Appeal from the United States District Court for the Northern District of Illinois, Eastern Division. No. 1:19-cr-00226-9 — Virginia M. Kendall , Chief Judge .

ARGUED DECEMBER 8, 2025 — DECIDED AUGUST 18, 2026

Before ROVNER, JACKSON-AKIWUMI, and MALDONADO, Circuit Judges .

JACKSON-AKIWUMI, Circuit Judge . Miroslaw Krezja was accused of participating in a multi-year scheme which ultimately caused the collapse of Washington Federal Bank for Savings. Pursuant to the scheme, Krezja and other known associates of the bank’s then-president received millions of dollars from the bank in the form of commercial real estate loans. Those loans were often insufficiently secured, improperly Page 2documented, and intentionally hidden from regulatory officials. After a multi-week trial, a jury convicted Krezja of conspiracy and aiding and abetting embezzlement. Krezja now challenges the sufficiency of the evidence to sustain his convictions and several of the district court's evidentiary decisions. Because we find no reversible error, we affirm.

Washington Federal Bank for Savings (WFBS) was a federally insured bank headquartered in Chicago. WFBS was regulated by the Office of the Comptroller of the Currency (OCC), which conducted on-site examinations of the bank every eighteen months. During the period relevant to this appeal, WFBS maintained a portfolio of residential and commercial real estate loans, including construction loans issued to developers throughout Chicago. At some point, at the direction of then-bank president John Gembara, senior bank officials started manipulating WFBS’s lending operations to benefit a select group of borrowers known as “friends of Gembara.' Those favored borrowers received accommodations unavailable to ordinary customers, including repeated payment extensions, continued loan advances despite loan nonperformance, and protection from collection and regulatory scrutiny. Miroslaw Krezja was among that group of favored borrowers.

Between 2005 and 2007, WFBS issued Krezja four residential construction loans secured by three Chicago properties: 4207 N. Lawndale Avenue, 3135 N. Spaulding Avenue, and 2020 N. Tripp Avenue. At the time they were issued, the loans functioned consistent with industry standards: WFBS issued Page 3the loans and created loan files, the parties executed promissory notes, and Krezja pledged collateral and began construction on the properties.

But the loans stopped functioning as traditional construction loans sometime after the real estate market collapsed in 2008. WFBS insiders began to conceal the true condition of Krezja's loans by manipulating accounting entries, altering loan files, and creating artificial banking records to prevent regulators from learning the truth. By 2017, the outstanding balance on Krezja’s loans totaled approximately $5.9 million, while the collective appraised value of the collateral stood at only $1.9 million. But that reality was not reflected in WFBS’s records, which were altered to create the appearance that those loans were performing. The OCC eventually learned the true status of WFBS’s loans, but by then it was too late.

During a 2017 examination, OCC examiners—led by Examiner-in-Charge Billy Lyons-discovered that WFBS had supplied them with inaccurate information that omitted loans to the “friends of Gembara.” Examiners then determined that loan distributions to members of that group, including Krezja, had exceeded the amounts stated in the corresponding promissory notes, without adequate collateral. At some point during that OCC examination, the true status of the bank became clear and it failed, leaving approximately $67 million in nonperforming loans. Lyons and his team subsequently prepared a memorandum summarizing the loan files associated with four of Krezja’s loan numbers and directing the WFBS Board of Directors to deem $4,018,928 of those loans uncollectible.

After WFBS’s failure, the Federal Deposit Insurance Corporation (FDIC) was appointed receiver and ultimately suffered a loss calculated at approximately $140 million. During Page 4a February 2018 interview, Krezja submitted an Affidavit of Financial Condition to the FDIC indicating that he had not re- ceived funds from the bank after 2011 and misrepresenting the extent of his travel during the conspiracy.

In 2021, a grand jury charged Krezja and several other individuals with committing offenses related to the alleged lending scheme from 2004 until 2018. The indictment charged Krezja specifically with conspiracy in violation of 18 U.S.C. § 371, and aiding and abetting embezzlement under 18 U.S.C. §§ 656 and 2. The conspiracy count alleged two objects: embezzlement in violation of 18 U.S.C. § 656; and false entries and related recordkeeping offenses in violation of 18 U.S.C. § 1005. In essence, the indictment alleged that Krezja knowingly participated in a scheme through which bank insiders diverted funds from WFBS while concealing the true condition of the loans from regulators and auditors.

