................................................... 49
E.
Rebuttal to Mr. Torres’s analysis as it pertains to statutory damages .................. 51
1.
Factor 1: Actual damage to the victim ...................................................... 52
2.
Factor 2: Whether the Defendant profited from the alleged
violation .................................................................................................... 52
VIII. CONCLUSION ................................................................................................................. 53
ii
I.
INTRODUCTION
1.
I am the Sloan Distinguished Professor of Management Science and Professor of
Marketing at MIT Sloan at the Massachusetts Institute of Technology (“MIT”) in
Cambridge, Massachusetts. I received an undergraduate degree in Politics, Philosophy
and Economics from Oxford University in the United Kingdom. I received a PhD in
Economics from Stanford University in 2005. I have been at MIT since completing my
PhD.
2.
My academic specialty lies at the intersection between the economics of the new digital
economy and advertising. I have conducted multiple studies on the diffusion of new
advertising technologies and have published various papers which examine how
marketing communications, in the form of advertising, perform, and which delineate the
factors that relate to their performance.
3.
I am Co-Editor of the Journal of Quantitative Marketing and Economics. I was an
Associate Editor at Management Science and a Co-Editor of the recent National Bureau
of Economics Research volume on the Economics of Digitization. I am also Associate
Editor for the Information Systems Research Special Issue on Social Media and Business
Transformation and Co-Editor of the Information Economics and Policy Special Edition
on the Economics of Digital Media Markets. I have published multiple academic papers
in leading marketing journals, including Marketing Science, the Journal of Marketing
Research, Management Science and the Journal of Marketing. I received a National
Science Foundation CAREER Award, which is the National Science Foundation’s most
prestigious award in support of junior faculty who “exemplify the role of teacher-scholars
through outstanding research, excellent education and the integration of education and
research within the context of the mission of their organizations.” I teach the core class
on “Strategic Marketing” to our Executive MBAs at MIT Sloan, as well as being the lead
faculty for marketing of the MIT Sloan specialized executive education offerings on
marketing, where senior executives come in for one or two day courses to refresh their
skills and discuss the current state of the art in marketing. I also teach a specialized class
on “Platform Strategy: Building and Thriving in a Vibrant Ecosystem” which discusses
strategies for building data-rich digital platforms.
4.
I have received a Paul Green Award for the paper “most likely to influence marketing
practice” for research I did on new kinds of online targeting. I also received the 2015 Erin
Anderson Award for the most notable Female Emerging Marketing Scholar and Mentor.
5.
I have testified on factors influencing privacy regulation before Congress and have
presented my research on privacy to the Federal Trade Commission, the Federal
Communications Commission, the European Commission, and the OECD.
6.
Further details of my experience are in my curriculum vitae, which I attach as Exhibit
DDD to my report. A list of my prior testimony is attached as Exhibit EEE.
7.
I am being compensated for my services in this matter at my customary hourly rate of
$1,050. While preparing this report, I have been assisted by certain employees of
Analysis Group. I receive compensation based on the professional fees of Analysis
Group. No compensation is contingent on the nature of my findings or on the outcome of
this litigation.
8.
Since my work on this matter is ongoing, I may review additional materials produced
subsequent to the issuance of this report, and/or conduct further analysis. I reserve the
right to update, refine, or revise my opinions, or form additional opinions, including in
response to further information from Plaintiffs’ experts, further clarification of Plaintiffs’
requested relief, and any additional information I may receive.
9.
A list of the materials I have considered to date in developing the opinions contained in
this report is attached as Exhibit FFF.
II.
ASSIGNMENT
10.
I have been asked by counsel for Facebook to:
a.
Use my knowledge of the economics of digitization to lay out the appropriate
economic framework for analyzing the use of aggregate behavioral information
on Internet link-sharing behavior.
b.
Assess whether proposed class members were commonly affected by the practices
challenged in the operative Complaint and the Motion for Class Certification, the
use of counts of URL links that are formed as URL message attachments to
2
provide “recommendations” to people who use Facebook, to provide analytics to
third-party websites and developers, as well as to increment the “Like” social
plugin counter.1
c.
Assess the framework for potential damages put forward in the Expert Report of
Fernando Torres submitted in support of Plaintiffs’ Motion for Class
Certification.2
III.
SUMMARY OF CONCLUSIONS
11.
Below is a summary of my opinions as of the date I submit this report. If additional
documents or information become available after I submit this report, I will review the
material and update my opinions as appropriate.
12.
The use of popularity information to organize content on the Internet is both ubiquitous
and helpful. It leads to a democratizing process, where content that may not usually have
been highlighted is brought to potential readers’ attention.
13.
The alleged practices, including the incrementing of the social plugin counter next to the
“Like” button, affected potential class members in a variety of ways. Though many of the
proposed class members were unaffected, some may have benefited, and in a few cases
some may have been harmed, individual inquiry is necessary to determine in which group
individual class members belong.
14.
Many potential class members were unaffected by the alleged practices. The fact that
many websites did not have social plugins that display counters limits the effects of the
practices. Furthermore, the aggregate and anonymous nature of the data means that any
of the incremental data stored may have had no practical effect even if a social plugin
1
2
See Plaintiffs’ Consolidated Amended Complaint ¶¶ 2, 4, 31, 38; Plaintiffs’ Motion for Class Certification at 2,
5-10.
I understand that Mr. Torres has filed a second report in this case; however, I received it only two days prior to
the filing of this report and my references and responses in this report are to his first report.
3
displaying a counter was present because these also relied on many potential other
sources of activity which potentially dwarfed those associated with this practice.3
15.
Many potential class members may have benefited from the alleged practices. Potential
class members would benefit if they viewed the URL they shared in a message positively,
and there was a chance – however remote - that the share boosted subsequent site
visitation. This is especially true if the potential class member had a financial interest in
the website associated with the URL.
16.
Although it is theoretically possible that a few potential class members may not have
benefited, such cases are hard to identify without individual inquiry. For example, there
may be cases where class members may not have wished the website they shared in their
message to receive an incremental boost to its social plugin count. However, to identify
such cases requires knowledge of the motivations of the potential class member for
including that URL in a message, which cannot be achieved without individual inquiry.
17.
Mr. Torres’s Report does not consider whether individuals were affected by the alleged
practices. Instead the focus of the report is on estimating the benefits of the alleged
practices to Facebook. Many third party websites never pay for advertising. However, the
two methodologies proposed by Mr. Torres do not actually relate to the ways that
Facebook could have benefited from the alleged practices:
a.
The first methodology proposed by Mr. Torres suggests that a measure of benefits
could be the incremental enhancement in advertising revenue attributable to the
increase in links in a “Social Graph” due to the alleged practices. However, none
of the alleged practices was associated with enhancing advertising. So the
assumed link between the alleged practices and an increase in Facebook’s
advertising revenue does not exist. There are also other technical and numerical
flaws in Mr. Torres’s analysis.
3
See January 15, 2016, Declaration of Alex Himel (“Declaration of Alex Himel”) ¶¶ 34 “it displayed the
combination of the following fields from the counters in the global share object record: share_count, like_count,
comment_count, post_count.”
4
b.
The second methodology proposed by Mr. Torres covers the period where social
plugin counters were potentially incremented if a URL was included in a
message. The proposed methodology, however, is disconnected from actual
benefits to Facebook. Instead, it tries to use data on the price that a third-party
website would have paid for a “Like” in a different context. However, this ignores
again the fact that many websites did not have a social plugin displaying a counter
so did not receive the benefits of the potential increment to the counter. Even in
the limited circumstances that websites may have been willing to pay for
additional “Likes” to be displayed on a social plugin counter, there is no evidence
that their savings on “Like”-getting expenditures would have been diverted to
Facebook in the form of advertising revenue as claimed by Mr. Torres in his
Report.
IV.
BACKGROUND TO THE SPECIFIC FACTS IN THIS CASE
18.
I understand from the Plaintiffs’ Motion for Class Certification that the behavior in
dispute is that Facebook allegedly “scanned” messages containing URLs and used the
counts of URL links that are formed as URL message attachments “to provide
‘recommendations’ to people who use Facebook, to provide analytics to third-party
websites and developers, as well as to increment the ‘Like’ social plugin counter.”4
19.
Therefore, in this report, I consider the ways that people who use Facebook could be
affected by each of these challenged practices.
a.
First, I consider the ways in which someone who uses Facebook could be affected
if a URL he or she shared in a message were used to help identify relevant
websites to highlight in recommendations if a website had a Facebook social
plugin that reported such recommendations. I understand that such use of the
4
Plaintiffs’ Motion for Class Certification at 2. See also page 31 of the Expert Report of Jennifer Golbeck
(highlighting the use of the incremented share count in “fueling recommendations algorithms, offering third
parties analytics data, and inflating engagement counts on social plugins”); Plaintiffs’ Consolidated Amended
Complaint ¶ 2 (“[W]hen [Plaintiffs’] ostensibly private messages contained links to other websites, also known
as ‘URLs,’ Facebook scanned those messages and then analyzed the URL in the link. If the website contained a
Facebook ‘Like’ button, Facebook treated the content of Plaintiffs’ private messages as an endorsement of the
website, adding up to two ‘Likes’ to the page’s count.”).
5
aggregate and anonymous data only occurred when the primary system of
providing such recommendations failed and that the practice ceased in 2014.5
b.
Second, I consider the ways in which someone who uses Facebook could be
affected if his or her gender, language, country, and age were part of the
aggregate statistics that Facebook offered on URL-sharing behavior to website
owners who accessed Insights or related APIs, as a consequence of the person
sharing a URL in a message between December 2011 and October 2012.6
c.
Third, I consider whether someone who used Facebook to share a URL in a
message, and did not modify the associated attachment, would be affected by the
possibility that between December 2011 and December 2012, Facebook may have
incremented a counter for a website that had that specific social plugin counter by
up to two “Likes”.
20.
I also understand from the Plaintiffs’ Motion for Class Certification that the proposed
class is defined as follows:
All natural-person Facebook users located within the United States who
have sent, or received from a Facebook user, private messages that
included URLs in their content (and from which Facebook generated a
URL attachment), from within two years before the filing of this action
[December 30, 2011] up through the date of the certification of the class.7
21.
In the next sections of this report I systematically consider whether these proposed class
members were unaffected, affected positively, or affected negatively by these practices.
By way of background, I first discuss the use of behavioral information to aid social
discovery from an economics perspective.
5
6
7
See January 15, 2016, Declaration of Dan Fechete (“Declaration of Dan Fechete”) ¶¶29, 34 “[D]uring the
proposed Class Period, URLs solely shared in messages did not appear to other people and were not searchable
using public APIs. They may have been included in data consulted for the backend PHP service that acted as a
backup resource for the Recommendations Feed if Taste was unavailable, but its use would be highly unlikely
and highly variable, if used at all.”
I understand from the Declaration of Alex Himel that this practice ended in October 2012. ¶61 “[O]n October
11, 2012, I changed the code to no longer include URL shares in messages in the aggregated, anonymous
counters visible to domain owners through Insights.”
Plaintiffs’ Motion for Class Certification at 10-11.
6
V.
THE USE OF BEHAVIORAL INFORMATION TO ORGANIZE THE INTERNET
22.
Over the past two decades, consumption of information has been revolutionized by the
Internet. The Internet has, of course, directly reduced the cost of disseminating and
receiving information, as digital data is virtually costless.8 However, another key advance
has been the use of user behavioral information to ease the process of social discovery of
content.
23.
Prior to the Internet, curation and recommendations were usually done by specialized
experts in the field. For example, newspaper editors decided what stories to feature on the
front page in order to attract occasional purchasers. Similarly, department store managers
decided which products to highlight in a store entry display that would appeal most and
induce browsing customers to purchase. Even in the non-digital world, organizations
used popularity information to help them connect with customers. For example,
musicians vied to make sure their songs ranked highly on the Billboard Hot 100 so they
would receive more exposure.9 National Public Radio (NPR Books) ranked books on the
basis of anonymous sales counts at independent booksellers.10 Similarly, toy shops would
examine sales data to identify a “hot toy” and feature it more prominently on their
display.11
8
9
10
11
Greenstein, Shane M., Avi Goldfarb, and Catherine E. Tucker, editors, The Economics of Digitization, Edward
Elgar Publishing, 2013.
