OPINION
CLIFTON, Circuit Judge:
Prabhat Goyal, former chief financial officer of Network Associates, Inc. (“NAI”), appeals from his convictions on fifteen counts of securities fraud and making materially false statements to auditors. The government alleged that NAI, under Goyal’s supervision, violated generally accepted accounting principles (“GAAP”) by recognizing revenue from certain software sales earlier than it should have. Goyal was indicted for concealing the allegedly improper accounting from NAI’s outside auditors and for filing reports with the Securities and Exchange Commission that, because of NAI’s accounting, allegedly misstated revenue in certain reporting periods between 1998 and 2000. Goyal argues that no jury could have found him guilty beyond a reasonable doubt, as the jury below did, based on the evidence the prosecution presented at trial. We agree, and we reverse his convictions on all counts.
I. Background
From approximately 1997 to 2001, Goyal was chief financial officer of NAI. NAI, formerly known as McAfee, was and remains a major vendor of antivirus and network security software.
Before 1998, NAI used a “direct sales” business model, meaning that it primarily sold its software directly to end users. In 1998, the company added a “distribution channel” model, selling products through distribution companies. These distributors in turn sold NAI’s software to retail stores that resold the software to end users.
The prosecution’s case against Goyal challenged the accounting method that NAI used, under Goyal’s supervision as CFO, to recognize revenue from sales to its largest domestic distributor, Ingram Micro. In particular, the government took issue with the accounting method NAI used to recognize large sales that it made to Ingram at the end of financial quarters between 1998 and 2000. Following a practice common in the software industry, which the government did not contend was [914]*914illegal, NAI negotiated significant quarter-end deals with Ingram, called “buy-in” transactions, to help meet its quarterly revenue projections.1 To close these sales, NAI granted Ingram substantial discounts, rebates, and other favorable sales terms. One enticement that NAI offered Ingram in the last two quarters of 1998 and the first quarter of 1999, was a guarantee that its wholly owned subsidiary, NetTools, would repurchase unsold product from Ingram in specified amounts. NetTools would then sell the repurchased product to customers.2
The government did not contend that any of the sales concessions that NAI gave Ingram in the buy-in deals were improper or that NAI claimed revenue that it never earned. Rather, the government objected to the timing of NAI’s recognition of revenue from these deals. The government maintained that NAI violated GAAP by using “sell-in” accounting to recognize revenue from these deals earlier than it should have and thereby overstated its revenue. Under sell-in accounting, a manufacturer like NAI recognizes revenue when it ships products to its distributors (i.e., “sells in” to the distribution channel). The manufacturer must estimate the amount of future rebates, discounts or returns and then reduce its stated revenue by this amount.
By contrast, a company using “sell-through” accounting recognizes revenue when its distributors sell the product to a reseller (i.e., “sells through” the distribution channel). Sell-through accounting recognizes revenue later than sell-in accounting does and nets out rebates, discounts, and returns. Thus the manufacturer does not need to estimate their effect on its revenue.
The jury convicted Goyal of one count of securities fraud and seven counts of making false filings with the SEC (collectively, “the securities counts”),3 and seven counts of making materially false statements to NAI’s auditors at PricewaterhouseCoopers (“PwC”) (the “lying-to-auditors counts”).4
After Goyal’s conviction, the district court denied his motions for judgment of acquittal and for a new trial. Goyal appealed.
II. Discussion
We review de novo the district court’s denial of Goyal’s Rule 29 motion for judgment of acquittal. See United States v. Mosley, 465 F.3d 412, 415 (9th Cir.2006). We must decide “whether ‘after viewing the evidence in the light most favorable to the prosecution, any rational trier of fact could have found the essential elements of the crime[s] beyond a reasonable doubt.’ ” United States v. Nevils, 598 F.3d 1158, 1163-64 (9th Cir.2010) (en banc) (quoting Jackson v. Virginia, 443 U.S. 307, 319, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979)); see also McDaniel v. Brown, 558 U.S. -, 130 S.Ct. 665, 673, 175 L.Ed.2d 582 (2010). We apply this standard to the securities [915]*915counts and the lying-to-auditors counts in turn.
