AMAZON WEB SERVICES, INC. v. USA

Filing 62

PUBLISHED OPINION with redactions. Signed by Judge Thomas C. Wheeler. (ss) Copy to parties.

Download PDF
In the United States Court of Federal Claims No. 13-506C (Filed Under Seal: October 31, 2013) (Reissued for Publication: November 8, 2013) 1 *************************************** * * AMAZON WEB SERVICES, INC., * * Plaintiff, * * v. * * THE UNITED STATES, * * Defendant, * * and * * IBM U.S. FEDERAL, * * Defendant-Intervenor. * * *************************************** * Post-Award Bid Protest; Cloud Computing Services Contract; Agency Corrective Action Following GAO Decision; No Prejudice to Party Lacking Any Chance of Contract Award; Reopening of Competition Unnecessary. Craig A. Holman, with whom were Kara L. Daniels, Lauren J. Schlanger, and Steffen Jacobsen, Arnold & Porter LLP, Washington, D.C., for Plaintiff Amazon Web Services, Inc. Steven M. Mager, Senior Trial Counsel, with whom were Stuart F. Delery, Assistant Attorney General, Jeanne E. Davidson, Director, and Kirk T. Manhardt, Assistant Director, Commercial Litigation Branch, Civil Division, U.S. Department of Justice, Washington, D.C., Edwin G. Doster, Avi M. Baldinger, and Arthur L. Passar, Office of General Counsel, Central Intelligence Agency, Of Counsel, for Defendant. Jason A. Carey, with whom were Thomas C. Papson, Luke W. Meier, John W. Sorrenti, and Katherine M. John, McKenna Long & Aldridge LLP, Washington, D.C., for Defendant-Intervenor IBM U.S. Federal. 1 The Court issued this opinion under seal on October 31, 2013, and gave the parties until November 7, 2013, to submit any proposed redactions of competition-sensitive, proprietary, confidential, or other protected information. The parties submitted their proposed redactions, which have been accepted by the Court. Redactions are indicated by [. . .]. OPINION AND ORDER WHEELER, Judge. Introduction This post-award bid protest arises from a competitive acquisition by the Central Intelligence Agency (“the agency”) for cloud computing services (“C2S”). At issue is the agency’s decision to take corrective action by inviting a new round of final proposal revisions from Plaintiff Amazon Web Services, Inc. (“AWS,” or “Amazon”) and Defendant-Intervenor IBM U.S. Federal (“IBM”). Previously, in its original evaluation of proposals, the agency had found AWS’s proposal to be far superior to IBM’s proposal, even though AWS had proposed a higher price. In a “best value” award decision, the agency determined that the higher price it would pay to AWS was justified. The agency awarded the C2S contract to AWS on February 14, 2013. After an agency debriefing, IBM filed a bid protest at the Government Accountability Office (“GAO”) on February 26, 2013 challenging various aspects of the agency’s procurement including the evaluation process. The agency initially stopped AWS’s contract performance pursuant to the automatic stay provisions of the Competition in Contracting Act, 31 U.S.C. § 3553(d)(3) (“CICA”), but later issued an override of the stay on March 15, 2003 to allow AWS’s performance to proceed. The GAO bid protest was sharply contested. After its initial protest, IBM filed three supplemental bid protests, and the parties submitted multiple legal briefs and comments during April and May 2013. The GAO conducted a lengthy evidentiary hearing on May 14, 2013 (505 transcript pages). On June 6, 2013, the GAO sustained IBM’s protest in part, and recommended that the agency take corrective action by reopening negotiations with offerors, amending the solicitation if necessary, and making a new award decision. IBM U.S. Federal, B407073.3 et al., 2013 CPD ¶ 142 (Comp. Gen. June 6, 2013). When the agency decided to follow the GAO’s decision and take the recommended corrective action, AWS filed suit in this Court on July 24, 2013. AWS asserted that the agency’s corrective action was overbroad, unreasonable, and in violation of federal law and regulation, and it asked the Court to grant declaratory and injunctive relief preventing the agency from considering revised final proposals. AWS also challenged the rationality of the underlying GAO decision. AWS contends that it handily won the competition with IBM, and that IBM suffered no prejudice and lacks standing because it has no substantial chance of receiving the contract award. Absent a legitimate, prejudicial procurement violation, AWS objects to competing again with IBM to win the same C2S contract, especially where so much information has been released to IBM during the debriefing process. 2 Defendant filed an extensive administrative record with the Court on August 7, 2013 (supplemented on September 9, 2013), consisting of 14 volumes and 13,048 pages. The administrative record contains the agency’s initial and second market survey documents, the solicitation documents including amendments, proposals, and agency evaluation records, the post-award bid protest documents, the contract implementation documents, the corrective action documents, and miscellaneous correspondence. Thereafter, the Court received the parties’ cross-motions for judgment on the administrative record, as well as response briefs and reply briefs. The Court heard oral argument on October 7, 2013. At the conclusion of the oral argument, due to the urgency and importance of the C2S contract, the Court issued a bench ruling in AWS’s favor that would allow the agency to go forward with the contract originally awarded to AWS on February 14, 2013. The Court outlined the reasons for its bench ruling, but stated that it would issue this formal opinion as promptly as possible. In assessing the reasonableness of the agency’s corrective action, the Court’s task is “to evaluate the rationality of the GAO’s decision.” Turner Constr. Co., Inc. v. United States, 645 F.3d 1377, 1383 (Fed. Cir. 2011). The Federal Circuit has observed that “an agency’s decision lacks a rational basis if it implements a GAO recommendation that is itself irrational.” Id. Central to the Court’s analysis here are fundamental principles of timeliness, standing, and prejudice. While the Court disagrees with the GAO’s substantive treatment of the discrete procurement issues presented, the essential finding underlying this decision is that the GAO completely overlooked the question of whether IBM suffered any prejudice and had standing to bring the protest in the first