As described here, the Billboard chart uses anonymized counts of streaming, sales and air-play to determine its
rankings. See Trust, Gary, “Ask Billboard: How Does the Hot 100 Work?” Billboard, September 29, 2013,
http://www.billboard.com/articles/columns/ask-billboard/5740625/ask-billboard-how-does-the-hot-100-work,
viewed December 11, 2015.
The NPR Bestseller Lists are compiled from weekly surveys of close to 500 independent book-stores
nationwide in collaboration with the American Booksellers Association. See, e.g., “NPR Bestseller List: Week
of Oct. 1, 2015,” NPR, http://www.npr.org/books/bestsellers/2015/week40/, viewed December 11, 2015.
An evolution of this is Kmart’s website: “2015 Fab 15 Toys,” Kmart, http://www kmart.com/en_us/dap/fab-15toys html, viewed January 3, 2016. For a discussion, see “Behind Kmart’s Fab 15 List: How We Identify Hot
Toy Trends,” SEARS HOLDINGS: SHC Speaks, September 24, 2014, http://blog.searsholdings.com/insideshc/behind-kmarts-fab-15-list-how-we-identify-hot-toy-trends/, viewed on January 15, 2016.
7
24.
However, the Internet has both automated and further democratized this process of social
discovery. Media websites can now organize their content based on popularity.12
Recommendation systems on websites such as Amazon now display automated
suggestions of products that might interest consumers based on other customers’
purchasing behavior.13
25.
Partly responding to and partly facilitating this general shift online in using digital
behavioral data to organize the Internet and facilitate social discovery, has been the
advent of electronic lists automatically recommending Internet content. A similar
evolution has taken place in the form of “counters” of social media website activity
surrounding a particular website. These counters take a variety of forms. The simplest
form they might take is as a simple counter at the bottom of the webpage. Figure 1
depicts such a counter for a fetal health charity that I support.14 Sadly, the Fetal Health
Foundation is a small charity which does not receive much publicity or support from the
general public, so its counts of social media support are negligible. However, if I shared
the link on my Facebook page, retweeted the link using Twitter, shared the link on
Google Plus or posted the website on my Pinterest page, I might be able to build interest
and support for this charity, and also potentially inform those who are pregnant or who
know people who are pregnant who may face such conditions.
26.
I can achieve the aim of increasing interest and readership of this website from my
sharing the URL through a Facebook message in two ways. The first way is the direct
effect I would have from sharing the URL with my friends and relatives. The second way
is the indirect effect I would have from an increment of the social plugin counter from
“Likes” from three to four, should I have sent the email in the limited period such a share
would have affected the count. Though individually my share is unlikely to have any
12
13
14
Even news websites which stick to the traditional curation model also feature browsing and sharing information
to help customers identify news articles. For example, the New York Times lists the ranking of the top ten most
shared articles: “New York Times Most Popular Articles,” The New York Times, http://www nytimes.com/mostpopular, viewed December 11, 2015. For more details, see Berger, Jonah, and Katherine L. Milkman, “What
Makes Online Content Viral?,” Journal of Marketing Research, Vol. 49, No. 2, April 2012, pp. 192-205.
See Resnick, Paul, and Hal R. Varian, “Recommender Systems,” Communications of the ACM, Vol. 40, No. 3,
March 1997, pp. 1-3.
The Fetal Health Foundation is focused on advances in technology that help prevent babies dying in utero.
8
effect, if one hundred thousand people also shared this website it would help it gain more
general prominence as visitors may take the website more seriously upon observing the
high count when visiting the website.
Figure 1: Snapshot of Social Media Counters from Charity Website15
15
“Amniotic Band Syndrome,” Fetal Health Foundation,
https://web.archive.org/web/20150910004526/http://www fetalhealthfoundation.org/amniotic-band-syndrome/,
viewed January 6, 2016. Interestingly, I initially viewed the image for this URL on October 29, 2015. When I
returned to the website in January 2016, the site had experienced a redesign – quite unrelated to anything to do
with this case – and there was no longer this visible social plugin counter. This change illustrates the extent to
which it can be problematic to identify whether or not a particular social plugin was or was not on a website
over time, which I discuss below.
9
27.
In general, this personal example shows three things:
28.
First, the use of data and anonymous counts on the sharing of content online is
ubiquitous. Social media platforms such as Facebook, Twitter, Google and other websites
such as Pinterest all offer such data. Second, behavioral data relating to the sharing of
website URLs is useful for identifying which websites may be of most interest. Third, the
process is somewhat democratizing. Websites or niche causes such as the Fetal Health
Foundation find it hard to attract attention, but if they can achieve shares and signal their
popularity and relevance on social media this can help them spread their message.
Indeed, my own research shows that the release of popularity information benefits niche
products or less common websites most. Everyone expects a major website such as
CNN.com to be visited, shared, and “Liked” a great deal; information that a smaller or
unusual website is also visited, shared, and “Liked” a great deal is more surprising and
can therefore attract more notice.16
29.
This research builds on an older (and growing) economic literature about the effects of
popularity information on people’s behavior. This literature uses the insight that when
quality of a product or service or piece of Internet content is uncertain, people may use
others’ behavior as a guide to quality.17 Other academic papers have subsequently also
noted the usefulness of such information for improving users’ browsing experience
online and decision making.18 Other work has emphasized that such algorithmic rankings
may improve a user’s experience by reducing the potential for information overload.19
16
17
18
19
See Tucker, Catherine, and Juanjuan Zhang, “How Does Popularity Information Affect Choices? A Field
Experiment,” Management Science, Vol. 57, No. 5, 2011, pp. 828-842.
The canonical academic works in this area are Bikhchandani, Sushil, David Hirshleifer, and Ivo Welch, “A
Theory of Fads, Fashion, Custom, and Cultural Change as Informational Cascades,” Journal of Political
Economy, Vol. 100, No. 5, 1992, pp. 992-1026, and Banerjee, Abhijit V., “A Simple Model of Herd Behavior,”
Quarterly Journal of Economics, Vol. 107, No. 3, August 1992, pp. 797-817.
See Tucker, Catherine, and Juanjuan Zhang, “Growing Two-Sided Networks by Advertising the User Base: A
Field Experiment, Marketing Science, Vol. 29, No .5, 2010, pp. 805-814.
See Ghose, Anindya, Panagiotis G. Ipeirotis, and Beibei Li, “Designing Ranking Systems for Hotels on Travel
Search Engines by Mining User-Generated and Crowdsourced Content, Marketing Science, Vol. 31, No .3,
May–June 2012, pp. 493-520.
10
Still more have shown explicitly that users make inferences from many different types of
user behavior once that information is available online.20
VI.
MANY PROPOSED CLASS MEMBERS WERE UNAFFECTED, SOME
BENEFITED, AND QUANTIFYING BENEFITS OR LACK THEREOF
REQUIRES INDIVIDUALIZED INQUIRY
30.
In order to assess whether the effect of the challenged practices was common across
proposed class members, I evaluate how different types of proposed class members were
affected. I conclude that many of the potential class members were unaffected by these
practices and some may have benefited. However, identifying whether or not, and how,
they are affected requires detailed individual inquiry.
A.
31.
Some potential class members were unaffected by the challenged practices
To understand whether a potential class member was affected, it is useful to understand
and assess the likelihood of the set of coinciding circumstances that are necessary for
there to be any concrete effect from the alleged practices.
1.
32.
Potential class members were unaffected if the website did not have a
relevant social plugin
There are many potential circumstances where a URL attachment was created but where
there was no actual effect on visible social plugin or visible data and therefore, there is no
concrete effect on a proposed class member. To illustrate this it is useful to consider an
example from one of the Named Plaintiffs. This example is shown in Figure 2, which is a
message that Plaintiff Campbell sent on July 27, 2011, to
20
:
See Tucker, Catherine, Juanjuan Zhang, and Ting Zhu, “Days on Market and Home Sales,” The RAND Journal
of Economics, 2013.
11
Figure 2: Message from Plaintiff Campbell, as produced by Plaintiffs21
33.
In this example, Mr. Campbell shared a link to
dated July 26,
2011, which is shown in Figure 3. The message was shared on July 27, 2011. However,
when I visit the indicated URL in 2016, I see no social plugin counter or any evidence of
the incremental “Like” generated by this share being used in any way. (See Figure 3.) I
also see no social plugin which could have shown recommendations. This suggests that it
is not the case that the “Like count [is] publicly displayed”22 on all websites for which a
URL attachment was created as claimed in Plaintiff’s Motion for Class Certification.
Similarly, when I visited the page as it was posted on August 25, 2011 through the
21
22
CAMPBELL000076-77.
Plaintiffs’ Motion for Class Certification at 9.
12
Wayback Machine,23 no social plugin counter or plugin which could have shown
recommendations was visible.24
Figure 3: Not all websites had social plugins that featured a counter or Recommendations
or Activity plugin25
23
A query on
returns a page with no Facebook social plugin counter (viewed Jan. 6, 2016). However, the Wayback Machine
reports that “[a]rchived web sites in the Wayback Machine do not always appear as they did on the live web for
reasons such as the previously mentioned difficulties in archiving web sites,” and further, “difficulties in
archiving web sites include the use of JavaScript, server side image maps, orphan pages (web sites that are not
linked to by any other web pages), and unknown sites.” This only underscores the difficulty in determining if
putative class members were affected by the alleged practices.
24
already offers a ranking of its most emailed stories as a way of using popularity
information to inform readers what story they might most enjoy. See “New York Times Most Popular Articles,”
The New York Times, http://www nytimes.com/most-popular, viewed December 11, 2015 for the current list.
This New York Times list of most popular emailed articles was studied in research from 2012. See “New York
Times Most Popular Articles,” The New York Times, http://www nytimes.com/most-popular, viewed December
11, 2015.
13
34.
I understand that there are no Facebook records that allow determination of whether or
not a website accessed Insights or associated APIs.26 However, even making the
assumption that the
uses Facebook’s Insights tools or associated APIs, it
is also possible that the
would not have received any new information
from Mr. Campbell sharing the URL in a message. This is because if Mr. Campbell had
already visited the
website – which seems likely, as he shared the story –
his demographics would have already been visible and accessible to the
.27 Obtaining information on whether proposed class members had already visited
websites for which they included a URL in a message would be difficult, if not
impossible, and determining whether or not such a potential class member had already
visited the website requires individual inquiry regarding website visitation for the specific
time period surrounding the sharing of the URL link.
35.
This example illustrates that trying to determine whether or not a social plugin displaying
a counter or recommendations is displayed, whether Insights or associated APIs was
affected or how that has changed over time would require exhaustive effort at tracking
down various permutations of a website’s display choices over time for each individual
link and the individual’s own website visitation data. Such an inquiry may not even be
possible,28 and would inevitably result in a very time-consuming effort.
36.
Moving from this example, motivated by one of the Named Plaintiffs, it seems important
to understand how widespread such instances of lack of meaningful effects are. Though it
is hard to obtain concrete retrospective data on how widespread such lack of visible
social plugins containing counters or the Recommendations Feed are, one 2012 article
25
,
viewed January 6, 2016.
26
Declaration of Alex Himel ¶ 63.
27
28
, viewed January 11, 2016.
See Declaration of Dan Fechete ¶¶ 32-33, 43-44; Declaration of Alex Himel ¶¶ 42-43, 64-65, 78-79.
14
suggests that at that time the number of websites with any social plugin was small, and
that of the top 10,000 websites, 75.7 percent of them were not integrated with Facebook.29
37.
One might conclude, therefore, that for approximately 75 percent of all Facebook
messages with URL attachments sent in the December 2011-December 2012 period,
there was no effect because there was no Facebook integration on the website, let alone a
visible social plugin counter or Recommendations Feed. However, without actual data on
site visitation and its relative spread over the 10,000 top websites (and websites not in
this top 10,000), it is difficult to assess the extent to which the 75 percent figure applies
to URLs shared in Facebook messages rather than simply the distribution of websites.