Krezja moved to admit evidence that, between 1999 and 2006, he had obtained and fully repaid four earlier WFBS construction loans on other properties. He argued that those repayments were probative of his lack of fraudulent intent as to the later loans. The government opposed the motion, arguing the earlier loans differed materially from the loans at issue. The government noted the earlier loans were fully repaid, the properties were completed and sold, the disbursements did not exceed the amounts secured by the corresponding mortgages, and the surviving payment records for the earlier loans were incomplete. The district court agreed and excluded the evidence.

Page 5C. Trial

During a ten-day trial in 2023, the parties litigated several evidentiary disputes and offered very different accounts of Krezja’s relationship with WFBS and the status of his loans. The government sought to introduce evidence concerning OCC examinations, the broader condition of WFBS, other favored borrowers, and WFBS’s eventual collapse. The district court admitted the evidence over Krezja’s objection that much of the proposed evidence was irrelevant and unduly prejudicial.

The government presented documentary evidence from WFBS’s loan files and testimony from former bank employees, regulators, and cooperating witnesses to prove that the loans WFBS provided to Krezja were vehicles for embezzlement rather than genuine commercial transactions. The government also introduced evidence showing that construction activity slowed on Krezja’s projects before ceasing altogether. Government witnesses described partially completed homes, deteriorating properties, prolonged delays, a lack of meaningful progress on the projects, and Krezja’s refusal to sell the properties—all during the same period Krezja received advances from the bank. The government additionally highlighted that Krezja's balances increased (due to disbursements) during the same period that his repayment activity remained minimal. Meanwhile, the evidence showed, Krezja used some of the disbursement money for international travel, car payments, and other personal expenses, instead of construction costs. According to the government, ordinary commercial lenders would have stopped funding Krezja's projects, declared defaults, or initiated collection efforts.

Page 6Instead, it argued, WFBS insiders repeatedly intervened to continue funding Krezja’s projects and shield them from scrutiny.

The defense offered a different view—that the evidence showed, at most, WFBS’s imprudent lending decisions and Krezja's failure to repay loans during a severe real estate downturn. According to the defense, Krezja’s projects were casualties of the economic collapse that affected real estate developers nationwide. The defense argued that construction delays, declining property values, and loan extensions were common features of the post-2008 market and did not establish criminal conduct. The defense also maintained that Krezja’s projects were delayed due to the poor quality of his construction materials (another common issue in the real estate industry), which became the subject of separate litigation.

After hearing from both sides, the jury convicted Krezja on both counts.

Following the jury’s verdict, Krezja moved for a judgment of acquittal pursuant to Federal Rule of Criminal Procedure 29. He argued that the evidence was insufficient to establish either the charged conspiracy or the substantive embezzlement offense. In particular, he contended that the government failed to prove embezzlement as opposed to imprudent or preferential lending, failed to establish that he knowingly joined any conspiracy, and failed to prove the intent required for aiding and abetting liability.

Krezja also moved for a new trial pursuant to Federal Rule of Criminal Procedure 33 on two grounds. First, he argued that the jury's verdict was contrary to the weight of the Page 7evidence. Second, he challenged several of the district court’s evidentiary decisions. Namely, the court’s admission of evidence and testimony related to the OCC's examination of WFBS; admission of evidence related to the bank's failure; and its exclusion of evidence that he repaid his prior loans with WFBS. Krezja argued that these decisions were errors that deprived him of a fair trial.

The district court denied both motions. As to Krezja’s Rule 29 motion, the court concluded that the evidence, viewed in the light most favorable to the government, permitted a rational jury to find the essential elements of the charged offenses beyond a reasonable doubt. In rendering its decision, the court recounted evidence showing WFBS continuously advanced loans on Krezja’s failing projects, and that insiders concealed Krezja’s delinquent loans and manipulated documents. In addition, the court highlighted testimony that a promissory note associated with one of Krezja’s projects had been created after the fact. As to Krezja’s aiding and abetting charge, the court concluded that the verdict was proper because the offense did not require Gembara to have personally benefitted from the funds and instead 'requires only the fraudulent conversion or appropriation of property by a person in lawful possession of the property with an intent to defraud or injure.” With respect to Krezja’s Rule 33 motion, the court concluded that the challenged evidentiary rulings fell within its broad discretion and reaffirmed its prior relevance determinations. The court further determined that any potential errors did not prejudice Krezja or otherwise result in a miscarriage of justice warranting a new trial. Krezja now appeals.