place. As a threshold matter, IBM lacked any chance of winning a competition with AWS for this C2S contract, and therefore IBM could not show any prejudice from either of the two grounds on which the GAO sustained IBM’s protest. The GAO’s decision does not even mention the existence of any “prejudice” to IBM, thus indicating that the GAO did not apply any “prejudice” requirement to IBM’s protest. Similarly, the GAO did not consider whether IBM had standing to bring the protest. If IBM did not have a chance of being awarded the contract, it did not have the necessary standing as an interested party to pursue its bid protest. See Labatt Food Serv., Inc. v. United States, 577 F.3d 1375, 1379-80 (Fed. Cir. 2009). Moreover, as the Court will explain, the GAO failed to address the way in which IBM manipulated its pricing to create a bid protest issue. IBM appears to have intentionally manufactured a protest argument relating to the Scenario 5 pricing requirement, which it hoped to pursue if it lost the C2S competition with AWS. Knowing full well from its pre-proposal questions what the Scenario 5 requirements were, IBM drastically departed from the approach followed in its initial proposal when it came to submitting its final proposal revision. If it did not win the award, IBM could argue that the agency did not evaluate Scenario 5 prices on a common basis. IBM was the only offeror who appeared to “misunderstand” the Scenario 5 pricing requirements, as 3 the other offerors all interpreted Scenario 5 in the same way. IBM strongly disagrees that it manipulated the procurement in this manner, but the Court does not see any other explanation for IBM’s final pricing strategy. The GAO made no mention of IBM’s manipulation of the procurement process, but instead sustained the protest allegedly for lack of a common basis to evaluate the offerors’ Scenario 5 prices. Even if the proposals from AWS and IBM presented a closer “best value” award decision, the Court could not justify rewarding IBM with another chance of competing for the C2S contract under these circumstances. The second ground for sustaining IBM’s protest relates to an alleged relaxation of requirements for AWS in waiving a clause certifying the absence of any virus in the software provided by subcontractors. The agency considered this clause redundant, and the Court agrees. As in the case of the Scenario 5 pricing issue, the Court sees no sound reason to afford corrective action through a new round of proposals where IBM suffered no prejudice and lacks standing. The agency conducted a proper procurement, and should be permitted to go forward with the contract previously awarded to AWS eight months ago. As a remedy, since the agency and AWS reportedly are now proceeding with the performance of the C2S contract, and have been complying with the Court’s October 7, 2013 bench ruling, declaratory relief should suffice here. Although Defendant asked for the entry of an injunctive relief order, the Court does not see a need for entry of an injunction. See generally PGBA, LLC v. United States, 389 F.3d 1219, 1224-27 (Fed. Cir. 2004) (discussing the discretion to issue declaratory or injunctive relief upon finding arbitrary and capricious agency action). The Court finds that the GAO’s decision recommending corrective action lacks a rational basis, and therefore that the agency’s decision to follow the GAO’s recommendation also lacks a rational basis. A full explanation of the Court’s decision is set forth below. A. Factual Background The agency issued a request for proposals (“RFP”) on June 15, 2012 in contemplation of awarding a single indefinite delivery, indefinite quantity (“IDIQ”) cloud computing contract. Administrative Record (“AR”) 1332, 1392, 1398. The agency issued the RFP after an extensive market survey involving communications with industry, and the circulation of a draft solicitation. In general, the RFP called for an established cloud service provider to build a custom clone of its public cloud for the agency. AR 1337-38. Although a number of companies expressed an interest in competing for this contract, and some of them submitted proposals, the agency’s evaluation ultimately came down to the proposals of two offerors, AWS and IBM. The agency found AWS to be the clear winner in this evaluation, and awarded a contract to AWS on February 14, 2013. AR 5090. The potential period of contract performance was ten years, with an initial 4 five-year ordering period, a three-year option, and a two-year option. AR 1332, 1392, 1400. The RFP provided that award would be made on a “best value” basis. AR 1458. Initially, the agency performed a pass/fail mandatory qualification evaluation to verify that an offeror was an established commercial cloud service provider with an existing, large-scale public offering. AR 1458-59. Following pass/fail evaluation, the agency applied four evaluation factors: (1) technical/management, including subfactors for technical approach (evaluated under a demonstration/oral presentation element and a written element), service level agreements, and management; (2) past performance; (3) security; and (4) price. AR 1459-63. Of these factors, price was “slightly less important than the other areas combined.” AR 1462. However, “as the relative difference in nonprice discriminators decreases,” price would become “more of a discriminator.” Id. The RFP also provided for an overall risk assessment rating to be assigned to each proposal. AR 1458; see also AR 1795 (amending the RFP on July 13, 2012 to formally add “risk” as a fifth evaluation factor to “be considered as part of the best value trade off determination”). The agency intended to evaluate price for completeness and reasonableness. Id. Offerors were required to submit a fixed price for task order 1 for program management to achieve initial operating capability, and a guaranteed minimum price for task order 2 for the provision of cloud services for the first year after initial operating capability. Id. Offerors also were required to provide a Cloud Services Catalog Price List containing fixed prices for various services. Id. The RFP contained six representative service-type ordering scenarios. AR 1793. Offerors were required to calculate the total costs of orders for each of these scenarios using their proposed catalog prices, on a yearly basis, for the base performance