Though it seems likely that smaller websites are less likely to have social plugins, it is
worth noting too that many of the most popularly visited websites (as such Google,
Twitter, Wikipedia, and YouTube) do not currently have Facebook integration.30 The 75
percent figure may also understate the effect because there are many different types of
social plugins, and if the website only had a “Like” button and not a counter next to the
“Like” button, then obviously there would not have been the effects associated with
counters.31 The number of websites featuring a counter or a Recommendations Feed has
29
30
31
This website reports its methodology as follows: “We examined the HTML code of the home-pages of the top
10k sites in the world according to Alexa. To determine Facebook integration, we looked for the official ways
of integrating Facebook on sites, with paths such as facebook.com/plugins, connect facebook net and
graph facebook.com.” “How Many Sites Have Facebook Integration? You’d Be Surprised,” Pingdom.com, June
18, 2012, http://royal.pingdom.com/2012/06/18/how-many-sites-have-facebook-integration-youd-be-surprised/,
viewed December 11, 2015.
Of course since this 2012 survey, it is likely that these statistics have changed. For example as reported by He,
Ray C., “Introducing New Like and Share Buttons,” Facebook for Developers, November 6, 2013,
https://developers.facebook.com/blog/post/2013/11/06/introducing-new-like-and-share-buttons/, viewed
December 11, 2015 by November 2013, the “Like” and “Share” buttons were being viewed over 22 billion
times daily across more than 7.5 million websites.
See “Top Sites in United States,” Alexa, http://www.alexa.com/topsites/countries/US, viewed on January 15,
2016 for current top websites. Based on my review on January 11, 2016, Google.com, Youtube.com,
Wikipedia.org, and Twitter.com do not have Facebook integration.
“Social Plugins,” Facebook for Developers, https://developers.facebook.com/docs/plugins, viewed December
12, 2015. There are not only “Like” buttons but also “Share,” “Follow,” and “Send” buttons available. The
“Like” button is displayed at “Like Button for the Web,” Facebook for Developers,
https://developers.facebook.com/docs/plugins/like-button, viewed December 12, 2015. That is of course
supposing I get the code from Facebook directly rather than using a template – as shown on “The Best
WordPress Facebook Widgets,” Elegant Themes Blog, January 15, 2015,
https://www.elegantthemes.com/blog/resources/the-best-wordpress-facebook-widgets, viewed December 12,
2015, there are many other potential ways of including Facebook interactions on a website that do not involve a
social plugin counter.
15
likely changed over time; however, the direction and magnitude of this change during the
relevant period is unclear, as is whether the number of URLs for such websites sent in
Facebook messages would have changed in parallel.
2.
38.
Potential class members were unaffected due to the aggregate and
anonymous nature of the data used
There is reason to think then there would have been no effect simply due to the lack of
pertinent social plugin. However, even if a pertinent social plugin was present the nature
of the data used limits the likelihood of any concrete effects. In general, the disputed
practices use aggregate and anonymous statistics.32 This was confirmed by Plaintiffs’
technical expert Dr. Jennifer Golbeck in her deposition.33
39.
Let’s assume that, for a given URL shared in a message, there was a social plugin counter
for the website where the URL attachment pointed. Even with this assumption it is not
clear that each of the proposed class members would be adversely affected. In order to
illustrate this, Figure 4 displays a message I shared through my personal Facebook
account with an old friend who has reservations about the Halloween holiday. However,
when I visit the actual webpage, it is unclear how I would have been adversely affected,
or how this instance involves the challenged practices at issue in Plaintiffs’ lawsuit.34 In
this example, there is no counter next to the “Like” button – but instead a social plugin
displaying a share counter which is shown in Figure 5. Even if the share counter had been
affected by the disputed practices it is not clear how an anonymous shift in the share
32
33
34
Furthermore, often URLs themselves were further anonymized by users. For example, Mr. Campbell received
such an anonymized URL on July 13, 2011 at 9:01 am. In the message, Mr. Campbell received the following
URL: “
” The URL, which happens to direct to a
was
anonymized effectively due to this “bit.ly”-style abbreviation.
Dr. Golbeck testified as follows: “[T]he activity feed presents an edge case where there could be personally
identifiable information exposed. The other cases all use aggregated data that reasonably seems to be not
personally identifiable.” Golbeck Depo. Tr. at 311: 6-11. On page 312 of her deposition, there is clarification
about what such an edge case could be: If I shared www.catherinetucker.com in a Facebook message and it had
been shared elsewhere on Facebook, then it is possible that if it appeared in a friend’s activity feed, they could
guess that I was the person who had shared it, should the unusual combination of events occur which mean that
it would ever actually appear in a Feed.
Of course, this is a result of my visiting the site in late 2015. It is possible that in 2014, the website may have
displayed different aspects of the social plugin and even an actual counter, but it is difficult for me to ascertain
that without time-consuming inquiry, and perhaps impossible altogether.
16
counter from 765 to 766 would have affected me or influenced other’s subsequent
behavior.
40.
Further, any analysis of whether there was any real effect of an increment in the social
plugin counter is complicated by the set of technical circumstances that need to be met
for a social plugin counter to have incremented. For example, I understand that if
multiple people share the same URL at the same time, only one increment to the social
plugin count may occur.35
Figure 4: Sharing a Story about Halloween with a Friend on Facebook
Figure 5: The story that I shared with my Friend (as of 2015)36
41.
For the Halloween story example depicted in Figure 4, it is difficult to imagine how I
would have been adversely affected if the information were used, as alleged by the
Plaintiffs, in a recommendation algorithm that tried to highlight interesting content in a
35
36
Declaration of Alex Himel ¶¶ 28.
Fisher, Max, “Why Australia Hates Halloween,” Vox, October 31, 2014,
http://www.vox.com/2014/10/31/7137369/why-australia-hates-halloween, viewed January 6, 2016.
17
social plugin displaying recommendations on the Vox website should the primary system
for providing such recommendations have failed.
42.
It is also unclear how I would be affected if Vox had accessed the Insights tool or
associated APIs, as alleged by the Plaintiffs, had learned that their audience was slightly
more female, closer in age to forty, and more English-speaking than before, especially as
I have visited their website on other occasions, meaning that Vox would have presumably
already accessed this information and there would be no new incremental information. Of
course, as this happened in 2014, rather than prior to October 2012, this could not have
happened in any case.
43.
In general, this second example illustrates that even if one supposes a relevant social
plugin was present on the website for which the URL attachment was created, trying to
identify whether or not the potential for an increment on the social plugin counter had
any meaningful effect on anyone is difficult (and sometimes impossible). Furthermore,
the aggregate and anonymous nature of the data collected limits effects of the other
disputed practices in the time periods when the occurred.
B.
44.
Some potential class members benefited from the challenged practices
Plaintiffs’ Motion for Class Certification suggests that Facebook “monetizes the content
of these private messages for its sole benefit.”37 However, my analysis suggests that many
people who use Facebook benefit directly from the usage of URL share counts to allow
them and others to identify relevant and useful websites. In this section, I lay out two
potential ways that people who use Facebook may benefit.
1.
45.
Some proposed class members benefited directly from incremental
publicity
First, Plaintiffs who shared URLs in which they had a direct financial or vested interest in
publicizing may have benefited directly from this practice. For example, Mr. Campbell
stated that his law firm’s Facebook page was designed to serve the “advertising purposes
for the law firm.”38 As a consequence, Mr. Campbell (through his law firm) benefits any
37
38
Plaintiffs’ Motion for Class Certification at 1.
Campbell Depo. Tr. at 45:1.
18
time a URL associated with the law firm or the “Blue Hog Report” blog listed on his law
firm webpage is highlighted on social media or other online venues.39 Therefore, any use
of the sharing of URLs via Facebook messages which led to more prominence or
publicity for Blue Hog Report would have benefited him.40 In terms of the other alleged
practices, the potential benefit would be limited by the extent to which such a share
incremented the likelihood of Facebook recommending the URL to other users on a
Recommendations Feed, which in itself would have required a very fortuitous
combination of events—indeed, since it seems that the Blue Hog Report did not have a
Recommendations Feed, this would be an impossible combination of events. It is not
clear there would be any effect on the demographic information shared through Insights
since, as it is Mr. Campbell’s own webpage, presumably he visited it often.
46.
Additionally, many class members may have been actively seeking publicity by sharing a
URL. Again, I can use my personal Facebook account for an illustrative example, such as
the one depicted in Figure 6. This message exchange is between myself and the owner of
a website named Boston Events Insider, in a failed attempt on my part to obtain movie
tickets for Kung Fu Panda 2 for my daughters. In order to win these tickets, the
gentleman asked me to follow the steps on his website which were: “Like & comment on
everything on the Facebook page,” “Retweet everything in the Twitter feed,” and share
the URL on several websites (Facebook, Twitter, Google Plus, as well as others).
39
40
“Generating traffic back to the website from the actual Blue Hog Report Facebook page is the primary benefit
that I’ve seen.” Campbell Depo. Tr. at 63:7-9.
For example, Mr. Campbell could have benefited directly from potential increases in the social plugin counter
on his website. It appears from his deposition that the counter was only installed after Facebook stopped
incrementing the counter using URLs contained in private messages. See Campbell Depo Tr. at 222:18-19 (“Q.
[W]hen [was] that Like button [ ] installed[?] A. Sometime in 2013 or possibly early 2014.”).
19
Figure 6: URL sender actively seeks Social Media Activity surrounding the URL
47.
Given that the website owner was actively soliciting social media activity in order to
boost the perceived popularity of his website, he would have directly (and
unambiguously) benefited from any incrementing of the internal social plugin counter for
the website as a result of sending me this message.41 In a case such as Figure 6, where the
owner was actively seeking publicity, anything that would boost the likelihood of his
website being recommended would benefit him though at this distance the website does
not appear to have a social plugin displaying either the Recommendations or Activity
Feed. Furthermore, since he is presumably already constantly visiting his own website,
his own demographics being shared with him make no difference to him.
48.
This is not an isolated example. For example, Mr. Campbell, a Named Plaintiff,
apparently received three messages containing URLs relating to a
.42 Although
41
42
included multiple URLs in two of the three messages, in
See Events Insider, http://bostoneventsinsider.com/subscribe html/, viewed December 17, 2015 for details. In
this case it seems apparent that “Johnny” clearly benefited from the disputed practices by Facebook.
See Plaintiff Matthew Campbell’s Corrected Objections and Responses to Defendant Facebook, Inc.’s First Set
of Interrogatories. This series of messages are summarized by rows messages 409, 411, and 412.
20
all three cases he included his own URL,
.43 It seems likely that
would have welcomed any boost to the popularity of his
as a consequence of sharing the URL via a message. However, these messages
would also qualify
2.
49.
to be part of the potential class.
Some proposed class members benefited indirectly from incremental
publicity
Second, there is an even broader set of instances when a class member may have
benefited from the incremental counting of URLs from Facebook messages when they
shared a URL in a message in which they had an indirect interest in helping to publicize.
For example, Figure 7 depicts a message I shared with my husband regarding the efforts
our church was making regarding helping homeless people who had recently lost shelter
due to the closure of Boston’s largest homeless shelter. As a result of my sharing this
message, my husband publicly clicked “Like” on a posting about this mission on
Facebook. It is worth noting, as an aside, that by the class definition, my husband would
be a class member as a recipient of the message despite clicking on “Like” after receiving
the message.
50.
Though I shared this with my husband through a message, it is a website that I would in
retrospect welcome publicity for. Many people in Boston were unaware of the negative
effects of the closure of the Long Island Shelter, and I would welcome any incremental
publicity my sharing of the URL could generate for this cause, though any benefits are
clearly indirect.
43
21
Figure 7: Message where I indirectly benefited
51.
Since I shared this message depicted in Figure 7 too recently for it to benefit me by
potentially incrementing the social plugin counter (because Facebook ceased this practice
in 2012), the main avenue of benefit would be if it ended up favoring internal
recommendations of URLs on a social plugin displaying recommendations that my
church had installed on their website supposing that, for some reason the primary
generator of recommendations had failed as described in the Fechete Declaration.
However, as of 2015, it does not seem that such a social plugin exists.44 If my church
accessed the Insights tool or associated APIs, there is a chance they would believe that
their audience was (slightly) more female than before and perhaps closer to forty than
before. Since I already provide far more detailed information to them as a member, and
my gender and age are apparent to them every Sunday I attend, I am not sure how this
would affect anything. Of course, because this is after October 2012, there is no
possibility that this occurred.
52.
In a similar spirit, there are examples among the Named Plaintiffs where there are
potential indirect benefits. Plaintiff Hurley received a URL concerning
, with the URL
, which is
shown in Figure 8.
44
See Home: Old South Church, http://oldsouth.org, viewed January 11, 2016.