Page 8Krezja presents multiple issues on appeal. First, he argues that the government’s trial evidence departed from the indictment such that he lacked notice of the charges against him. Second, he contends that the government’s evidence was insufficient to sustain his convictions. Third, Krezja raises several evidentiary challenges. He challenges the district court’s exclusion of his prior loan repayment evidence, as well as its admission of OCC examination testimony and exhibits, and evidence relating to WFBS’s collapse and the FDIC’s corresponding losses. We address each argument in turn.

Krezja argues that the government proved a materially different scheme at trial than the one the grand jury charged him with. As Krezja sees it, the government charged a conspiracy involving embezzlement and false entries but transformed his trial into a sweeping examination of corruption, irresponsible lending practices, and institutional dysfunction at WFBS. This shift invited the jury to convict him, he says, based on broader conduct and not the charged offenses. Because Krezja did not develop this argument below, we review his claim for plain error. United States v. Scheuneman , 712 F.3d 372, 378–79 (7th Cir. 2013).

At times Krezja labels the problem-the divergence between the government’s indictment and its proof at trial—a “fatal variance.” But elsewhere in his briefs he argues that the government “effectively amended the Indictment.” The difference between a constructive amendment of an indictment and a variance can be significant. United States v. Heon Seok Lee , 937 F.3d 797, 806 (7th Cir. 2019). In a constructive Page 9amendment, the government has presented evidence to support an offense not alleged in the indictment, or the court has instructed the jury to find an offense not alleged in the indictment. Id . A constructive amendment of the indictment is a per se reversible error because the Fifth Amendment “limits the available bases for conviction to those contained in the indictment.” United States v. Willoughby , 27 F.3d 263, 266 (7th Cir. 1994). A variance, on the other hand, refers to a situation where the government has proved the same crime alleged in the indictment, but using facts materially different from those set forth in the indictment. United States v. Ratliff-White , 493 F.3d 812, 820 (7th Cir. 2007). A variance is only fatal if the defendant can demonstrate prejudice “because he cannot anticipate from the indictment what evidence will be presented against him or [he] is exposed to the risk of double jeopardy.” Id . (citation omitted).

Krezja’s contention, whether it alleges a fatal variance or a constructive amendment of the indictment, overstates the record. The indictment alleged a scheme through which Krezja and other favored borrowers received disbursements from nonviable loans that bank insiders concealed by manipulating records, recreating documents, and misleading regulators. The indictment alleged that insiders at WFBS embezzled bank funds by advancing purported real estate loans to Krezja and others without expectation of repayment.

The government’s proof tracked that theory throughout trial. The government presented evidence that Krezja continued to receive disbursements long after his loans were no longer in good standing. It also showed that he signed postdated documentation related to his loans and possessed Page 10documents reflecting payments on his loans that he had never made.

The record additionally revealed that Krezja’s loans were not isolated lending irregularities but rather part of a broader scheme through which WFBS insiders artificially maintained favored borrowers’ failing loans. The government’s theory— that favored borrowers benefitted from noncompliant loans that they helped bank insiders conceal from regulators—was therefore materially consistent from indictment through closing argument. In keeping with this theory, the evidence concerning other favored borrowers, the scheme writ large, and WFBS’s institutional practices was introduced to explain the conspiracy to the jury and show how the scheme functioned. Whereas other evidence—a document bearing Krezja’s signature, data showing the true condition of his loans, and testimony about his personal expenditures from disbursements— was offered to prove his guilt. We see no fatal variance or constructive amendment of the indictment.

Moreover, Krezja cannot show prejudice, which is necessary to succeed on a fatal variance claim. This was not a case where Krezja was unfairly surprised or unable to prepare a defense. Krezja was aware before trial that the government intended to rely on evidence concerning his and others’ favorable treatment, bank insiders’ concealment practices, regulatory examinations, and broader misconduct within WFBS. The defense vigorously contested this approach throughout the proceedings and repeatedly argued that the government had criminalized Krezja’s inability to repay his loans rather than proving the charged offenses. Thus, even if the government varied in its theory of the case, the absence of prejudice dooms any fatal variance claim. With that, we proceed to Page 11Krezja’s challenge to the sufficiency of the evidence against him.