A. Securities Counts
All of the securities counts — one count of securities fraud and seven counts of making false filings with the SEC— required the government to prove that NAI materially misstated the revenue it earned in certain quarters and years through its choice of accounting method. NAI’s reports of allegedly inflated revenue furnished the “untrue statement of material fact” required for each of the false filing counts, as well as the “misleading statement or omission of a material fact made with scienter” needed to sustain a fraud conviction under the general antifraud provision of § 10(b) of the Securities Exchange Act of 1934. United States v. Smith, 155 F.3d 1051, 1063 (9th Cir.1998) (internal quotation marks and numbering of elements omitted).
The government’s contention that NAI materially overstated its revenue necessarily entailed two claims: (1) that NAI recognized revenue at a different time than it should have; and (2) that NAI’s accounting produced artificially higher revenue figures in certain periods that “would have been viewed by the reasonable investor as having significantly altered the ‘total mix’ of information made available.” Basic Inc. v. Levinson, 485 U.S. 224, 231-32, 108 S.Ct. 978, 99 L.Ed.2d 194 (1988). The government relied on GAAP to make its case, on the first point, that sell-through accounting was required in instances where NAI used the sell-in method. But we need not decide whether NAI actually violated GAAP, because the government clearly failed to carry its burden on the second point, materiality. The prosecution offered no evidence adequate to prove that any GAAP violations materially affected the revenue that NAI reported.
The government relied at trial on the parties’ stipulations that applying sell-through accounting to NAI’s entire business would have resulted in “a revenue figure that is materially less than the reported figure” for the periods charged in the false filing counts.5 These stipulations are fatally overbroad, however, because the government did not contend that GAAP required NAI to use sell-through accounting for all sales. The government only offered evidence that sell-through accounting was required for the Ingram buy-in transactions, and the stipulations did not provide that applying sell-through accounting to those transactions alone would have made a material difference in any given period. Without evidence of how much less revenue NAI would have recognized on the Ingram deals if it had used sell-through accounting, the jury had no basis to conclude that the misstatement of reported revenue resulting from the Ingram transactions was material. Even presuming, as we must, that the jury drew all reasonable inferences in the prosecution’s favor, see Nevils, 598 F.3d at 1164, there was no way it could have properly inferred materiality from the evidence it had before it.
Confronted with this problem, the government argued after the verdict that the jury could have inferred materiality from [916]*916the mere fact that the buy-in deals with Ingram were substantial. The dollar amounts in the purchase orders for these transactions, according to the government’s calculations, “represented approximately 24% of the total revenue for NAI during 1998, 1999, and 2000” and “between approximately 7% to 40% of NAI’s total revenue” on a quarterly basis. But this argument failed to bridge the materiality gap because Goyal’s jury had to make the materiality findings that his convictions required, and it never saw these figures. See United States v. Rigas, 490 F.3d 208, 231 n. 29 (2d Cir.2007) (declining, in a sufficiency-of-the-evidence challenge, to “consider in the first instance arguments regarding materiality that were not presented to the jury”).
The jury could not have inferred materiality from this evidence even if it had seen it, because it was not the absolute magnitude of the buy-in deals that mattered. The jury needed to assess whether NAI’s use of sell-in accounting for the Ingram transactions materially increased NAI’s overall revenue when compared to using sell-through accounting. But the jury had no evidentiary basis for making the required comparisons. There was simply no evidence that the effect of using sell-in rather than sell-through accounting for the Ingram transactions was so substantial that it made a material difference. By the government’s own calculations, non-Ingram sales always accounted for most of NAI’s revenue. It would have been “mere speculation, rather than reasonable inference,” Nevils, 598 F.3d at 1167, for the jury to conclude that applying sell-through accounting to the revenue from the Ingram sales by themselves would have made a material difference in the company’s total revenue figures. Because Goyal’s jury had no competent evidence of materiality before it, it could not have properly convicted him on any of the securities counts.
