Kumandan et al v. Google LLC et al

Filing 453

ORDER RE MODIFIED SAMPLING PROTOCOL. Signed by Judge Susan van Keulen on 10/17/2024. (svklc1, COURT STAFF) (Filed on 10/17/2024)

1
2
3
4

UNITED STATES DISTRICT COURT

5

NORTHERN DISTRICT OF CALIFORNIA

6
7

IN RE GOOGLE ASSISTANT PRIVACY
LITIGATION.

Case No. 19-cv-04286-BLF (SVK)

8
ORDER RE MODIFIED SAMPLING
PROTOCOL

9
10

United States District Court
Northern District of California

11
12

Pursuant to Judge Freeman’s Order Regarding Motion for Relief from Non-Dispositive

13

Pretrial Order (Dkt. No. 370), the Court has revisited the scope of the October 20, 2022 Sampling

14

Order (Dkt. No. 331) in light of the Class Certification Order (Dkt. 360). Given that only the

15

Purchaser Class was certified to pursue breach of contract and Unfair Competition Law claims

16

predicated on breach of contract and a violation of California Business and Professions Code

17

§ 22576, the proportionality analysis has shifted dramatically since the Court issued the initial

18

Sampling Order.

19

As a result of these developments, the undersigned requested updated briefing regarding an

20

appropriate sampling protocol and held hearings on August 27 and October 3, 2024. See Dkt. 438,

21

443, 451, 452. Through this process, certain parameters on sampling were revised, however, the

22

Parties remained in vastly different universes as to sample size of queries, with Plaintiffs arguing

23

for a sample in excess of 20 million queries (see, e.g., Dkt. 452 at 1) and Google offering to

24

sample merely 12,500 queries. Dkt. 450 at 16:2-9. At the October 3 hearing, the Court

25

admonished both sides that substantial compromise would be required and, if after one final effort

26

the parties could not agree, the Court would determine the sample size. Id. at 44:15-19. Both

27

sides were amenable to a final effort and, if they could not agree, accepting a number set by the

28

Court. Id. at 43:15-44:3. The Parties’ further submissions reflected some, but not enough,

1

movement, with Plaintiffs reluctantly suggesting 2.6 million (Dkt. 452 at 2) and Google moving to

2

104,000 (Dkt. 451 at 1). Having reviewed the transcripts of the hearings on this issue, the Parties’

3

final submissions, and the declaration of Google’s expert Jonathan Borck (Dkt. 451-2), the Court

4

determines that a sample size of 260,000 queries is appropriate and proportional to the needs of

5

this case.

6

Accordingly, the Court ORDERS as follows:

7

1.

Google will produce a sample of speech log data within 45 days of the date of this

2.

Over the 26 quarters that the Parties have agreed are relevant, Google will perform

8
9
10

United States District Court
Northern District of California

11

Order.

sampling for 1 randomly selected day per quarter.
3.

For each sample day, Google will randomly select 10,000 Google Assistant

12

queries that were (1) hotword-initiated; and (2) made on Google Assistant Enabled Devices

13

manufactured by Google. The total number of queries to be sampled across all 26 quarters is

14

260,000.

15

4.

For the set of randomly sampled queries for each sample day, Google will provide

16

(1) the number of queries that did not contain a hot word in the “top hypothesis” field, and (2) the

17

number of those queries that were sent to a human reviewer.

18
19

5.

Google shall also provide Plaintiffs with the raw speech log data for all 104,000

queries, with the following modifications:

20

a.

Google may anonymize the raw data before producing it to Plaintiffs.

21

b.

Google may redact the content of each query, leaving only the hotword

22

visible. Where there is no hotword, the entire query will be redacted.

23

SO ORDERED.

24

Dated: October 17, 2024

25
26
SUSAN VAN KEULEN
United States Magistrate Judge

27
28
2