period. AR 1793-94. The total evaluated price was to be comprised of the sum of the task order 1 pricing and the prices for the six ordering scenarios for the base period. AR 1792. In the evaluation of the non-price evaluation factors, the agency deemed Amazon’s proposal superior to IBM’s proposal in every category except management, and except for “security” where each proposal received a “pass” rating. AR 5061. A summary of the agency’s evaluation of the Amazon and IBM proposals is contained in the following chart: 5 Evaluation Factor Technical/Management  Technical Approach (Demo)  Technical Approach (Written)  Service Level Agreements  Management Approach Past Performance (Confidence) Security Proposed Price Evaluated Price Guaranteed Minimum Overall Proposal Risk Amazon IBM Very Good Exceptional Very Good Satisfactory High Pass $149.06 million $148.06 million $25 million Low Marginal Very Good Satisfactory Very Good Moderate Pass $64.8 million $93.9 million $39 million High In the price evaluation, Amazon had a higher price, but the agency determined that Amazon’s technical proposal was sufficiently superior to IBM’s proposal to warrant a significant price premium. AR 5068. In the Source Selection Authority’s trade-off analysis, the SSA reached the conclusion that Amazon offered the best value to the Government, noting Amazon’s “superior overall approach, which will lower barriers to entry for [cloud computing] users and increase the likelihood of adoption.” Id. In the category of overall proposal risk, the agency rated Amazon as “low” and IBM as “high.” AR 5061. B. The Scenario 5 Pricing Issue Much of the controversy in this protest centers on the pricing of Scenario 5, one of the six hypothetical scenarios described in the price template. The RFP’s Scenario 5 requirement stated in part: This scenario centers around providing a hosting environment for applications which process vast amounts of information in parallel on large clusters (1000s of nodes) of commodity hardware in a reliable, fault-tolerant manner (MapReduce). The solution to this scenario should automatically provision clusters of compute for the segmentation and parallel processing of input datasets via the MapReduce framework (3.4.1) where the vendor is responsible for the management of the OS [operating system] and MapReduce implementation. Assume a cluster large enough to process 100TB [terabytes] of raw input data entirely on direct attached storage. Assume input data set was loaded from available object-based storage that realizes 6 reads/second and 2 writes/second. Assume 6 100% duty cycle on all virtual machines associated with this scenario. AR 1943. Prior to initial proposal submission, IBM had internal discussions regarding the meaning of the Scenario 5 requirements. AR 10513, GAO Hearing, Rhoades Test. (“[T]here were many [IBM] opinions of what scenario 5 meant.”). One IBM representative interpreted the 100% duty cycle instruction as requiring the data analytics tools to process continuously for a full year. AR 6300, IBM 6/26/12 email (“I don’t believe these can possibly be orders to run jobs . . . esp. with the requirement for 100% duty cycle.”). Another IBM representative disagreed, calling Scenario 5 “ambiguous.” Id., IBM 6/26/12 email (“The table is entitled ‘Cloud Services Prices’ so I think it is ambiguous.”). Due to the perceived uncertainty of the Scenario 5 requirements, IBM submitted questions to the agency prior to submitting its initial proposal. AR 1715-16. IBM inquired whether “orders” as used in the instructions meant “the number of new images of that scenario type” or the “number of instantiations/runs of the scenario type in the year.” AR 1715 (Question 49). The agency responded: “As outlined in the scenario, the servers should be treated as operating on 100% duty cycle and should be priced out as simultaneous orders.” Id. IBM also requested Scenario 5 specifications, including the data size and “anticipated average number of instantiations/runs of each scenario type (daily, monthly, etc.?).” AR 1716 (Question 50). The agency declined this request, reiterating its 100% duty cycle instruction, identifying the orders as simultaneous, and soliciting “commercial best practices.” Id. Although IBM was not satisfied with the agency’s answers, it did not seek further clarification or file a bid protest. Instead, IBM submitted its initial proposal with a Scenario 5 data analytics solution that processed 100 TB of data continuously throughout the year, at a price of approximately $[. . .]. AR 2310. The other offerors, including Amazon, interpreted Scenario 5 in the same way, as requiring continual data analytics deployment for a full year, or 8,760 hours. AR 2076-77, 2363, 2413. On October 24, 2012, the SSA established a competitive range of AWS, IBM, and one other offeror. AR 2991. Thereafter, the agency conducted written and oral discussions with these three offerors. AR Tabs 36-38, 41-43. During discussions with IBM, the agency identified each instance in which IBM failed to follow the pricing scenario directions. However, the evaluators did not identify any issue regarding IBM’s approach to Scenario 5, because IBM had offered a solution operating at a 100% duty cycle for 12 months, and IBM’s proposed price was consistent with the prices of the other offerors. AR 10388, Holloway Test.; AR 3120 (noting issues with IBM scenario pricing assumptions, but not Scenario 5). The agency thus effectively conveyed to IBM that it had correctly followed the Scenario 5 instructions. 7 On November 20, 2012, the agency issued RFP Amendment 4, calling for the submittal of final proposal revisions (“FPRs”). AR 1899. The agency permitted offerors to submit further questions prior to the FPR deadline, and while IBM did submit six questions, it did not raise any questions regarding Scenario 5. AR 1948-49. The agency then received FPRs on December 20, 2012 from the offerors within the competitive range. In the pricing portion of the FPRs, AWS and the third offeror consistently interpreted Scenario 5’s 100% duty cycle instruction as requiring continuous data analytic operations for the full year. AR 3817, 4532. IBM continued to price Scenarios 1 through 4, which also specified a 100% duty cycle, based on the continuous operation of the called-for servers for the year. AR 4443, 4445-47. However, IBM adopted a dramatically different interpretation for Scenario 5. In its FPR, IBM interpreted Scenario 5 as soliciting data analytics tools to perform a single 100 TB processing run, thereby slashing its Scenario 5 price from approximately $[. . .] to $[. . .]