B. Lying-to-Auditors Counts
The lying-to-auditors counts depended on two independent statements that Goyal made in management representation letters to PwC: (1) that NAI’s financial statements complied with GAAP and (2) that NAI had disclosed all “sales terms.” The government charged Goyal with willfully and knowingly making false statements to PwC. Willfulness requires that Goyal’s actions were “voluntarily and knowingly wrongful.” United States v. Tarallo, 380 F.3d 1174, 1189 (9th Cir.2004). Knowledge requires that Goyal “was aware of the falsification and did not falsify through ignorance, mistake, or accident.” United States v. Reyes, 577 F.3d 1069, 1080 (9th Cir.2009) (internal quotation marks omitted). The government therefore had to prove that Goyal voluntarily made statements to PwC that he knew were false.6
[917]*917When the events in this case occurred, NAI’s accounting for sales to distributors was governed by two GAAP guidelines: Financial Accounting Standards Board Statement No. 48 (“FAS 48”) and Statement of Position 97-2 (“SOP 97-2”). FAS 48 allowed the use of sell-in accounting only when:
1. The seller’s price to the buyer is substantially fixed or determinable at the date of sale;
2. The seller does not have significant obligations for future performance to directly bring about resale of the product by the buyer; and
3. The amount of future returns can be reasonably estimated.7
FAS 48, ¶ 6 (Fin. Accounting Standards Bd., June 1981). The only SOP 97-2 provision at issue in this case required that a “[vjendor’s fee” be “fixed or determinable” when revenue was recognized. The parties agreed that this condition was functionally equivalent to the “substantially fixed or determinable” prong of FAS 48.
1. Compliance with GAAP
The first basis for the lying-to-auditors counts requires us to decide whether the jury could have concluded that NAPs method of recognizing revenue violated GAAP. The government had to prove that Goyal’s representations to PwC that NAI complied with GAAP were materially false or misleading, and that he knew that.
It aids the analysis to group NAPs alleged GAAP violations into two categories. The first category encompasses allegations that sell-in accounting was improper under GAAP because terms of the Ingram deals allegedly involved prices that were not “substantially fixed or determinable,” or allowed “future returns” that could not “be reasonably estimated” at the time of sale. FAS 48, ¶ 6(a), (f). The second category involves NetTools, the subsidiary NAI used in 1998 and 1999 to repurchase product from Ingram. The government alleged that NAPs use of NetTools precluded sell-in revenue recognition because the arrangement amounted to NAPs “significant obligation[ ] for future performance to directly bring about resale of the product.” FAS 48, ¶ 6(e).
a. Uncertain prices and returns
The government argues that terms in the Ingram deals allowing for future adjustments to prices, such as sell-through rebates, meant that prices were not “substantially fixed or determinable” when NAI made the sales.8 It is undisputed, however, that future contingencies do not render sell-in revenue recognition improper if the seller can reasonably estimate the effect of the contingencies and set aside reserves adequate to cover them. The prosecution, therefore, could only prove this species of GAAP violation beyond a reasonable doubt by showing that NAI’s reserves did not reasonably account for the terms of the quarter-end sales. The prosecution failed to make its case on this point.
The government offered no evidence that NAPs reserves were, in fact, inadequate. Instead, it relied on equivocal and conclusory statements from witnesses— principally Hans Winters and Robert Stavers, two of NAPs auditors at PwC — to the effect that sales terms like rebates would “raise questions about the accounting the company was using.” Even where the wit[918]*918nesses testified that certain terms would preclude sell-in accounting, the government admits that adequate reserves would resolve these GAAP problems, and the witnesses had no basis to opine on whether NAI’s reserves actually were inadequate.
As the district court observed, no witness had “done the work, [or] perform[ed] the computations that would establish a basis to render an opinion” about how FAS 48 applied to NAI’s financial statements, including the reserves it set aside.9 With no evidence of what NAI’s reserves were, or how they fell short of amounts that the Ingram sales required, no reasonable juror could have found a GAAP violation that depended on insufficient reserves.
A similar problem arose with respect to sales terms that allowed Ingram to return software to NAI. GAAP only allows sell-in accounting to be used when “[t]he amount of future returns can be reasonably estimated.” FAS 48, ¶ 6(f). The government argued that certain terms in the buy-in transactions gave Ingram an unlimited right to return software it purchased, and that an unlimited right of return automatically defeated sell-in accounting.