We review a challenge to the sufficiency of the evidence de novo. United States v. Johnson , 874 F.3d 990, 998 (7th Cir. 2017). In doing so, we view all evidence in the light most favorable to the government and ask whether the jury could have found the defendant guilty of the charged offenses beyond a reasonable doubt. Id . We do not reweigh evidence or reassess witness credibility. United States v. Medina , 969 F.3d 819, 821 (7th Cir. 2020). Our task is simply to determine whether the jury was entitled to adopt the government’s theory based on the evidence presented at trial. Johnson , 874 F.3d at 998.

Krezja challenges the sufficiency of the evidence to sustain his convictions on two grounds. First, he argues that the government failed to prove that he knew of or participated in any conspiracy to falsify bank records. Second, he maintains that the government failed to show that Gembara embezzled any funds from his loans, which he says is a prerequisite to proving that he aided and abetted Gembara’s embezzlement. We address each argument separately.

To establish a conspiracy under 18 U.S.C. § 371, the government must prove: (1) the existence of an agreement to commit an unlawful act; (2) the defendant’s knowing and intentional participation in that agreement; and (3) the commission of an overt act in furtherance of the conspiracy. See United States v. Soy , 454 F.3d 766, 768 (7th Cir. 2006). The overt act committed need not be the substantive offense or even an Page 12element thereof; the act is sufficient so long as it is committed in furtherance of the underlying substantive offense. Id .

Because conspiracies are inherently covert enterprises, the government need not produce direct evidence of an express agreement. United States v. Ford , 21 F.3d 759, 762 (7th Cir. 1994); United States v. Useni , 516 F.3d 634, 646 (7th Cir. 2008). A conspiracy may be proved entirely through circumstantial evidence, including coordinated conduct, repeated interactions among participants, and actions that would be difficult to explain absent a shared unlawful objective. United States v. Morales , 145 F.4th 792, 796–97 (7th Cir. 2025). The key question is whether the conspirators are performing different functions in pursuit of common criminal objectives. Id .

The indictment in this case alleges two objects of the conspiracy: (1) to knowingly make false entries in WFBS’s books, reports, or statements with 'intent to injure and defraud' WFBS in violation of 18 U.S.C. § 1005; and (2) to “embezzle, abstract, purloin, and willfully misapply” WFBS funds and funds entrusted to WFBS “with intent to injure and defraud the bank” in violation of 18 U.S.C. § 656.

To convict Krezja, the government had to prove that he acted with “at least the degree of criminal intent necessary for the substantive offense itself.” Soy , 454 F.3d at 768. The requisite criminal intent for both offenses is intent to defraud or injure.¹ 'Intent to defraud requires a willful act by the defendant with the specific intent to deceive or cheat, usually for the Page 13purpose of getting financial gain for one’s self or causing financial loss to another.” United States v. Howard , 619 F.3d 723, 727 (7th Cir. 2010) (citation modified). Importantly, “intent to defraud may be established by circumstantial evidence and by inferences drawn from examining the scheme itself.' United States v. Paneras , 222 F.3d 406, 410 (7th Cir. 2000) (citation modified). Moreover, under both § 656 and § 1005, “[a] reckless disregard . . . of [the] bank’s interest is sufficient to establish the requisite intent to defraud.” United States v. Larson , 581 F.2d 664, 667 (7th Cir. 1978) (discussing § 656); accord United States v. McAnally , 666 F.2d 1116, 1118-20 (7th Cir. 1981) (discussing § 1005).

In sum, the question before us is whether the government proffered sufficient evidence that Krezja 'specifically intended that some conspirator commit each element of [either] substantive offense,” that he acted with at least a reckless disregard of the bank’s interest, and that a conspirator committed an overt act in furtherance of either substantive offense. Ocasio v. United States , 578 U.S. 282, 287–88, 292 (2016) (citation modified); accord United States v. Bucey , 876 F.2d 1297, 1312 (7th Cir. 1989) (“When an indictment alleges a conspiracy with multifarious objectives, a conviction will be sustained so long as the evidence is sufficient to show that the defendants agreed to accomplish at least one of the alleged objectives.”).