. AR 4449. There is no explanation in the record for this drastic IBM pricing change. The agency’s Price Evaluation Team (“PET”) evaluated the offerors’ FPRs for reasonableness, completeness, and risk in accordance with the RFP. AR 4594-4614. With regard to IBM’s Scenario 5 solution, the evaluators noted: Scenario was mis-priced. . . . Instead of following the directions which stated “Assume 100% duty cycle on all virtual machines associated with this scenario,” IBM calculated and proposed a cost for a single run through of a single 100TB data set. This resulted in a grossly under-priced dollar figure for scenario 5 of IBM’s proposal ($[. . .]/order). Once the Sponsor adjusted the price for 100% duty cycle over the entire year, using IBM’s optimal settings for completing each individual 100TB data set resulted in requiring 243 runs over the course of the year. Applying the catalog pricing to this adjustment results in a normalized cost of $[. . .] per order—which is consistent with [IBM’s] original proposal. AR 5014. The agency’s calculation using IBM’s catalog pricing and Scenario 5 technical solution yielded a price of approximately $[. . .], nearly the same as IBM’s Scenario 5 price in its initial proposal. AWS’s total evaluated price was $148,061,628 and IBM’s total evaluated price was $93,917,785. AR 5067. However, there were two pricing nuances that the agency thought were significant. First, IBM proposed a guaranteed minimum of $39 million, as compared to AWS’s guaranteed minimum of $25 million. AR 5061. IBM’s guaranteed 8 minimum was nearly double the anticipated Year 1 amount for IBM’s services, meaning that the agency likely would need to make a large year-end payment to IBM. AR 5069. Second, the agency noted that IBM’s proposed contract terms would allow IBM to request restructuring of the entire agreement after Year 2 if the service price in that year did not exceed the guaranteed minimum. AR 4827-28, 5069. These terms allowed IBM to propose a low price for the agency’s proposal evaluation purposes, but then to argue for negotiation of a higher price in the later years of performance. The agency’s SSA performed a best value trade-off analysis and selected AWS for award of the contract. The agency and AWS executed the contract document on February 14, 2013. Analysis A. Standard of Review In bid protest cases, courts review agency actions under the “arbitrary and capricious” standard. See 28 U.S.C. § 1491(b)(4) (adopting the standard of review “set forth in [5 U.S.C. § 706]”). This standard “is highly deferential” and “requires a reviewing court to sustain an agency action evincing rational reasoning and consideration of relevant factors.” Advanced Data Concepts, Inc. v. United States, 216 F.3d 1054, 1058 (Fed. Cir. 2000) (citing Bowman Transp., Inc. v. Arkansas-Best Freight Sys., Inc., 419 U.S. 281, 285 (1974)). Thus, as long as there is “a reasonable basis for the agency’s action, the court should stay its hand even though it might, as an original proposition, have reached a different conclusion.” Honeywell, Inc. v. United States, 870 F.2d 644, 648 (Fed. Cir. 1989). If, however, “the agency entirely fail[s] to consider an important aspect of the problem[ or] offer[s] an explanation for its decision that runs counter to the evidence before the agency,” then the resulting action lacks a rational basis and, therefore, is defined as “arbitrary and capricious.” Ala. Aircraft Indus., Inc.-Birmingham v. United States, 586 F.3d 1372, 1375 (Fed. Cir. 2009) (quoting Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983)) (internal quotation marks omitted). Where the issue is “an agency’s decision to follow a GAO recommendation, . . . [the] agency’s decision lacks a rational basis if it implements a GAO recommendation that is itself irrational.” Turner, 645 F.3d at 1383 (citing Centech Grp., Inc. v. United States, 554 F.3d 1029, 1039 (Fed. Cir. 2009); Honeywell, 870 F.2d at 648). Thus, “the controlling inquiry is whether the GAO’s decision was a rational one.” Id. at 1384 (alteration and internal quotation marks omitted). 9 B. IBM’s Lack of Standing and Failure to Show Prejudice “[B]ecause the question of prejudice goes directly to the question of standing, the prejudice issue must be reached before addressing the merits.” Labatt, 577 F.3d at 1378 (internal quotation marks omitted). A bid protestor has been prejudiced when it can show that, “but for [a significant error in the procurement process], it would have had a substantial chance of securing the contract.” Id. In deciding whether a protestor would have had a substantial chance of securing the contract, it is necessary to show proper deference to the views of the procuring agency, for “[i]t is well settled that COs are given broad discretion in their evaluation of bids. When an officer’s decision is reasonable, neither a court nor the GAO may substitute its judgment for that of the agency.” Turner, 645 F.3d at 1383 (citation omitted). “De minimis errors in the procurement process do not justify relief,” and “[t]he protestor bears the burden of proving that a significant error marred the procurement in question.” Glenn Def. Marine (Asia), Pte Ltd. v. United States, 720 F.3d 901, 907 (Fed. Cir. 2013). This burden “is greater in negotiated procurement, as here, than in other types of bid protests because ‘the contracting officer is entrusted with a relatively high degree of discretion.’” Id. (quoting Galen Med. Assocs., Inc. v. United States, 369 F.3d 1324, 1330 (Fed. Cir. 2004)). Contracting officers are afforded “an even greater degree of discretion when the award is determined based on the best value to the agency.” Id. at 908. This is just such a case. Here, based on the Source Selection Evaluation Team’s ratings, the Source Selection Authority determined that AWS “clearly” offered the best value: Amazon’s proposal contained a number of unique, differentiating capabilities that are considered highly advantageous to the Government. In several areas they exceeded the government’s requirements, providing enhanced capabilities and an overall superior technical solution. . . . Taking the significant technical advantages of Amazon’s proposal, a tradeoff analysis was performed. Amazon’s price is $148,061,628, while IBM’s is $93,917,785 . . . . I do not believe that the $54 million difference, over five years, outweighs Amazon’s strengths—specifically their superior technical solution. The additional cost to the Government of awarding to Amazon is justified by their proposed superior overall approach, which will lower barriers to entry for C2S users and increase the likelihood of customer adoption. . . . 