Stavers testified, in response to a hypothetical question, that “if the distributor had an unlimited right of return, then ... we do not believe it would have been possible to estimate the return.” But he did not base this statement on an analysis of NAI’s actual return estimations. FAS 48 lays out four “factors [that] may impair the ability to make a reasonable estimate” of future returns. FAS 48, ¶ 8. No witness applied these factors to NAI’s buy-in deals or concluded that NAI could not accurately estimate returns. The prosecution needed to prove what NAI did, not what the buy-in terms made hypothetically possible. Without evidence that Ingram’s returns were not amenable to reasonable estimation, no reasonable juror could have found that using sell-in accounting for these sales violated GAAP.
b. NetTools
The NetTools arrangement violated GAAP if NAI’s “commitments” to repurchase prespecified quantities of its software from Ingram amounted to “significant obligations for future performance to directly bring about resale of the product.” FAS 48, ¶ 6(e). Goyal contends that the NetTools “commitments” were not “significant obligations” because they were not technically binding agreements, but merely “forecast[s]” of how much software Net-Tools would buy back from Ingram.
One of Ingram’s associate buyers testified that although “there [were] no guarantees, ... the forecast came in spot-on every time for Net Tools.... If they said Net Tools was going to do 25 million through the end of the quarter, on the last day of the quarter we were receiving Net Tools’[purchase orders] up until 2:00 in the morning to make sure that we hit 25 million.” Ingram relied on the NetTools forecasts to predict its own inventory. And NAI used its NetTools forecasts, in quarter-end negotiations with Ingram, to assure it that certain quantities of software would sell in subsequent quarters. Even if these commitments were not formally enforceable, the evidence makes clear that both parties treated them as reliable and expected them to be honored. A reasonable juror could have concluded that the [919]*919NetTools commitments fell within the meaning of “significant obligations,” and thus that using sell-in accounting violated GAAP.
Even so, the prosecution had to prove that Goyal willfully and knowingly misled PwC when he asserted that NAI complied with GAAP. We are mindful that the mere fact that “evidence submitted by the government is wholly susceptible to innocent explanations,” such as Goyal’s alternative understanding of “commitments,” is not enough to reverse a conviction on appeal. United States v. Wiseman, 25 F.3d 862, 866-67 (9th Cir.1994) (quoted and overruled by Nevils, 598 F.3d at 1166-67). As long as the jury could have accepted a culpable explanation consistent with the proof of defendant’s conduct, we must assume, in the prosecution’s favor, that the jury did so. Nevils, 598 F.3d at 1166-67; see United States v. Dinkane, 17 F.3d 1192, 1196 (9th Cir.1994) (“The relevant inquiry is not whether the evidence excludes every hypothesis except guilt, but whether the jury could reasonably arrive at its verdict.”). The government nonetheless must offer some evidence to support a culpable explanation. “[WJhere there is a total failure of proof’ of the required mental state, we cannot affirm a conviction. Nevils, 598 F.3d at 1167 (quoting Briceno v. Scribner, 555 F.3d 1069, 1079 (9th Cir.2009)) (internal quotation marks omitted). The government’s failure to offer any evidence supporting even an inference of willful and knowing deception undermines its case.