22
Figure 8: Message to Plaintiff Hurley, as produced by Plaintiffs45
53.
Although I cannot be sure without viewing the content of the message that was redacted,
may have been sharing this URL
that she appears to support.46
would likely benefit
from any publicity this organization would receive as it may lead to
Plaintiff Hurley’s precise level of interest
is not clear without further
individualized inquiry, but it seems unlikely he would view it as harmful if more people
as an indirect consequence of him receiving this message.
54.
Similarly, Mr. Campbell also sent or received messages that contained URLs related to
.47 Again, it seems likely that Mr. Campbell would have weakly
but indirectly benefited if any of these causes had been boosted as a result of the
inclusion of the URL in a message.
C.
55.
It is difficult to determine the effect of the at-issue practices on some
potential class members
One issue for assessing whether proposed class members were negatively affected by the
disputed practices in this case is that the Named Plaintiffs in their depositions revealed
that they have divergent ideas of what negative effects they could potentially have
suffered which also are not necessarily based on fact or the current class certification
motion.
45
46
47
HURLEY000001.
See
Depo. Tr. at 157:18-160:7.
See Plaintiff Matthew Campbell’s Corrected Objections and Responses to Defendant Facebook, Inc.’s First Set
of Interrogatories, Exhibit 1.
23
56.
For example, former Named Plaintiff David Shadpour said, “I object to the selling of the
data obtained from scanning the messaging.”48 However, the anonymous aggregate share
count data at issue in this case was never sold for money.49 Hurley states a concern
surrounding the monetization of his messages that a “company could use information
from my messages to – I don’t know – further narrow or target their advertising.”50
However, URL sharing behavior in Facebook messages was not used to refine targeted
advertising.51 Plaintiff Campbell focuses on the social plugin counter and emphasizes the
harm done due to “using my conversation with somebody else to generate ‘Likes’ that
don’t exist, to generate fictitious Likes.”52 Therefore, Mr. Campbell focuses on a practice
that was limited in time between December 2011 (the start of the proposed class period)
and December 2012 and would have left unaffected many potential class members who
did not use Facebook Messaging until after December 2012.
57.
Both Named Plaintiffs (Mr. Campbell and Mr. Hurley) stated that they did not suffer any
monetary harm.53 Three friends of the Named Plaintiffs were unsure of what harm they
suffered. Putative class member
states that “I don’t know enough about it
to know whether or not I’ve been harmed.”54 Similarly,
not aware of a direct negative impact front and center of my life.”55
stated “I’m
states “I
don’t know.”56
48
49
50
51
52
53
54
55
56
Shadpour Depo. Tr. at 91:2-3.
Defendant Facebook, Inc.’s Supplemental Responses and Objections to Plaintiffs’ First Set of Interrogatories at
40, 44. (“During the relevant time period (December 30, 2011 to approximately December 20, 2012), data or
information derived from messages (including URLs shared in messages) was not a criterion available to
advertisers in choosing the audience for their ads, and Facebook did not use data or information derived from
messages (including URLs shared in messages) to match ads to users.”)
Hurley Depo. Tr. at 162:21-24.
Defendant Facebook, Inc.’s Supplemental Responses and Objections to Plaintiffs’ First Set of Interrogatories at
40, 44
Campbell Depo. Tr. at 194:5-8.
Campbell Depo. Tr. at 192:5-9; Hurley Depo. Tr. at 153:20-24.
Depo. Tr. at 171:2-8.
Depo. Tr. at 248:16-17.
Depo. Tr. at 157:20. Mr.
may have benefited from the alleged practices as he has two Facebook pages
for his blog and his business venture, and uses these to “increase visibility” online.
Depo Tr. at 36:9, 69:12, 17- 19.
24
58.
The above testimony demonstrates the difficultly in determining if potential class
members have been harmed by the challenged behavior. In order to assess harm from an
economics perspective, one must have a clear definition of what is harmful, which the
Plaintiffs have failed to consistently provide.
1.
59.
A “Like” button does not necessarily imply endorsement
Let us start with the most concrete statement of harm which was Plaintiff Campbell
stating that “my speech was corrupted in that Facebook created ‘Likes’ for a page
irrespective of whether that was something that I would like.”57
60.
Underlying this argument appears to be the assumption that a “Like” is unambiguously
an endorsement. However, it is not clear that Facebook or more general web users view it
as such. Table 1 reports results from a marketing research survey conducted by
ExactTarget where they asked people who use Facebook why they “Liked” a company’s
webpage.58 What is immediately striking is that there are many different reasons why
people click “Like.” Table 1 shows that even in 2010, only 39 percent of users used the
“Like” button to “show my support of the company to others.” Instead, there are a myriad
of ways that the “Like” button was being used that do not necessarily imply
endorsement.59 This multi-purpose use of the “Like” button means that users already
anticipate that a count of Likes does not necessarily imply multiple endorsements, but
could derive either from users wanting discounts or offers from a particular website or
because they wanted to stay informed (for whatever purpose).
57
58
59
Campbell Depo. Tr. at 190:7-10.
According to the webpage, the survey was fielded from April 9, 2010 through April 13, 2010. The survey was
fielded through a MarketTools TrueSample online panel and completed by 1,506 U.S. respondents, aged 15 and
older, and stratified by age so that each age bracket contained no less than 200 responses. Responses are
weighted by age and gender according to U.S. Census Bureau population estimates and Pew Internet Project’s
online activity data to reflect the online U.S. consumer population.
A recent paper by researchers from Harvard found that consumers respond enthusiastically to invitations to Like
brands – popular or unpopular, new or established – and that such indiscriminate “Liking” suggests that
expressing a “Like” may not reflect deep preferences. John, Leslie et al., “What are Facebook ‘Likes’ Really
Worth?,” HBS Working Paper, 2015, http://rady.ucsd.edu/docs/events/lesliejohn.pdf.
25
Table 1: Why do people click “Like” for a company, brand, or association?
Motivation
To receive discounts and promotions
To show my support for the company to others
To get a ‘freebie (e.g., free samples, coupon)
To stay informed about the activities of the company
To get updates on future products
To get updates on upcoming sales
For fun or entertainment
To get access to exclusive content
Someone recommended it to me
To learn more about the company
For education about company topics
To Interact (e.g., share ideas, provide feedback)
Percentage
40%
39%
36%
34%
33%
30%
29%
25%
22%
21%
13%
13%
Source: “The Thin Line between Liking a Brand and Liking Its Social Marketing,”
eMarketer, September 8, 2010, http://www.emarketer.com/Article.aspx?R=1007912,
viewed January 8, 2016.
61.
Mr. Torres testified about one particularly clear example where a “Like” is not an
endorsement: “Facebook has hinted at introducing other alternatives for people to express
their response or reaction to posts and things like that” because, for example, “it’s always
been a curious thing that if somebody posts a death or reports a death in the family, that
the summary way to show your, your awareness of the message, or anything else, is to
click on ‘[l]ike.’”
62.
Such variance makes it difficult to assess whether potential class members have been
harmed and whether that harm is common across class members because the context in
which URLs are shared varies across messages and that context cannot be known without
individual inquiry.
2.
63.
Due to the use of aggregate counts it is very unlikely any single increment
of the social plugin counter had a negative effect for that individual
As well as a Like not necessarily implying endorsement, it is unlikely that a small
perturbation in the number of “Likes” displayed on a social plugin counter due to the
sharing of a URL by one individual will affect outcomes substantially and any potential
effect will vary substantially by website and time. Indeed, Mr. Torres explained this in
his deposition by pointing that a website with “[l]ike counts of, in the order of one or two,
26
then it’s a 100 percent increase” in the count. However, if the social plugin incremented
was for “Coca Cola, and they already have 500,000 ‘Likes’ on their third-party website,
that is a miniscule less than a 1 percent, so, they won’t be as influenced or as impressed
by the increase.”60 In other words, even if there is an effect, the effect would not be
common across potential class members and would depend on the nature of the URL
shared and the date.61
64.
Though this analysis focuses on the potential for negative effects of sharing, it also
applies for the potential positive effects of sharing. In many cases, due to the small likely
effects of any one potential increment of the social plugin counter, the potential for
positive indirect benefits of the type discussed in Section VI.B is small. It seems more
likely that there would be a positive effect in the cases described in Section VI.A, simply
because an individual promoting their own website via messages is more likely to create
the volume of URL attachments that could lead to a more sizable increase in the social
plugin counter should it be in a context where that was a possibility.
3.
65.
It is difficult to determine potential negative effects of any sharing of a
URL without intrusive inquiry
The facts that Likes are not necessarily interpreted as endorsements and that the potential
marginal effects of any one Like on a counter is small, limit any potential negative effects
from the alleged practices. However, even without these constraints, there are only very
unusual and individualized circumstances where I can envisage harm. Indeed, the only
circumstance I can identify when there could have been a potential negative effect on
60
61
Torres Depo. Tr. at 174:3-175:4 (“Q. Why does it make it appear that the integration is more effective than it is?
A. Because the like count is increasing, despite the fact that the person is not clicking on the like button on the
third party website. Q. And does that opinion depend on how much the like counter is increasing, based on
messages? A. Not necessarily. Q. Why not? A. Because it depends, it would depend on exactly what the
proportion of the enhancement is. During some, at some point, according to some of the experiments reported
on The Wall Street Journal, the like count was increasing twice, or, or, in a two-to-one ratio, to including the
URLs in the messages. So, if that happens to a website, a third party website that has like counts organic like
counts of, in the order of one or two, then it’s a 100 percent increase. If it happens to Coca Cola, and they
already have 500,000 likes on their third party website, that is a miniscule less than a 1 percent, so, they won’t
be as influenced or as impressed by the increase.”).
This is further complicated by the fact that rather taking notice of absolute numbers of social plugin counts,
consumers are more influenced by the location of the link on the homepage. This is something demonstrated
with my research that shows the importance of website location relative to the influence of popularity
information. Tucker, Catherine, and Juanjuan Zhang, “How Does Popularity Information Affect Choices? A
Field Experiment,” Management Science, Vol. 57, No. 5, 2011, pp. 828-842.
27
people who use Facebook would be if the proposed class member shares a URL with a
friend that they wanted to alert their friend about, but they would prefer for other people
to not visit the URL. It is difficult to imagine how to determine this rather nuanced and
complex set of circumstances without a great deal of individual inquiry.
66.
Indeed, I found it problematic to identify a straightforward example of a URL being
shared in a message that the sender would prefer not to be publicized. The closest
example I can find is as follows. My husband chairs a Fourth Amendment organization
called “Restore The Fourth,” whose previous website was at www.restorethefourth.net,
and whose current website is at www.restorethe4th.com. He shared a message over
Facebook
with
a
colleague
in
October
2015,
regarding
the
former
URL
“restorethefourth.net”. In the message, he noted that [an unknown] someone was
updating that URL. It could be argued that my husband would prefer traffic where
possible to not be diverted to restorethefourth.net as a result of his message, as he was
questioning whether having two parallel websites was potentially confusing. However,
even in this case – supposing it was affected by the alleged practices, which it was not
since the message was sent in October 2015 – it is not straightforward.
Figure 9: Example of a message where a social plugin count of the URL in the message did
not necessarily benefit the sharer
67.
Restorethefourth.net
has
no
apparent
social
plugin.
From
its
appearance,
restorethefourth.net is not advertising-supported or linked to Facebook in any way. Of
course, this would have to be verified, and one issue my husband faced in managing this
issue is that he is not sure who has control of this website and has been unable to find this
out from the domain registrar who hosts the website.
28
68.
If there had been an operational social plugin displaying a counter or Recommendations
Feed which led to the website content being somehow boosted, it is not clear that my
husband is harmed. He is not averse to the content of the website, but wants to be able to
coordinate messaging for the Restore the Fourth movement better across websites.
Generally, he would prefer that more people actively contact their Congress member to
express support for the Fourth Amendment, which is what the reactivated
www.restorethefourth.net was trying to do. Indeed, he would prefer the URL to be
recommended over any other URLs (such as celebrity gossip websites), with the sole
exception of the more current and comprehensive URL for restorethe4th.com. Finally, the
degree of harm, if any, is likely to change over time, as his organization may be able to
contact and work with the individual who revived the old URL.
69.