10 Although not a basis for my best value tradeoff decision . . . , I note for the record that the price risk in IBM’s proposal likely overstates the overall price difference between Amazon and IBM. IBM’s price proposal has two factors that contribute to additional price risk, and may result in additional costs to the Sponsor: • The proposed ‘guaranteed minimum’ is more than double their expected year 1 prices and as a result the Government is unlikely to see the benefits of the proposed low catalog prices. • The contract terms and conditions indicate that IBM will seek to restructure the contractual agreement in Years 2+ if the service price in that year does not exceed the guaranteed minimum. .... Being mindful of the fact that non-price factors are only slightly more important than price factors, I found the above described advantages of Amazon’s proposal to be well worth the price premium over IBM’s proposal and clearly the best value. AR 5068-69. In sum, AWS’s offer was superior in virtually every way but price, and IBM’s advantage in that area was likely not as great as IBM attempted to make it appear. Nevertheless, the GAO sustained IBM’s protest on two grounds: (1) the agency’s Scenario 5 price evaluation lacked a common basis and was therefore unreasonable; and (2) the agency materially relaxed a solicitation requirement for AWS, but not for the other offerors. See AR 10706. Regarding the first ground, the GAO makes no mention of prejudice whatsoever, despite its being raised and argued by both AWS and the agency. Regarding the second, the GAO notes—without any explanation—that “[i]n [its] view, IBM’s assertion that the level of risk to the contractor was reduced by this modification is sufficient to establish prejudice.” AR 10712 n.4, IBM U.S. Federal, 2013 CPD ¶ 142. In neither instance is there any consideration of the proper legal standard for prejudice, nor is there any evidence that IBM met its burden of proof for establishing such prejudice. Consequently, there is no justification for even reaching the merits of IBM’s protest. Regarding the Scenario 5 price evaluation, the GAO found that, unlike IBM’s solution, “there was no way” to ascertain the speed of AWS’s solution, and therefore 11 “there [was] no basis for concluding that Amazon was evaluated for scenario 5 using the same or otherwise comparable level of performance as included in IBM’s adjusted price.” AR 10709-10. However, at the GAO hearing, the agency’s C2S experts testified that speed was purposely not specified as a performance metric in Scenario 5 because performance depended on a wide variety of factors. M. Jason Holloway, the chair of the agency’s Technical/Management Evaluation Team and advisor to its Price Evaluation Team, explained that “Scenario 5 is a platform service that is very specific to each offeror’s individual capabilities,” and many variables could influence the approach any one offeror might take. AR 10430-31, Holloway Test. Because the agency was “interested in taking advantage of the expertise that [the offerors] could provide,” it intentionally left performance characteristics, such as speed, unspecified. AR 10372-74. Instead, the agency stated only the function to be performed and allowed each offeror the flexibility to propose and price its best commercial practice. Mr. Holloway explained: We have to validate that [the offerors] followed commercial best practices, but we did not confine them with how they did that. . . . That’s just one characteristic of what would affect performance . . . , just one of many characteristics. We can’t—there’s no way that we could potentially list all the technical specifications in order to have the best solution capable, specifically with a different technical functionalities of each of the diverse set of platform services . . . the offerors provided. AR 10437, Holloway Test. Thus, once the agency validated each of the proposed Scenario 5 solutions for reasonableness—“[a]nd each of the offerors met that requirement,” AR 10436, Holloway Test.—it compared those solutions based on the same duty cycle (100%) and duration (one year). Given the agency’s emphasis on commercial best practices and recognition of a variety of performance metrics, it is impossible to see how the agency’s decision not to reduce the Scenario 5 comparison to a simple price-to-speed calculation prejudiced IBM. Regarding the material relaxation of a solicitation requirement, the GAO accepted IBM’s assertion that the result of this relaxation was “sufficient to establish prejudice.” AR 10712 n.4, IBM U.S. Federal, 2013 CPD ¶ 142. In rejecting AWS’s counterargument, the GAO stated, “Our conclusion is not changed by Amazon’s assertion that IBM also sought numerous proposed changes to provisions in the RFP, including a proposal that it would not provide warranties with respect to third party software.” Id. The problem with the GAO’s conclusion is it ignores the relevant rule that if an agency’s “improper deviation from the solicitation” equally affects all offerors, then it causes prejudice to none. Labatt, 577 F.3d at 1380. Naturally, because AWS was awarded the contract, it was the only offeror to engage in post-solicitation negotiations, 12 but prior to the contract award, IBM had proposed negotiating the very same modification. See, e.g., AR 9336 (“Sponsor . . . receives no warranties, indemnities or express or implied patent or other license from IBM with respect to any third party software.”). In other words, AWS’s successful post-solicitation modification had no effect on IBM’s ability to pursue the same result, and therefore no effect on IBM’s proposal or the evaluation of that proposal. Moreover, the agency’s later removal of the requirement at issue because “it [was] redundant to other RFP requirements,” AR 12689, further emphasizes that even if an error was made, its effect was not prejudicial. The bottom