The government offers several ways that the jury could have inferred fraudulent intent from Goyal’s conduct, but none can support the inferences the government would draw. Goyal’s desire to meet NAPs revenue targets, and his knowledge of and participation in deals to help make that happen, is simply evidence of Goyal’s doing his job diligently. See, e.g., Anderson v. First Sec. Corp., 249 F.Supp.2d 1256, 1270 (D.Utah 2002) (effort “to meet analysts’ numbers and not disappoint Wall Street is merely an example of a company’s shared motives to look good” that does not imply “that the company was engaged in fraudulent conduct”). Similarly, Goyalis presumed knowledge of GAAP as a qualified CFO does not make him criminally responsible for his every conceivable mistake. If simply understanding accounting rules or optimizing a company’s performance were enough to establish scienter, then any action by a company’s chief financial officer that a juror could conclude in hindsight was false or misleading could subject him to fraud liability without regard to intent to deceive. That cannot be. Cf. Merck & Co., Inc. v. Reynolds, 559 U.S. -, 130 S.Ct. 1784, 1796-97, 176 L.Ed.2d 582 (2010) (holding that “facts that tend to show a materially false or misleading statement” do not always suffice “to show scienter as well”). That Goyal’s compensation was linked to NAI’s success — half of his bonus was based on achieving corporate goals — does not change matters. Such a general financial incentive merely reinforces Goyal’s preexisting duty to maximize NAI’s performance, and his seeking to meet expectations cannot be inherently probative of fraud. See Aldridge v. A.T. Cross Corp., 284 F.3d 72, 83 (1st Cir.2002) (financial incentive linked to company’s performance “cannot be enough to establish scienter” if it does not “go far beyond the usual arrangements of compensation based on the company’s earnings”).
The government also argues that Goyal must have known about various revenue recognition problems because others at the company claimed that they were aware of accounting improprieties. Eric Borrmann, NAI’s vice president (and later treasurer), testified that he believed there were “balance-sheet issues that relate to the lack of reserves. Or perhaps not suffi[920]*920cient reserves. There is also ... lack of disclosure around issues relating to the [distribution] channel.” Evan Collins, NAI’s former corporate controller, offered his opinion that “some of the terms ... in certain [buy-in] agreements ... would ruin [sell-in] revenue recognition.”
Neither of these witnesses found fault with the NetTools commitments, however, so neither shed any light on what Goyal knew or did not know about that arrangement. Borrmann’s testimony concerned NAI’s accounting for uncertain terms in the buy-in deals such as rebates and discounts; these are unrelated to NetTools. Collins’s statement was about perceived problems with written buy-in letters; NAI made its NetTools commitments orally. Because neither Borrmann’s nor Collins’ testimony said anything about NetTools, that testimony did not prove anything about Goyal’s knowledge on that subject.
The government also suggests that a jury could infer intent from Goyal’s statement that “he couldn’t know about” a memo NAI’s vice president of sales distribution had sent him questioning NAI’s accounting for the terms of several buy-in deals and enclosing letters documenting those terms. The government appears to argue that Goyal did not want to know about the letters because that would have made his statements about GAAP compliance knowingly misleading. But this incident does not prove anything of the sort because the memo and letters were not related to whether NAI’s use of sell-in accounting complied with GAAP and did not raise any red flags about it. The NAI vice president of sales did not claim that the terms in the letters violated GAAP; indeed, he admitted that he had no idea what the payments documented in the letters were for. The government therefore produced no evidence that Goyal’s knowledge of the letters’ contents would have made his statements to the auditors knowingly misleading. Cf. SEC v. Retail Pro, Inc., 673 F.Supp.2d 1108, 1142-43 (S.D.Cal.2009) (CFO violated Rule 13b2-2 by signing management representation letters stating he had no “knowledge of allegations of fraud or suspected fraud” after company’s contract administrator sent him an email credibly alleging that there was a potential fraud); SEC v. Espuelas, 579 F.Supp.2d 461, 487 (S.D.N.Y.2008) (finding no violation of Rule 13b2-2 where there were no “red flags” indicating that statements to auditors were misleading).
In sum, of the various GAAP violations alleged, only the problem with NetTools was supported by the government’s proof. But no evidence supported a finding that Goyal knew that NetTools’s commitments violated GAAP. The lying-to-auditors counts, therefore, cannot be sustained on the ground that Goyal’s assertion of GAAP compliance to PwC was willfully and knowingly false.
A Disclosure of sales terms
Goyal also affirmed in seven signed management representation letters to PwC that NAI had “fully disclosed to [PwC] all sales terms, including all rights of return or price adjustments, and all warranty provisions.” The government argued that this statement was willfully and knowingly false, independent of any GAAP violation, because NAI, at Goyal’s direction, did not turn over the buy-in letters memorializing the quarter-end Ingram transactions.