It seems unlikely that the website in question accesses the Insights tools or related APIs
from Facebook, but if they do, again it seems immaterial whether or not my husband’s
demographic data is included in their data, since he is representative of many of their
supporters and had visited the website before emailing the URL to his colleague. And
again, because the message was sent in October 2015, there was no potential for any
social plugin counter or Insights to be affected.
70.
Another example of this ambiguity over the potential for negative effects is the instance
of Mr. Campbell sharing the URL in Figures 10-11, which is
Given that Mr. Campbell’s
political views are not public information, and his blog, The Blue Hog Report, is focused
on government transparency, it is difficult to know whether or not he supports or opposes
and therefore it is difficult to know if he views the content
negatively or positively.
71.
However, understanding whether the intention behind the messages was to restrict public
support for such a viewpoint requires an understanding of Mr. Campbell’s political
views, how closely they related to the content of the URL, and whether the intention was
to challenge or support the sender’s or receiver’s political viewpoint.
29
Figure 10: Message sent by Plaintiff Campbell, as produced by Plaintiffs62
62
CAMPBELL000075-77.
30
Figure 11: Example of a message where a share count of the URL in the message did not
necessarily benefit the sharer (URL from message reflected in Figure 10)63
63
viewed January 6, 2016. I understand that this website currently displays a “Recommend” button as
opposed to a “Like” button. Nonetheless, the webpage remains illustrative as an example of the kinds of
political URLs Mr. Campbell shared.
31
72.
In any case, as can be seen, this is a nuanced and complex analysis, which would be
difficult to resolve without a great deal of intrusive and individualized inquiry. This can
be seen by quickly reviewing the variety of politically-themed URLs shared by the
Named Plaintiffs.64,65 Furthermore, as I discussed in the prior two sections, there are
reasons to think even in these highly individualized circumstances the possibility for
harm is limited.
VII.
REBUTTAL TO MR. TORRES’S REPORT
73.
Mr. Torres’s Report describes two potential methodologies for calculating damages. The
first is based on extrapolating the alleged benefits to Facebook of “enhancing the Social
Graph by including data intercepted in private messages.”66 It is not clear how Mr. Torres
proposes to extrapolate this value, or more importantly, why the methodology should be
connected with advertising revenues.67 Mr. Torres’s second potential damages
methodology is a more limited analysis that focused on benefits to Facebook connected
64
65
Similarly, Mr. Hurley received a message containing the URL as noted in HURLEY000002
, viewed January 3,
2016. Again, the benefit or detriment to the sender from publicizing the included URL would depend on the
sender’s political views.
See, e.g., Mr. Campbell’s message to
on August 28, 2013, with the following URL:
(CAMPBELL000005)
viewed January 8, 2016; Mr. Campbell’s
21, 2013, with the following URL:
on December
, viewed January 8, 2016; and Mr. Campbell’s message to
on August 30, 2013, with the following URL:
(CAMPBELL000052)
August 30,
2013,
66
67
Torres Report ¶ 35.
The Plaintiffs’ Motion for Class Certification explains Mr. Torres’s first proposal as follows: “The unlawfully
intercepted private message content contributes meaningful data to the Social Graph, increasing the quality of
its ability to provide predictive value, and, consequently, increasing Facebook’s advertising revenue and value...
A reasonable value to Facebook of the intercepted content can be assigned on a per-URL basis, and can be
allocated to class members on that basis.” Plaintiffs’ Motion for Class Certification at 22.
32
with increments to the social plugin counter. This methodology reflects a measure of the
costs a URL owner may have faced of obtaining the “Likes” through other means or the
benefits they may have obtained. This analysis is removed from any actual harm, and also
highlights the huge degree of variation and lack of commonality in the proposed
methodology.68
74.
Mr. Torres also opines in his Report that “Class membership [is] identifiable and
ascertainable based upon Facebook’s records.”69 However, Mr. Torres made clear during
his deposition that he was not offering an opinion on ascertainability70 and when asked
about paragraph 11.a. of his Report, stated that the “technical issue as to what records to
look at to identify the membership in the class, that’s not, that’s outside of my scope.”71
Therefore, my rebuttal of his report does not consider ascertainability; this is instead
addressed in the technical Report of Dr. Benjamin Goldberg.
75.
In my rebuttal to the Torres Report, I begin by observing that Mr. Torres has not
calculated “damages” to putative class members, but rather alleged “benefits” to
Facebook. I then consider each of the proposed methodologies in turn and whether these
two methodologies can be reconciled with each other. Last, I consider whether this
analysis informs underlying factors that relate to the appropriateness of statutory damages
from an economics perspective.
A.
76.
Mr. Torres estimated “benefits” to Facebook, not “damages” suffered by
putative class members
Although his report claims to describe the “Measure of Damages,”72 both of Mr. Torres’s
methods for estimating damages purport to be related to the benefits received by
68
69
70
71
72
The Plaintiffs’ Motion for Class Certification explains Mr. Torres’s second proposal as follows: “In addition,
Facebook generated value from its inflation of third-party Like counters. The economic benefit derived by
Facebook attributable to this conduct lies between two bounds: a higher bound represented by the cost that
client websites saved by not having to acquire additional Likes; and a lower bound determined by the market
value of artificially acquired Likes.” Plaintiffs’ Motion for Class Certification at 22.
Torres Report, ¶ 11.a.
Torres Depo. Tr. at 34:2-3 (“Q. [Are you offering an opinion on] [a]scertainability? A. No.”).
Torres Depo. Tr. at 93:7-14 (“Q. And are you offering an opinion in this case that class membership is
identifiable and ascertainable based upon Facebook’s records? A. To the extent that’s a technical issue as to
what records to look at to identify the membership in the class, that’s not, that’s outside of my scope.”).
Torres Report Section IV heading.
33
Facebook. This is confirmed in his deposition when Mr. Torres repeatedly noted that he
estimated the benefits allegedly received by Facebook, not damages suffered by putative
class members. For example, Mr. Torres stated: “So my report and methodology they
developed was asked to analyze the benefits to Facebook. So that’s, so, it doesn’t
calculate the detriment to the class members, or the potential class members, because it
wasn’t meant to.”73 Mr. Torres reiterated this several times in his deposition.74
77.
Therefore, Mr. Torres has not presented any method for estimating the actual damages or
loss, if any, suffered by individual putative class members. Furthermore, there is no
attempt to consider or evaluate any benefits enjoyed by putative class members and
integrate these into an evaluation of net damages.
B.
It is not clear what the proposed methodology relating to the Social Graph is
or why the alleged practices are being related to advertising
1.
78.
Summary of Mr. Torres’s method for estimating the alleged benefit to
Facebook of enhancing the “Social Graph”
Mr. Torres does not present a finalized methodology for estimating the benefit he alleges
Facebook received from enhancing the “Social Graph.” Instead he has “[laid] out the
methodology and the beginnings of the calculations that can be done with publiclyavailable information.”75 He states that he has not “finalized the calculations because I
haven’t received the precise data from Facebook.”76
73
74
75
76
Torres Depo. Tr. at 48:11-21 (“Q. Why doesn’t it examine, your methodology examine, instead of examining
benefit to Facebook, why doesn’t it examine detriment to the putative class? A. So, my report and methodology
that I developed was asked to analyze the benefits to Facebook, so that’s, so, it doesn’t calculate the detriment
to the class members, or the potential class members, because it wasn’t meant to.”).
See, e.g., Torres Depo. Tr. at 48:23-49:1 (“Q. So, you have not developed a methodology to calculate damages
to putative class members[?] A. That, that was not my task, no.”); 108:11-17 (“Q. . . . [H]ave you attempted to
calculate detriment to the putative class? A. As I said, that, that’s not part of my scope. My scope is to analyze
the benefits to Facebook.”); 279:7-11 (. . . [T]he methodology is attributing, is not measuring the effect, the
detriment, for example, to the class member, so it’s allocating to class members as a whole the benefits to
Facebook as a whole.”).
Torres Depo. Tr. at 107:2-9 (“Q. And do you lay out these calculations anywhere in your report? A. Well, in the
body of the report, in section 4, I lay out the methodology and the beginnings of the calculations that can be
done with publicly-available information. I haven’t finalized the calculations, because I haven’t received the
precise data from Facebook.”).
Torres Depo. Tr. at 107:2-9. I understand that Plaintiffs have not even requested most of this information from
Facebook. Declaration of Christopher Chorba ¶ 8.
34
79.
The methodology that Mr. Torres does present is premised on the assumption that
Facebook used information gathered from messages to expand and enhance the “Social
Graph,” and thereby allow Facebook to enhance the “value of its own social media
advertising platform.”77
80.
The most concrete statement of his methodology appears in paragraph 51: “Therefore, the
economic value of the benefits Facebook derives from the unlawfully gathered user URL
links is proportional to the impact of this additional information on the total information
on the Social Graph. In principle, the benefit to Facebook in this respect would be
measured by attributing the corresponding portion of the incremental value of the Social
Graph to the accretion of the unlawfully gathered links.” Mr. Torres then goes on to say
that the value of the Social Graph is the “product of the number of links (L) in the
Graph.”78
81.
In other words, Mr. Torres intends to calculate the benefit to Facebook by multiplying his
estimate of the value of the Social Graph, multiplied by the percentage of links in the
Social Graph obtained from Facebook messages as a percentage of all links in the Social
Graph.
2.
82.
Mr. Torres’s method is based on a false assumption
Mr. Torres’s methodology is based on the assumption that Facebook uses information it
obtained from Facebook messages to refine its targeting and increase advertising
revenues.79 However, Facebook did not incorporate any information from Facebook
77
78
79
Torres Report ¶ 36.
Torres Report ¶ 52. Mr. Torres also gives an equation for damages, D, which equal (Lt+1 – Lt)wt, where Lt+1 is
the next period’s number of links and Lt is today’s number of links. It is unclear what is meant by “next period”
and “today” in this equation. These labels may actually be intended to contrast the actual world with the “but
for” world where there was no counting of aggregate numbers of any URLs in messages. However, that is not
specified or clear. wt is the value of each link. It is also unclear what is meant by links or how they relate to the
storage of aggregate URL counts.
Torres Depo. Tr. at 45:3-13 (“Q. And what, based on your understanding of the allegations in the complaint,
and your assumption that those allegations are true, what was the benefit to Facebook, as you understand it? A.
Well, the accumulation of the information gleaned from the messages, basically, the edges between members
and the marketers and entities identified by the URLs, is accessible through, as part of the social graph, it’s
accessible to Facebook in developing the targeted advertising services that, that generate this revenue.”).
35
messages into the Social Graph to target advertising.80 Indeed, Dr. Golbeck confirmed at
her deposition that she found no evidence that URLs shared in messages were used to
target advertisements.81 However, in his deposition Mr. Torres reiterated his assumption
that the data was used to generate targeted advertising revenue.82
83.
Because of the assumption that Facebook incorporated information from Facebook
messages into the Social Graph to enhance targeted advertising, Mr. Torres’s
methodology is fundamentally flawed. In fact, Mr. Torres himself acknowledged that his
opinion crucially depended on how Facebook actually used the information from these
practices.83
84.
From a theoretical perspective, the value of a Social Graph84 lies in using friendship data
to help predict the relevance of products and services, as Mr. Torres notes when he cites
my research in which I say that such advertising “efficacy seems to stem mainly from the
ability of targeting based on social networks to uncover similarly responsive
consumers.”85 This refers to the fact that if people are friends, they are more likely to
80
81
82
83
84
85
Defendant Facebook, Inc.’s Supplemental Responses and Objections to Plaintiffs’ First Set of Interrogatories at
40, 44 (“During the relevant time period (December 30, 2011 to approximately December 20, 2012), data or
information derived from messages (including URLs shared in messages) was not a criterion available to
advertisers in choosing the audience for their ads, and Facebook did not use data or information derived from
messages (including URLs shared in messages) to match ads to users.”)
Golbeck Depo. Tr. at 223: 3-6 (“Q: You have information that some shares were used for recommendations but
not for that kind of targeted advertising. A: That’s right.”)
Torres Depo. Tr. at 45:3-13. See also Torres Depo. Tr. at 102:1-2 “Q: So, the first part of your analysis works in
conjunction with the second part of your analysis? A “Well they are related because, ultimately, the only
benefits are from advertising.
Torres Depo. Tr. at 47:12-14. “Q. And I’m saying, if the facts were as I described them, and not as you are
assuming them, would it impact your expert opinion in this case.?”…A. It would still depend on exactly how
Facebook would be using the information in that hypothetical.”