line is that IBM did not lose the competition because of the Scenario 5 price evaluation or AWS’s post-solicitation negotiations, but because of the overall inferiority of its proposal. This proposal contained numerous weaknesses, including some “significant” weaknesses, a technical deficiency, and an overall high risk rating. AR 4638-69; AR 5065-67 (describing “multiple weaknesses,” a technical “deficiency,” and “multiple concerns” creating a high price risk). For example, the Technical/Management Evaluation Team determined that IBM “[did] not demonstrate the capability to auto-scale all required services” and therefore “fail[ed] to meet the [autoscaling] requirement.” AR 4644-45. This inability was “in direct conflict with the [agency’s] C2S goal ‘to deliver scalable, balanced, and fault tolerant solutions,’” and thus was deemed “unacceptable.” AR 4645. Although IBM protested that rating, the GAO denied this part of IBM’s protest, stating, “We see no basis to question the reasonableness of the agency’s concerns (expressed as a deficiency and a significant weakness under the technical approach subfactor) . . . .” AR 10715. The GAO also affirmed the agency’s consideration of IBM’s guaranteed minimum price, which contributed to the overall high risk rating. See AR 10717. “[S]tanding is a threshold jurisdictional issue,” and “prejudice (or injury) is a necessary element of standing.” Myers Investigative & Sec. Servs., Inc. v. United States, 275 F.3d 1366, 1369 (Fed. Cir. 2002). To establish prejudice, IBM had to “show that there was a ‘substantial chance’ it would have received the contract award but for the alleged error in the procurement process.” Info. Tech. & Applications Corp. v. United States, 316 F.3d 1312, 1319 (Fed. Cir. 2003) (quoting Alfa Laval Separation, Inc. v. United States, 175 F.3d 1365, 1367 (Fed. Cir. 1999)). IBM failed to make such a showing, and the GAO failed to make relevant findings or apply the proper legal standards. In fact, other than the GAO’s unexplained acceptance of IBM’s speculation that it had suffered prejudice, see AR 10712 n.4, IBM U.S. Federal, 2013 CPD ¶ 142, the GAO made no mention of prejudice to IBM at all. Such a “fail[ure] to consider an important aspect of the problem” is, by itself, sufficient to render the GAO’s decision arbitrary and capricious. Ala. Aircraft, 586 F.3d at 1375 (quoting Motor Vehicle Mfrs. Ass’n., 463 U.S. at 43). 13 C. Timeliness Issues in IBM’s Protest Timeliness, like prejudice, is a threshold issue that must be addressed prior to reaching the merits of a bid protest. See, e.g., Goel Services, Inc., B-310822.2, 2008 CPD ¶ 99 (Comp. Gen. May 23, 2008) (“Our timeliness rules reflect the dual requirements of giving parties a fair opportunity to present their cases and resolving protests expeditiously without unduly disrupting or delaying the procurement process. In order to prevent these rules from becoming meaningless, exceptions are strictly construed and rarely used.” (citation omitted)). Where, as here, an offeror misses its opportunity to fairly challenge the terms of a solicitation, it cannot then be allowed to avoid the timeliness bar by mischaracterizing its case as an evaluation challenge. See Blue & Gold Fleet, L.P. v. United States, 492 F.3d 1308, 1313 (Fed. Cir. 2007). 1. Scenario 5 “As an initial matter, [the GAO found] untimely IBM’s challenge to the agency’s interpretation that scenario 5 called for repeated 100 TB data runs throughout the year, rather than a single run under each order.” AR 10707. The GAO explained: Under our Bid Protest Regulations, a solicitation defect apparent on the face of the solicitation must be protested prior to the time set for receipt of initial proposals or quotations, when it is most practicable to take effective action against such defects. Furthermore an offeror who chooses to compete under a patently ambiguous solicitation does so at its own peril, and cannot later complain when the agency proceeds in a way inconsistent with one of the possible interpretations. Here, by its own actions, IBM evidenced its recognition that the scenario 5 instructions were ambiguous as to the frequency of the expected 100 TB data runs. IBM requested clarification of the requirements in this regard and then, not having received meaningful clarification, first adopted one interpretation (that is, continual runs) in its initial proposal and then a different interpretation (a single run per order) in its FPR. Having chosen to compete despite its recognition of the patently ambiguous nature of the solicitation in this area, IBM cannot now complain when the agency proceeds in a manner inconsistent with one of the possible interpretations and adjusts IBM’s price to match the government’s (and Amazon’s apparent) interpretation of the requirement. 14 AR 10708 (citations omitted). This portion of the GAO’s analysis was correct, and the analysis should have ended there. As the Federal Circuit has explained: In the absence of a waiver rule, a contractor with knowledge of a solicitation defect could choose to stay silent when submitting its first proposal. If its first proposal loses to another bidder, the contractor could then come forward with the defect to restart the bidding process, perhaps with increased knowledge of its competitors. A waiver rule thus prevents contractors from taking advantage of the government and other bidders, and avoids costly after-the-fact litigation. Blue & Gold Fleet, 492 F.3d at 1314 (applying the patent ambiguity doctrine to postaward bid protests). This timeliness rule aptly describes what IBM attempted to do. Nevertheless, the GAO concluded that the agency’s “price evaluation” method was unreasonable and sustained IBM’s protest on that ground. AR 10707-10. This conclusion, however, completely ignored the blatant manner in which IBM manipulated the situation to its advantage. Prior to the submission of its initial proposal, IBM had an internal disagreement regarding the meaning of Scenario 5. See, e.g., AR 10513, Rhoades Test. (stating that “there were many [IBM] opinions of what scenario 5 meant because of the flexibility of the directions”). One IBM architect understood the 100% duty cycle instruction to mean continuous processing for a full year, while another disagreed, calling Scenario 5 “ambiguous.” AR 6300. Consequently, IBM asked the agency several questions about Scenario 