Goyal contends that his affirmations of disclosure were not false, because PwC had access to all “sales terms” through Ingram’s debit memos. Whenever Ingram claimed a concession from a buy-in deal, it sent NAI a debit memo requesting that the promised discount or rebate be honored. Ingram typically cited the deal term it sought to apply, and sometimes attached [921]*921its copy of the relevant buy-in letter for reference.
The government maintains that NAI could only satisfy its duty to disclose “sales terms” by disclosing the buy-in letters themselves. Equivalent information, trickling in piecemeal in the form of debit memos from Ingram, did not, according to the government, measure up to what Goyal claimed to have disclosed.
Whether this is correct is a close question. NAI recognized revenue based on Ingram’s purchase orders, invoices, and debit memos, not the buy-in letters that spelled out what sales terms would be. Goyal reasonably could have thought, in good faith, that conveying the sales terms in one of the forms NAI used to recognize revenue was sufficient. Our review of a motion to acquit limits us, however, to what a reasonable juror could have found, and there was enough evidence for a juror to conclude that nondisclosure of the buy-in letters mattered. Stavers testified that “[i]t would have been important for [PwC] to have seen these letters in order to determine the proper accounting for the sales transactions.” He further explained that a debit memo did not adequately provide him with “the terms of the deal” because the debit memos came in “after the fact”: “It doesn’t help us on June 30th to look at a document issued in August, in order for us to determine the accounting at the end of June.” A reasonable juror could have concluded that buy-in letters were necessary for a “full[ ]” disclosure of “all sales terms,” and that Goyal’s representations to PwC on that subject were materially false.
But even if sales terms were not disclosed to PwC, the government’s case suffered from a “total failure of proof,” Nevils, 598 F.3d at 1167, that Goyal willfully and knowingly misled PwC. Several of the government’s arguments for inferring mens rea in connection with GAAP violations, see part II.B.l.b supra, apply equally to Goyal’s knowledge of withholding of buy-in letters. The inferences the government would have a juror draw — from Goyal’s accounting knowledge and participation in buy-in transactions, from his incentives to use the deals to meet NAI’s projected revenue, and by attributing Borrmann and Collins’ claimed knowledge to Goyal — are no more valid in this context. We reject them for the reasons already articulated and turn now to arguments specific to the buy-in letters.
The government points to Collins’ testimony that he, as corporate controller, withheld buy-in letters from PwC because that is what he believed Goyal wanted. Collins left NAI before the time periods covered by the lying-to-auditors counts. His opinion, therefore, is only pertinent to the extent it supports an inference that Goyal’s intentions in the relevant period were the same as Collins assumed they were during his tenure. More fundamentally, though, Collins offered no basis for his belief about Goyal’s intentions. He admitted that Goyal had never told him to withhold buy-in letters from PwC, and the government introduced no other evidence that Goyal wanted them withheld. To the contrary, Collins told Goyal that he would “[e]nsure [SOP 97-2] compliance for the year end audit,” and “[m]eet all filing requirements.” Baseless speculation by a cooperating witness is not proof of fraudulent intent, especially when it contradicted evidence that Goyal had reason to expect compliance with auditing requirements.
Finally, the district court relied on the notion that Goyal could be convicted of lying to PwC because he “had an affirmative responsibility — as set forth in the management representation letters — to disclose” the buy-in letters. But even if he did, his failure to do so does not indicate scienter. The district court’s theory is [922]*922therefore untenable because it makes a strict-liability crime out of one that requires willful and knowing deception. Cfi United States v. Smith, 155 F.3d 1051, 1068 n. 25 (9th Cir.1998) (declining to adopt a construction of 15 U.S.C. § 78ff(a) that “de facto eliminates the mens rea requirement”). Absent any proof that Goyal willfully concealed buy-in letters from PwC, his convictions on this basis must be reversed.
III. Conclusion
Even viewing the evidence in the light most favorable to the prosecution, no reasonable juror could have found Goyal guilty beyond a reasonable doubt of any of the charges against him. The judgment of the district court is REVERSED, and the case is REMANDED for entry of judgment of acquittal on all counts.