This term was used before Facebook to describe the use of graph theory to apply to social networks. See, e.g.,
Walsh, Toby, “Search in a Small World,” IJCAI, Vol. 99, 1999.
Tucker, Catherine, “Social Advertising,” February 15, 2012, SSRN (http://ssrn.com/abstract=1975897); see also
Torres Report, note 37. This reflects the concept of homophily discussed in the Golbeck deposition. Golbeck
Depo. Tr. at 100-101.
36
share similar interests and be interested in similar products and services.86 However, this
research highlights that it is the social connections that make Social Graph data
potentially valuable.
85.
However, none of the disputed practices embedded any social relationships in connection
with their use of URLs in messages. It was not the case that Facebook used or could have
used URL aggregate counts to identify the nature or intensity of social relationships.
Indeed, the inherent value of such aggregate URL share data is hugely diminished by the
fact that aggregate counts of website visitation are broadly and freely (or at least
inexpensively) available from many websites and providers such as Alexa, Compete,
Hitwise and comScore.87 Therefore, even supposing there was some link, which there is
not, between the alleged practices and advertising, it is unclear why a Social Graph is
relevant for Mr. Torres’s analysis.
86.
Indeed, more generally, the Social Graph does not drive all advertising revenue at
Facebook and the extent to which drives advertising revenue has changed over time. as
noted by AdAge in 2013, “Facebook has since introduced its ad exchange, FBX, and has
shifted its focus from social ads to more traditional web-advertising models, such as retargeting.”88,89 that do not rely on social relations. Furthermore, since 2013, Facebook has
also offered advertisers the potential to use custom audiences which offers access to an
86
87
88
89
This is a point emphasized by Dr. Golbeck in her TEDxMidAtlantic talk at minute 4:40 – the technical term
which she refers to in her talk for this idea is “homophily.” Golbeck, Jennifer, “The Curly Fry Conundrum:
Why Social Media ‘Likes’ Say More than You Might Think,” TEDxMidAtlantic 2013,
https://www.ted.com/talks/jennifer_golbeck_the_curly_fry_conundrum_why_social_media_likes_say_more_th
an_you_might_think, viewed December 11, 2015. Dr. Golbeck expanded on this in her deposition: “A. Yeah, so
homophily, H-O-M-O-P-H-I-L-Y, is a concept from sociology actually that basically birds of a feather flock
together that we tend to be friends with people who share our traits more than people randomly pulled from the
general population would share our traits.” Golbeck Depo. Tr. at 101:7-13.
See my recent paper, “Can Big Data Protect a Firm from Competition?,” jointly with Anja Lambrecht, for a
richer discussion of this point. Tucker, Catherine, and Anja Lambrecht, “Can Big Data Protect a Firm from
Competition?” December 18, 2015, SSRN (http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2705530).
Delo, Cotton, and Michael McCarthy, “GM Returns to Facebook Advertising after Public Split a Year Ago,”
AdvertisingAge, April 9, 2013, http://adage.com/article/digital/gm-returns-facebook-advertising-publicsplit/240785/, viewed January 3, 2016. This is the same trade publication noted by Mr. Torres’s Report in
Footnote 101.
See, e.g., Delo, Cotton, “Facebook Launches New Retargeting Alternative to FBX: Targeting to Use Tracking
Software That Marketers Can Attach to Websites and Mobile Apps,” AdvertisingAge, October 15, 2013,
http://adage.com/article/digital/facebook-launches-retargeting-alternative-fbx/244746/, viewed January 3, 2016.
This describes an advertising platform that is based on behavior outside of the Social Graph.
37
audience based on the audience the advertiser already has, such as email addresses or
phone numbers.90 Such analysis is further complicated by the fact that for any one click
there may be several drivers which interlink in complicated ways, making identifying
what drives any one piece of advertising revenue problematic. This is unsurprising given
literature in economics which highlights the difficulty of measuring the economic drivers
of advertising effectiveness.91
3.
87.
The parts of the proposed methodology where Mr. Torres does give details
have several flaws
As discussed, Mr. Torres’s proposed methodology is unrelated to how Facebook
benefited from the challenged behavior, as it did not use the aggregate stores of
anonymous social plugin count data to target advertising, which is the fundamental
assumption of his methodology. However, even supposing that the data collected was
related to advertising (which it was not), issues remain with Mr. Torres’s three
calculations.
88.
The first calculation in Mr. Torres’s description of this methodology is a table of
estimated messages (Table 2 in his Report, at page 19). However, the total number of
messages seem irrelevant to the key aspect of the data which is needed, which is how
many of these messages had URLs that created attachments. 92 Crucially, even a count of
URLs that generated a URL attachment does not reflect whether they were used in any
disputed practice—that is, whether the data was used as part of a social plugin counter
between 2011 and 2012, or used in the background in the provision of aggregate
90
91
92
“More Matching Capabilities with Custom Audiences,” Facebook Marketing Partners, November 30, 2015,
https://facebookmarketingpartners.com/partner-news/more-matching-capabilities-with-custom-audiences/,
viewed January 3, 2016. Facebook offers potential advertisers a number of ways of targeting customers beyond
relationships. Specifically, on its business website, Facebook offers that advertisers can target users not only
through location variables such as country, state, zip code, or local area but also through demographics, user
selected interests, and shopping or use behavior. See “Facebook Advertising Targeting Options,” Facebook for
Business, https://www facebook.com/business/products/ads/ad-targeting/, viewed January 6, 2016.
Lewis, Randall A., and Justin M. Rao, “The Unfavorable Economics of Measuring the Returns to Advertising,”
The Quarterly Journal of Economics, first published online July 6, 2015 doi:10.1093/qje/qjv023.
Torres Report ¶ 45. In his deposition he appeared to restate this to say, “[t]he data that I would need is mainly
the number of those messages that were intercepted that contained URLs, and the total number of messages for
the same time periods.” Torres Depo. Tr. at 27:20-23.
38
demographic data to website owners, or used as part of a recommendation. Individual
enquiry is necessary to make these determinations.
89.
The second set of calculations surrounds the alleged presence of 15.9 billion friendship
ties on the Social Graph.93 These data come from May 2011. Mr. Torres states, “I would
estimate the value of the enhancement to the Social Graph as commensurate with the
ratio of (1) intercepted URLs in private messages during the Class period to (2) the total
number of links on the Social Graph.” However, these second set of calculations does not
make sense as a denominator in Mr. Torres’s proposed ratio for two reasons. First, the
value of friendship ties that are used to target advertising is completely distinct from
aggregate URL counts, which are not used to refine targeted advertising. Second, even
supposing the aggregate link counts were used to produce advertising revenue, which
they were not, the number of friendship ties would be the wrong denominator. The
correct denominator, which would be orders of magnitude larger, would include not just
friendship ties but every interaction between friends on Facebook—every “Like,” every
“share,” every piece of demographic information, and the content of every public posting.
Further complicating the analysis, each of these different drivers of the potential for
Facebook to generate advertising revenues have different efficacy in different
circumstances and at different times.94
90.
The third set of calculations surrounds the value of the Social Graph. However, Mr.
Torres has also overestimated the value of the Social Graph for at least five reasons.
91.
First, there is an error in the calculation of the value of the Social Graph. The average
quarterly revenue Mr. Torres based his estimate on was total revenue, not advertising
revenue.95 This means that the estimates also include revenues from Facebook’s activities
93
94
95
Torres Report ¶ 49.
See, e.g., Tucker, Catherine E., “Social Networks, Personalized Advertising, and Privacy Controls,” Journal of
Marketing Research, Vol. 51, No. 5, 2014, pp. 546-562, where I show that different undergraduate institutions
have different values for advertisers, as does the rarity of information - for example, liking Oprah Winfrey may
be less informative than liking an obscure 1970s poet.
Looking at slide 9, 2015 Q2 Results PowerPoint, the average of total revenues over the past four quarters equals
the $1,771 figure noted by Mr. Torres in footnote 66 of his report. The average of total advertising revenues
could potentially be estimated from Slide 10 of the same document, at $1,622.25.
39
including payments in online games.96 Correcting this error (which Mr. Torres
acknowledged in his deposition) reduces Mr. Torres’s estimate of the Social Graph’s
value by $1.267 billion dollars.97
92.
Second, the choice of revenue numbers appears selective and problematic. The equations
in Mr. Torres report suggest that the change in value was contemporaneous with the
alleged practice, suggesting the use of revenue from the span of years governed by the
class definition. However, the Torres report instead uses just the most recent four quarters
in 2014 and 2015 as a basis for advertising revenue. Using the span of years covering the
class definition as a basis for average revenue, suggests a valuation of the Social Graph
that is $7 billion lower than the one suggested in the Torres report.98
93.
Third, Mr. Torres’s allocation of costs is as follows: “the additional information collected
through the accused activities has arguably zero incremental cost. Therefore, from an
economic perspective, virtually all of the incremental advertising revenue generated from
the enhancement can justifiably be considered incremental profit to Facebook.”99 This
seems arbitrary, as it is not clear from this description to what incremental part of
96
97
98
99
This led the estimates in Table 1 in his report to be off by $1.2 billion. This error was confirmed in his
deposition. Torres Depo. Tr. at 195:10-204:9.
Mr. Torres initially estimated the value of the Social Graph to be $15.087 billion. Torres Report ¶ 43, Table 1.
In his deposition, he stated that he intended the valuation to be $13.820 billion. Torres Depo. Tr. at 204:4-9 (“Q.
So, those three corrections on page 15, is that all, Mr. Torres? A. Yes. And then that feeds into the table 1,
where the annual profit numbers would be 3,459,000,000, and the discounted values in that line, for the whole
line, for the full column, would be 2915, 2457, 2070, 1745, 1470, 1239, 1044, and 880, for a total of
13,820,000,000.”). I understand that Mr. Torres has made corrections to the report to rectify this error but these
corrections were submitted too close to the deadline for the submission of my report for me to be able to review
them.
The actual amount of the overstatement is $7.056 billion ($7.056 billion = $15.087 billion - $8.031 billion (see
Exhibit HHH)). While it does not affect his estimate of the value of the Social Graph, Mr. Torres made yet
another error related to his revenue estimate. He claims his revenue estimate is based on “quarterly advertising
revenue from the activities of users located in the U.S. and Canada during the four quarters between April 2014
through June 2015.” Torres Report ¶ 39 note 66). The period April 2014 to June 2015, however, contains five
quarters, not four. A review of his calculations, after taking into account the $1.267 billion error identified
above, indicates Mr. Torres is using quarterly advertising revenue for the four quarters between July 2014 and
June 2015. Torres Report ¶ 39, n. 66, and Facebook, Inc.’s 2015 Q2 Earnings Report (July 29, 2015), slide 10.
Torres Report ¶ 44.
40
Facebook’s revenue-generating functions Mr. Torres thinks Facebook’s considerable
costs should be allocated.100
94.
Fourth, Mr. Torres excludes research and development costs from Facebook’s expenses
when calculating Facebook’s profit margin. His argument is that expenditures for
research and development are intended to yield benefits in the future and are therefore not
appropriate to be accounted for today to determine current period profits. Mr. Torres
claims this is consistent with “accepted valuation standards.”101 However, though Mr.
Torres is correct that valuation practitioners often exclude current period research and
development from current period calculations of profit, they still include research and
development expenses from prior periods that are resulting in benefits today.102 In fact,
the text that Mr. Torres cites as the basis for his Income Valuation Approach103 includes
research and costs as an expense in a sample income valuation case study.104 Moreover,
Mr. Torres assumes that the benefit to Facebook related to the Social Graph will accrue
over eight years. In order for the Social Graph to remain a valuable asset to Facebook, it
will need to continue to invest in the Social Graph. To the extent that this has historically
required Facebook to invest in research and development to support and develop the
Social Graph, this need will continue into the future and through Mr. Torres’s eight-year
time horizon. By failing to account for research and development expenses, Mr. Torres is
biasing Facebook’s profit margin up, which then biases his estimate of Facebook’s
benefits up as well. Including research and development expenses for the years Mr.
Torres considered in his valuation as a proxy for historical research and development
100
101
102
103
104
Considerable costs as defined by Mr. Torres (cost of revenue, marketing and sales, and general and
administrative expenses) and outlined in Exhibit GGG have ranged from 35 percent (Q2’14) to 103 percent
(Q2’12) as a percentage of revenue. Torres Report ¶ 39 and Exhibit 1.