5 before submitting its initial proposal. AR 1715-16. The agency’s answer was simply that “the servers should be treated as operating on 100% duty cycle and should be priced out as simultaneous orders” and that “the contractor should propose commercial best practices.” AR 1715-16. Thereafter, IBM submitted its initial proposal with a data analytics solution based on processing 100 TB of data continuously throughout the year. AR 2310-11. The other three offerors submitted proposals that interpreted Scenario 5 precisely the same way. See AR 2076-77, 2363, 2413. After reducing the competitive range to AWS, IBM, and a third offeror, the agency conducted written and oral discussions. AR Tabs 36-38, 41-43. During these discussions, the agency identified questions and concerns about the offerors’ initial proposals. AR 10381-82. The agency did not identify any concerns with IBM’s initial approach to Scenario 5, which the agency found “consistent” with the other proposals, AR 10388, Holloway Test., thereby effectively conveying to IBM that it had correctly followed the Scenario 5 instructions. 15 Then, in its FPR, IBM suddenly deviated from its initial Scenario 5 proposal. IBM changed its solution from one that processed 100 TB of data continuously throughout the year to one that performed a single 100 TB processing run in a 36-hour period, thereby reducing its price from approximately $[. . .] to $[. . .]. AR 4449. Because IBM was the only offeror that deviated from its initial proposal in such a manner, the agency extended IBM’s FPR Scenario 5 solution from a single run to a full year. See AR 5066 (stating that “[n]ormalization was required because the Offerors did not uniformly follow the instructions”). The result was a corrected Scenario 5 price of approximately $[. . .], which was “consistent with [IBM’s] original proposal.” AR 5014. Given the absence of a rational alternative explanation, it is obvious that when IBM deviated from its initial approach, it did so as a way to manipulate the situation in its favor. If the agency accepted IBM’s “gross[ly] under-priced dollar figure for Scenario 5,” AR 5014, then IBM would gain almost a $[. . .] advantage over its properly priced solution. If, however, the agency normalized IBM’s price (and IBM lost the competition), then it could protest such normalization by arguing the lack of a common basis, whether it be speed or some other intentionally unspecified performance metric. When the agency made the latter choice, IBM filed its protest and feigned ignorance of the reasons for the agency’s actions, pretending not to understand why “the agency determined that the solution should be available throughout the year.” AR 5961-62. In reality, IBM was well aware that the agency had made this determination long before its final price evaluation. Such gamesmanship undermines the integrity of the procurement process and should not be rewarded with circumvention of the timeliness requirement. 2. Post-Selection Negotiation IBM’s post-selection challenge to AWS’s revision of RFP Commercial Clause § 152.204-706(a) was also untimely. This clause provided for each offeror to “certif[y] that it will undertake to ensure that any software to be provided . . . under [the] contract will be provided . . . free from computer virus.” AR 1409. During post-selection negotiations, AWS proposed, and the agency accepted, a revision stating that “[u]nder AWS Terms and Conditions, only software developed and provided by AWS would be subject to this requirement.” AR 5074. IBM protested, arguing that “[a]voiding the requirement to certify third party software and returned government software reduces the burden of compliance, and the risk of non-compliance,” AR 9606, and the GAO agreed, AR 10713. The challenged actions, however, were clearly contemplated by the RFP, which provided: [T]he offeror’s solution may include commercial license/terms and conditions that are customarily included 16 within their commercial transactions. The offeror shall propose any such language within this section for the Government to review for potential inclusion within this acquisition. AR 1787; see also AR 1778 (stating the Government’s intent to “select an offeror for final negotiations”). The agency then specified five types of clauses that it excluded from this negotiation provision, leaving § 152.204-706(a) among the terms and conditions that could be revised. AR 1787; see also AR 10562, Ross K. Test. (stating that the agency permitted offerors to propose their own commercial terms and conditions). Finally, during discussions, the agency specifically advised IBM that the agency “reserve[d] the right to negotiate all commercial terms and conditions, if selected for award, under the select to negotiate process.” AR 3121. These rules were made explicitly clear to IBM, but IBM did not challenge them prior to the award. On the contrary, IBM itself took advantage of the opportunity to propose revised terms and conditions, even proposing terms limiting its own responsibility for third-party software, see AR 9336, 9343, 9345-46, then challenging this approach after the fact. This tactic of “rolling the dice” to see if it could receive the award “and then, if unsuccessful, claim[ing] the solicitation was infirm,” is simply not allowed. Blue & Gold Fleet, 492 F.3d at 1314 (quoting Argencord Mach. & Equip., Inc. v. United States, 68 Fed. Cl. 167, 175 (2005)). “[W]here there is a ‘deficiency or problem in a solicitation . . . the proper procedure for the offeror to follow is not to wait to see if it is the successful offeror before deciding whether to challenge the procurement, but rather to raise the objection in a timely fashion.’” Id. (quoting N.C. Div. of Servs. for the Blind v. United States, 53 Fed. Cl. 147, 165 (2002)). If an offeror fails to do so, it cannot then circumvent the timeliness requirement by recasting its challenge to the terms of the solicitation as a challenge to the evaluation of the proposals. Id. at 1313. IBM’s objection was not raised in a timely fashion and therefore should have been barred. As with the timeliness of the Scenario 5 challenge, the GAO’s failure to consider this threshold matter renders its decision irrational. See Ala. Aircraft, 586 F.3d at 1375. D. Overbreadth of Corrective Action Although contracting officers are given “broad discretion to take corrective action where the