Torres Report ¶ 39, note 67.
See, e.g., Damodaran, Aswath, “Research and Development Expenses: Implications for Profitability
Measurement and Valuation,” NYU Stern School of Business,
http://people.stern nyu.edu/adamodar/pdfiles/papers/R&D.pdf, in which he argues that research and
development expenses should be capitalized and amortized as opposed to being charged to the quarter in which
they are incurred. Importantly, in both positions it is assumed that research and development costs will be
accounted for somewhere in the valuation.
See Torres Report note 63 in which he cites Smith, G.V. and R.L. Parr, Valuation of Intellectual Property and
Intangible Assets, John Wiley & Sons, 2000; Reilly, R. F. and R.P. Schweihs, Valuing Intangible Assets,
McGraw Hill, 1999. Mr. Torres also cites Smith and Parr in footnotes 64 and 96.
See Smith and Parr, Table 18.3 on pages 510 and 511.
41
expenses would reduce the Social Graph valuation by over $7 billion down from the
number presented in the report.105 In combination with the correction to the selectivity of
the years used, the Social Graph valuation would drop from the $15 billion figure stated
by $10 billion.106
95.
Fifth, at a more conceptual level, Mr. Torres’s decision to give the Social Graph a
lifetime of eight years based on geographical mobility misses a critical fact: The nature of
Internet advertising makes geography not that relevant as a targeting variable relative to
friendship ties or expressed interests.107 Furthermore, the history of social networks has
shown the vulnerability of any social network site to turmoil and displacement and users
leaving the site.108 For example, it would have been wrong to assume that the Social
Graph embedded in MySpace in 2008 would have a lifetime value of eight years, given
that within less than a year its users had left the site in droves.109 Mr. Torres was in fact
posed with this hypothetical in his deposition and stated that in order to value the
MySpace Social Graph he would have to “perform a series of due diligence and
preliminary analyses.”110
105
106
107
108
109
110
The actual amount of the overstatement is $7.456 billion ($7.456 billion = $15.087 billion - $7.631 billion (see
Exhibit III)).
The actual amount of the overstatement is $10.704 billion ($10.704 billion = $15.087 billion - $4.383 billion
(see Exhibit JJJ)).
Indeed, my own research emphasizes that geography becomes meaningful as a targeting variable only when
offline advertising channels are not available to the advertiser. See Goldfarb, Avi and Catherine Tucker,
“Advertising bans and the substitutability of online and offline advertising,” Journal of Marketing Research
48.2 (2011): 207-227.
Tucker, Catherine, and Alexander Marthews, “Social Networks, Advertising, and Antitrust,” George Mason
Law Review, Vol. 19, 2012, pp. 1211-1227.
Torkjazi, Mojtaba, Reza Rejaie, and Walter Willinger, “Hot Today, Gone Tomorrow: On the Migration of
MySpace Users,” Proceedings of the 2nd ACM Workshop on Online Social Networks, 2009.
Torres Depo. Tr. at 211:21-212:5 (“Q. If you were tasked with valuing the social graph of Myspace in 2007,
would you have used a similar methodology as one that you’ve used here? A. Well, in that hypothetical
situation, I would have to, to perform a series of due diligence and preliminary analyses. I’m not sure that
Myspace had the same revenue mode, so I would have to reconsider the revenue model then, and, to see if that
is sufficient.”).
42
C.
It is not clear how the proposed methodology related to allegedly inflated
social plugin counters is linked to the disputed practice
1.
96.
Summary of Mr. Torres’s method for estimating the alleged benefit to
Facebook related to allegedly “inflated” social plugin counters
Mr. Torres’s second proposed analysis, which is related to the “Like” button next to a
social plugin counter, describes two potential bounds for damages related to each URL
attachment created.111 The first is to try and establish how much the website owner might
benefit from additional “Likes.” The second is to establish the market value of these
“Likes” in order to determine what website owners would have needed to pay in order to
acquire the “Likes.” However, both of these proposed methodologies are unrelated to the
claims made by Plaintiffs over the harm they suffered and seem to misunderstand the
reasons why website owners value “Likes.”
2.
97.
The analysis focuses on the value of “Likes” to website owners, which has
no reliable link to Plaintiffs’ allegations of harm
Mr. Torres’s methodology for estimating the benefits from inflating the social plugin
counter on third-party websites attempts to quantify the amount of money that third-party
website owners either received from the allegedly inflated “Likes” or would have been
willing to pay to acquire the allegedly inflated “Likes.” Even if Mr. Torres were to
measure these amounts accurately the benefit to the subset of third-party website owners
willing to pay for Likes are not benefits received by Facebook.
98.
Mr. Torres suggests that “In the Facebook environment, the number of ‘Likes’ measured
is typically interpreted as an indicator of the reach of an advertising strategy and, given
the particular brand/product combination, as a factor in generating sales.”112 However,
since “Likes” incremented were never used on the Facebook advertising platform to
measure the reach or success of a Facebook advertising strategy, this analogy is
misguided. Mr. Torres then attempts to link the benefit to third-party website owners to
Facebook by claiming that
111
112
Torres Report ¶¶ 62-71.
Torres Report ¶ 64.
43
“The amounts identified in this analysis – the cost savings to advertisers
from the accrual of Likes from the intercepted messages – were, in
principle, made available to spend on additional Facebook marketing
campaigns. This would have been particularly true in light of the false
appearance of increase [sic] Fan engagement that an inflated [social
plugin] count would present. To that extent, a fraction of this benefit may
have been converted to advertising revenue benefiting Facebook.”113
99.
The link is tenuous. Mr. Torres provides no method for determining if the cost savings
were actually spent on Facebook advertising or, if so, how much was spent. He does not
even argue with certainty that any of it resulted in incremental revenue to Facebook, just
that “in principle” it was available to be spent on Facebook marketing and that it “may
have been converted.” In his deposition, Mr. Torres confirmed only “a fraction [of an
advertiser’s cost savings] would have been converted,”114 to Facebook revenue, but was
unable to state what fraction, stating, “I can’t tell you because I don’t have the
information to determine it.”115
100.
Instead, the argument in Mr. Torres’s Report is “this practice gave its clients, Marketers,
an incremental impression of effectiveness of their Facebook marketing campaigns.
Marketers perceiving an incremental return of their spending on Facebook campaigns
were undoubtedly encouraged to allocate additional funds to these campaigns.”116 The
argument is that when a third-party website observed an increase in a social plugin
counter, they diverted the funds that they would have spent on incrementing the social
plugin counter towards Facebook advertising. However, this argument is flawed for at
least four reasons.
101.
First, as discussed above, many third-party websites do not have social plugin counters.
Second, among those third-party websites that have social plugin counters many do not
pay to advertise on Facebook. Indeed, much advice on social media emphasizes the
113
114
115
116
Torres Report ¶ 73.
Torres Depo. Tr. at 295:6-13 (“Q. And does your report assume that advertisers would have passed 100 percent
of their cost savings on to Facebook? A. Is that my assumption, that they would – Q. Yes. Is that your
assumption? A. No. Q. What is your assumption, then? A. That a fraction would have been converted.”).
Torres Depo. Tr. at 295:14-22 (“Q. Which fraction? A. I don’t have the information to determine that fraction.
W. Can you tell me it’s more than 50 percent? A. I can’t tell you, because I don’t’ have the information to
determine it.”).
Torres Report ¶ 68.
44
extent to which it is desirable often to not spend money on advertising.117 In his
deposition, Mr. Torres agreed that his definition of “Marketers” means that the focus is
on third-party websites who purchase advertising.118 However, the class definition
includes many URL messages where the website did not and would not spend money on
advertising on Facebook. Indeed, some examples from Named Plaintiff Mr. Hurley
include websites that explicitly do not spend money on external advertising, such as
.119 Similarly, Mr. Campbell shared URLs for government departments such as
.120 There are many
for the
reasons to think it unlikely that such government websites would be likely to divert
advertising dollars to Facebook.121
102.
Second, the mechanism by which Facebook allegedly benefited may in fact have had the
opposite effect. Mr. Torres argues that Marketers would have concluded that Facebook
marketing was more effective because of the incremental “Like” and devoted more
money to Facebook advertising.122 If the social plugin counter incremented without any
extra effort or expenditure on advertising from the firm itself, the firm may take this as
suggestive that its organic (or non-paid) marketing efforts were successful and be less
likely to divert money to advertising.
103.
Third, there are many reasons to think that website owners understood the varied
providence of “Likes” displayed on the social plugin counter, especially given that the
117
118
119
120
121
122
Edelman, David, and Brian Salsberg, “Beyond Paid Media: Marketing’s New Vocabulary,”
McKinsey&Company, November 2010,
http://www mckinsey.com/insights/marketing_sales/beyond_paid_media_marketings_new_vocabulary, viewed
January 11, 2016.
Torres Depo. Tr. at 98:2-8 (“Q. What do you mean by, marketers? A. In this report, I mean by marketers the
same thing that Facebook defines as marketers, which are their clients, the people responsible for advertising,
companies, entities, organizations, and whether they are direct entities or agencies in the advertising market.”).
See HURLEY000001 where the URL
was shared for
example.
See Plaintiff Matthew Campbell’s Corrected Objections and Responses to Defendant Facebook, Inc.’s First Set
of Interrogatories.
For example, the IRS itself imposes a long list of restrictions on potential advertisements that anyone connected
with the IRS can use. See “Advertising Standards,” IRS, last updated 07-Jan-2016,
https://www.irs.gov/uac/Advertising-Standards, viewed January 15, 2016.
Torres Report ¶¶ 68, 73.
45
instructions for installing the counter explicitly stated it would include “Likes” created
from URL attachments.123
104.
Without a bridge between the alleged “benefit” received by third-party website owner
and any alleged “benefit” to Facebook, Mr. Torres’s damage theory for the allegedly
inflated social plugin counter is divorced from the way that Plaintiffs described the harm
they suffered.
3.
The analysis fundamentally misunderstands or distorts why website
owners value “Likes”
a.
105.
The analysis focuses on the value of “Likes” that allowed a
continuing relationship between the website and an individual
rather than social plugin counters
By themselves “Likes” have little value to third-party websites. Recent research broadly
contradicts Mr. Torres’s assertion that “Likes can be profitable.”124 Harvard researchers
found in multiple experiments that “Liking” a brand has no effect on subsequent
consumer attitudes or behavior, including advertisement choice and actual purchase.125
Indeed, it appears likely that the study that Mr. Torres cites in Table 3 of his report126 does
not actually represent anything profitable that is causally connected with a “Like.” This
table compares the cost of inducing a “conversion” between a “Fan” and a non-”Fan” for
a variety of products.127 However, people who have a greater tendency to become a Fan of
a product are also easier to convert irrespective of whether they click a “Like” button.
There is no causal relationship implied by this data or profitability that can be attributed
to the “Like” button.
123
124
125
126
127
See FB000000163 from March 2011 (captured by the Wayback Machine) for an example of the text available
on Facebook’s developer website. The text explicitly says that the count includes “Likes” deriving from the
creation of URL attachments in messages. See also FB000000166 from October 2012 (also captured by the
Wayback Machine) with similar information.
Torres Report ¶ 70.
John, Leslie et al., “What are Facebook ‘Likes’ Really Worth?,” HBS Working Paper, 2015,
http://rady.ucsd.edu/docs/events/lesliejohn.pdf. This is also illustrated by the wide variety of motivations for
“Liking”, such as the desire to receive a discount or an offer, displayed in Table 1.
Torres Report at 26.
Note that this “Fan” language represents an earlier incarnation of Facebook, where users could be “Fans” of,
rather than “Like” an organization, so it is not quite certain how relevant it is for an analysis of “Likes” in any
case.
46
106.
Instead, the “value” of a “Like” to a third-party website or to a Facebook page is that it
enables that organization to form a relationship with that user and share communications
with them. Indeed, research shows128 that the only value of “Likes” to advertisers is that
they allow the user to subscribe to the conventional marketing communications put out
by that advertiser’s main Facebook page. This implies that the kind of “Like” that is an
anonymous increment of a social plugin counter, and that does not allow a website to
form a relationship with the user, has little worth. Therefore, trying to ascribe value to all
“Likes” based on valuations of “Likes” that allowed or implied a continuing relationship
between the organization and an individual is misguided.