agency determines that such action is necessary to ensure fair and impartial competition,” DGS Contract Serv., Inc. v. United States, 43 Fed. Cl. 227, 238 (1999) (quoting Rockville Mailing Serv., Inc., B-270161, 96-1 CPD ¶ 184 at 3 (Comp. Gen. Apr. 10, 1996)) (internal quotation marks omitted), such corrective action must be “reasonable under the circumstances and appropriate to remedy the impropriety,” Reema Consulting Servs., Inc. v. United States, 107 Fed. Cl. 519, 527 (2012) (internal quotation marks omitted). Therefore, even where a protest is justified, any corrective action must 17 narrowly target the defects it is intended to remedy. Sheridan Corp. v. United States, 95 Fed. Cl. 141, 153 (2010). In this case, the GAO “recommend[ed] that the agency reopen the competition and amend the RFP as necessary” based on its identification of two discrete defects, AR 10717, and the agency followed that recommendation, AR 12679. Neither defect, however, warranted reopening the entire competitive process. With respect to the first defect, the GAO identified a price evaluation error affecting only Scenario 5: “the agency’s uncertainty regarding just what performance (e.g., number of 100 TB data runs) was included in each evaluated price.” AR 10710. Given the narrowness of this finding, reopening the competition to include unaffected scenarios and proposal areas would be overbroad. See Sheridan, 95 Fed. Cl. at 153 (“[T]his Court has rejected corrective action to resolicit proposals because of a perceived evaluation error.” (citing Delaney Constr. Corp. v. United States, 56 Fed. Cl. 470, 476 (2003); MCII Generator & Elec., Inc. v. United States, No. 1:02-CV-85, 2002 WL 32126244 (Fed. Cl. Mar. 18, 2002))). Instead, appropriate corrective action would be limited to a revised Scenario 5 price evaluation. Second, the GAO found that “waiving a material term of the solicitation for one of [the offerors], after the selection decision was made, was improper.” AR 10712 n.4. As above, targeted correction of this defect would not require reopening the entire competition, but only addressing the affected aspects of the offerors’ proposals. Nonetheless, the agency did not stop with merely reopening the competition. On the contrary, in soliciting new offers from AWS and IBM, the agency “elected to use this as an opportunity to amend other aspects of the solicitation,” AR 12686, despite the fact that no “other aspects” were at issue. Such corrective action is inherently overbroad. When flaws occur during the evaluation of properly submitted proposals, “a reevaluation of the proposals may be warranted, but a resolicitation of the proposals compromises the integrity of the procurement system, especially where the winning price has been disclosed.” Sheridan, 95 Fed. Cl. at 154. In this case, considerable information regarding the competition and the agency’s evaluation of AWS has been disclosed to IBM, including a 45-page debriefing (AR Tab 62) and a lengthy question and answer session (AR Tab 63). Reopening the entire competition under these circumstances would lack a rational basis and undermine the integrity of the procurement process. In short, the GAO’s recommendation was irrational because it was not narrowly tailored to address discrete procurement defects, and, as a result, the agency’s decision to follow that recommendation was likewise irrational. E. Appropriate Relief to Amazon The Tucker Act grants the Court broad discretion to “award any relief that the court considers proper, including declaratory and injunctive relief.” 28 U.S.C. 18 § 1491(b)(2); see also PGBA, 389 F.3d at 1223 (“We give deference to the Court of Federal Claims’ decision to grant or deny injunctive relief, only disturbing the court’s decision if it abused its discretion.”). In deciding whether a permanent injunction is proper, a court considers “(1) whether, as it must, the plaintiff has succeeded on the merits of the case; (2) whether the plaintiff will suffer irreparable harm if the court withholds injunctive relief; (3) whether the balance of hardships to the respective parties favors the grant of injunctive relief; and (4) whether it is in the public interest to grant injunctive relief.” Id. at 1228-29. Because the Court issued a bench ruling on October 7, 2013, allowing the agency and AWS to continue performance of the C2S contract, it is unnecessary to grant injunctive relief in this instance. Specifically, regarding the second factor, it is clear that an absence of injunctive relief will not cause AWS to suffer irreparable harm. AWS has received the contract award and is now performing under that contract. Consequently, the Court finds that declaratory relief is sufficient and proper. Conclusion There is no such thing as a perfect procurement. Thus, a bid protestor must show prejudice, not mere error, for “[n]ot every error compels the rejection of an award.” Grumman Data Sys. Corp. v. Dalton, 88 F.3d 990, 1000 (Fed. Cir. 1996). Rather, it is “the significance of errors in the procurement process [that determines] whether the overturning of an award is appropriate,” and it is the protestor who “bears the burden of proving error in the procurement process sufficient to justify relief.” Id. IBM never met that burden, and the GAO neglected to address it. Even if IBM’s arguments regarding the price evaluation and modified solicitation requirement were persuasive, it remains implausible that there would be any effect on the outcome of the procurement. AWS’s offer was superior, and the outcome of the competition was not even close. Indeed, if there has been any prejudice in this process, it has been to AWS, for improper corrective action in the form of reopening competition is not harmless. The unfairness inherent in such an action is that the winner must resubmit a new proposal with the information from its original offer already disclosed. In effect, AWS would have to bid against its own winning proposal. This Court will not allow such an unjust result. For the reasons set forth above, Plaintiff’s motion for judgment on the administrative record is GRANTED. Defendant’s motion for judgment on the administrative record is DENIED, and Defendant-Intervenor’s motion for judgment on the administrative record is DENIED. 19 IT IS SO ORDERED. s/ Thomas C. Wheeler THOMAS C. WHEELER Judge 20

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?