107.
In general, Mr. Torres’s Report fails to distinguish between users actually clicking on
“Like” buttons on third-party websites with changes in the display of counters on those
third-party websites. For example, Mr. Torres cites an internal Facebook email chain for
the proposition that “from [the Like button’s] launch in April 2010, the impact of social
plugins was significant, generating 815 million clicks on ‘Like’ buttons daily in the first
few weeks.”129 However, the document indicates that Facebook’s partners had a wide
range of outcomes with respect to implementing social plugins – which are themselves
broader than a social plugin counter. For example, traffic on the Rotten Tomatoes movie
reviews website actually fell after implementing social plugins, suggesting that any
effects are not straightforward or uniform.130 Similarly, the document Mr. Torres uses to
demonstrate “Benefits of Using Like Button Plugins” conflates the potential for
anonymous incrementing of the social plugin counter with users clicking the “Like”
button.131
108.
Given this, any attempt to use a valuation for a “Like” that might include a meaningful
and ongoing relationship between the website and website user is wrong.
128
Mochon, Daniel, Karen Johnson, Janet Schwartz, and Dan Ariely, “How much is a like worth? A field
experiment of Facebook pages,” Tulane University Working Paper – Advances in Consumer Research, vol. 42,
2015. This paper is under the review process so is not publicly available.
129
Torres Report ¶ 29.
“Partners: social plugins,” Internal Facebook Email Chain, FB000011715.
“Connecting Outside of Facebook,” PowerPoint Presentation at Slide 4, FB000026793.
130
131
47
109.
Mr. Torres argues that “the average cost of advertising on Facebook to encourage a user
to become a Fan – Like the advertiser’s Facebook page – was $1.07. This cost also varies
across sectors and over time. In 2012, the cost per acquired Fan (i.e., cost per click in Fan
acquisition campaigns) averaged $0.55.”132
110.
There are four things to note about these estimates. First, they refer to “Fans,” not
“Likes.” Second, they refer to a situation where an organization will subsequently, as a
result of the Fan relationship, be able to communicate with that audience via the
Facebook platform and so do not reflect the market value of an anonymized +1 increase
in a plugin counter on a third-party website. Third, these estimates themselves show the
huge variability in potential estimates of the costs of obtaining a “Like” (which again, is
distinguishable from the anonymous incrementation at issue here). Indeed, there are
estimates that suggest a cost of obtaining a “Like” can via Facebook advertising is
$0.08.133 Estimates which range, depending on the study used, from $0.08 to $1.07 are not
a reliable guide for damages. Fourth, as shown in the earlier example of the promotion of
the BostonEventsInsider website shown in Figure 6, there are many other ways of
incentivizing users to give “Likes” which might even be cheaper than paying for them –
in that particular case, the website had not paid money for the movie tickets it was using
to incentivize customers to “Like” their website.
111.
It might be supposed that the estimates of “phony” purchases of “Likes” cited by Mr.
Torres, such as the case where “Likes” were sold for $0.075, are therefore more
relevant.134 However, there are at least two issues with such numbers. First, “Likes” are
often actually cheaper than the article cited.135 One website test suggests that “Likes” can
132
133
134
135
Torres Report ¶ 70.
Chieruzzi, Massimo, “Buying Facebook Likes Sucks, Here’s The Data To Prove It!,” AdEspresso, November
19, 2014, https://adespresso.com/academy/blog/buy-facebook-likes/, viewed December 12, 2015.
National Public Radio, Planet Money: “For $75, This Guy Will Sell You 1,000 Facebook ‘Likes,’” originally
broadcast on May 16, 2012, http://www npr.org/sections/money/2012/05/16/152736671/this-guy-will-sell-yousell-you-1-000-facebook-likes, viewed December 12, 2015.
For example, http://www.buylikesandfollowers.net/buy-facebook-likes-cheap html suggests that it would cost
$0.03 a “Like” if you buy 10,000 “Likes.” “Buy Real Facebook Likes,” Buylikesandfollowers.net,
http://www.buylikesandfollowers.net/buy-facebook-likes-cheap html, viewed December 12, 2015.
48
be bought as cheaply as $0.01. 136 Second, the market price of such “Likes” may reflect
the potential belief among buyers (whether true or not) that “Likes” might actually
translate into real people taking real actions. As such the price would be higher than for
an anonymous increment of the social plugin counter where there was definitely not such
a possibility.
b.
112.
The Proposed Methodology For Social Plugin Counters Does
Not Address The Fact That Many Proposed Class Members
Were Unaffected Or Benefited From These Practices.
Mr. Torres’s proposed methodology does not distinguish between the many cases where
the user was unaffected as there was no counter or social plug-in that displayed counts.
Indeed, it seems to presume the presence of a social plugin counter on the website for
every message where an attachment was created. However, many websites do not have
social plugins and many social plugins do not provide a counter.137
113.
Mr. Torres’s proposed methodology also does not consider the cases where a user was
invested in the website, meaning they would have welcomed or benefited from the
potential for an increment of the social plugin counter, supposing the website did indeed
have a plugin that contained the counter.
D.
114.
Mr. Torres’s two potential methodologies cannot be reconciled with each
other
Last, these two separate proposed methodologies cannot be reconciled with the different
claims that proposed class members may have. In particular, it is not clear how the
proposed methodology would avoid double-counting the benefits in instances where a
message contained a URL during the period that such a share could have potentially
incremented a social plugin displaying a counter. Mr. Torres has two competing
suggestions for how to resolve this issue.
115.
First, in his Report, Mr. Torres suggests: “the calculated effect from incremental
advertising revenue during the time when the Like counters were being affected (through
136
137
Chieruzzi, Massimo, “Buying Facebook Likes Sucks, Here’s The Data To Prove It!,” AdEspresso, November
19, 2014, https://adespresso.com/academy/blog/buy-facebook-likes/, viewed December 12, 2015.
Declaration of Alex Himel ¶¶ 34-35, 37.
49
December 2012) . . . shall be deducted from the benefits calculated for this period under
the methodology described in the previous section [the Social Graph method] for affected
Class Members.”138
116.
This proposal leads to conflicts in the interests of different putative class members. The
following thought experiment provides an example of possible conflicts, taking as given
that these methodologies are capable of producing concrete numbers and that the
numbers would be relevant.
117.
Suppose that between 2011 and 2015, 50 million URLs in messages were affected.
Suppose that in the first year of this period (2011-2012), 10 million URL messages were
affected. Suppose that the Social Graph method produced a calculation of 1 cent per
message-URL. Suppose also that the “Like”-counter valuation method produced a value
of five cents per message-URL in the 2011-2012 period. Under Mr. Torres’s Social
Graph method, the available damages to be split among class members would be
$500,000. Under the “Like”-counter valuation method, the available damages to be split
among affected class members would also be $500,000. However, under the
reconciliation proposal in Mr. Torres’s Report, that “Like”-counter total of $500,000
would need to be subtracted from the Social Graph method total of $500,000, implying
zero dollars available for any class members who sent messages containing URLs after
December 2012. Now that might be correct, given the negligible effects of the URL
counts after December 2012, but it does suggest a conflict of interest of the proposed
class members inherent in the two methodologies. Any proposed class member who sent
messages mainly prior to December 2012 would have an interest in maximizing the value
calculated by the “Like”-counter valuation method; any proposed class member who only
sent messages after December 2012 would prefer that the “Like”-counter valuation
method provided very low valuations.
118.
Second, in his deposition, Mr. Torres testified that ultimately his goal was to make sure
the overlap was taken into account and that “when everything is said and done . . . only
138
Torres Report ¶ 74.
50
one of the two calculations will prevail.”139 In response to a thought experiment similar to
the one in the previous paragraph, he said that ultimately, “you wouldn’t add them
together. You would just have one.”140 Similarly, Mr. Torres made clear that his
methodology could not give rise to a negative number because “if the overlap
overwhelms the situation, then only one of [the figures] would be appropriate.”141
119.
Mr. Torres’s suggested solution during his deposition is fundamentally different than the
solution proposed in his Report. Therefore, it is unclear how Mr. Torres would actually
reconcile his competing damages methodologies. Further, his testimony suggests that he
thinks that only one set of putative class members may recover and therefore, the
conflicts in the interests of different Class Members remain unresolved.
E.
120.
Rebuttal to Mr. Torres’s analysis as it pertains to statutory damages
I understand that the Court has discretion regarding whether to award statutory damages
and, if so, the amount. I also understand that the Court may consider several factors in
this determination including, among others, the actual damage to the victim and whether
the Defendant profited from the alleged violation. I have no opinion regarding whether
statutory damages are appropriate or not, but I note where my analysis and rebuttal to the
Torres report addresses these two factors. Mr. Torres explicitly stated in his deposition he
was not offering an opinion relating to statutory damages, so I emphasize that these are
139
140
141
Torres Depo. Tr. at 300:3-19 (“Q. But how would the net, if you are saying that you would deduct the amounts,
the analysis in this section shall be deducted from the benefits calculated under the methods described in the
previous section, okay, I’m saying, if the benefits were greater than the calculated – A. Now, what this means is
that . . . what this means is that the overlap has to be taken into account. That overlap can be calculated, when
everything is said and done, and that overlap means that only one of the two calculations will prevail. Q. One of
the two, meaning A or B? A. So, if you add A and B, you would then have to take away the overlap.”).
Torres Depo. Tr. at 299:4-8 (“So, if it were to be the case that benefits from one perspective are the same as the
benefits from the other perspective, then, yeah, the overlap with, would mean that you wouldn’t add them
together. You would just have one.”).
Torres Depo. Tr. at 299:10-23 (“Q. And what if the benefits were greater than the calculated effect from the
incremental advertising revenue? That would result in a negative number? A. In, it would be a very strange
hypothetical situation where that would even be the case, because of the length of the time period. Q. But, if it
were the case, it would be a negative number? A. So, whatever the methodology determines for those two
numbers would have to do the analysis of the overlap, and, if the overlap overwhelms the situation, then only
one of them would be appropriate.”).
51
not critiques of his conclusions but instead critiques of his analysis in terms of how it
informs statutory damages.142
1.
121.
Factor 1: Actual damage to the victim
As discussed in Section VI.C, according to their depositions, several proposed class
members that were deposed could not articulate the potential harms or losses they
suffered as a result of the challenged behavior. Furthermore, some Named plaintiffs
would have benefited directly, as in the case of Mr. Campbell when he shared a URL
, or indirectly. Mr. Torres has
not provided an opinion related to the actual loss, if any, suffered by proposed class
members.143 In general, given the discussion in Sections VI.A and VI.B, there are reasons
to believe that many potential class members suffered no damages or harm and in some
cases as actually benefited from the alleged behaviors.
2.
122.
Factor 2: Whether the Defendant profited from the alleged violation
Mr. Torres claims to measure the benefit received by Facebook from the challenged
behavior, but as I discuss in Sections VII.B and VII.C, he fails to do so. His first method
is based on the false assumption that Facebook uses information shared in URLs
contained Facebook message in the Social Graph for targeted advertising, which the
plaintiff’s own technical expert stated is not true.144 His second method is based on the
unsupported assumption that marketers may have shifted a fraction of their marketing
budget to additional Facebook advertising, but he has no idea what fraction it would be if
any. Therefore, for these and other reasons discussed above, neither of Mr. Torres’s
methods are a valid measure of Facebook’s profits resulting from the challenged
behavior.
142
Torres Depo. Tr., at 37:4-9 Q. And are you offering an opinion in this case as to whether or not statutory
damages should be awarded? A “No. That would be a legal conclusion”
143
Torres Depo Tr. at 48:11-21 (“Q. Why doesn’t it examine, your methodology examine, instead of examining
benefit to Facebook, why doesn’t it examine detriment to the putative class? A. So, my report and methodology
that I developed was asked to analyze the benefits to Facebook, so that’s, so, it doesn’t calculate the detriment
to the class members, or the potential class members, because it wasn’t meant to.”).
144
Golbeck Depo. Tr. at 223: 2-6 (“Q: You have information that some shares were used for recommendations but
not for that kind of targeted advertising? A: That’s right.”)
52
Disclaimer: Justia Dockets & Filings provides public litigation records from the federal appellate and district courts. These filings and docket sheets should not be considered findings of fact or liability, nor do they necessarily reflect the view of Justia.
Why Is My Information Online?