According to Krezja, the government's evidence established only a borrower-lender relationship between him and WFBS and did not show that he agreed to participate in any misconduct with bank employees. He highlights that he Page 14exercised no authority over accounting systems, regulatory submissions, document creation, loan classifications, or bank examinations. Instead, he says, those functions remained entirely within the control of WFBS employees.

Krezja’s arguments have some force. Neither preferential treatment nor repeat borrowing necessarily prove conspiracy. Indeed, a customer may receive favorable terms, an improper accommodation, or even unethical assistance from a lender without agreeing to join a criminal enterprise. Accordingly, to prove conspiracy here, the government was required to offer evidence showing Krezja knowingly participated in the scheme, not that he passively received improper benefits. Soy , 454 F.3d at 768.

The government’s trial evidence satisfied that burden. The evidence showed that the challenged loans were not isolated transactions but instead involved multiple projects and extensions, repeated disbursements, and continuing interactions over an extended period. Government witnesses testified that the disbursements continued despite obvious warning signs that ordinarily would have resulted in default, foreclosure, or cessation of disbursements. The government’s evidence also showed that several insiders hid loan delinquency, manipulated records, recreated documents, and misrepresented the condition of Krezja’s loans to regulators.

The recreated promissory note with Krezja's signature provides an additional basis for the jury’s verdict. The government introduced the note as evidence of the conspirators’ efforts to create support for transactions that otherwise lacked legitimacy. Although the defense offered innocent explanations for the document, the jury was not required to accept them. Instead, the jury could infer that Krezja was not merely Page 15an uninformed bystander but rather a knowing participant interested in preventing discovery of the underlying scheme. The basis for such an inference is even stronger considering that the note bore Krezja’s signature and was recovered in his possession.

From this evidence, a reasonable juror could infer that Krezja and WFBS maintained a nontraditional lending relationship. From there, a juror could further infer that Krezja, along with his coconspirators, attempted to conceal the scheme and preserve the appearance of legitimate lending relationships that otherwise would withstand regulatory scrutiny; in other words, to disguise the conspiracy. Considering this permissible inference and viewing the evidence in the light most favorable to the government, a reasonable juror could conclude that Krezja knowingly participated in a conspiracy to embezzle funds. The district court therefore properly denied the Rule 29 motion as to the § 656 object of the conspiracy.

1. b. Sufficiency of the Evidence as to the § 1005 Object: Conspiracy to Falsify Bank Records

Krezja next argues that even if the evidence was sufficient to establish that his lending relationship with WFBS was unorthodox, the government failed to prove he conspired with bank insiders to falsify documentation. Here again, Krezja is mistaken.

Section 1005 prohibits bank employees from knowingly making false entries in bank records with intent to deceive regulators or examiners. White v. Keely , 814 F.3d 883, 888 (7th Cir. 2016); 18 U.S.C. § 1005. According to Krezja, he cannot be guilty of this offense because bank insiders-not Krezja- Page 16controlled loan files and accounting systems and communicated with regulators. He maintains that because he neither created nor maintained bank records, the government failed to establish his knowing participation in a false-entry scheme.

But those arguments ignore both the standard for a conspiracy conviction and the government’s theory of the case. Because Krezja was not a bank employee, the government did not contend that he personally entered information into WFBS's accounting systems or drafted regulatory submissions. Rather, the government argued that Krezja knowingly participated in a broader scheme where success depended on the creation and maintenance of false or misleading records and that he signed documents in furtherance of that scheme.

Sufficient evidence supports that theory. The government presented evidence of corruption at WFBS generally but also evidence and documentation associated with Krezja’s development projects. Witnesses testified about recreated promissory notes, altered documents, and records allegedly designed to conceal the deteriorating condition of Krezja’s loans. According to the government, those materials were deliberate efforts to create the appearance that Krezja’s loans were legitimate and performing when in reality, as the evidence showed, Krezja had not made the requisite payments on his loans, his projects were stalled, and the collateral properties were deteriorating. From this evidence, a juror could reasonably infer that Krezja, who continued receiving benefits and disbursements over an extended period, understood that the transactions depended on documentation that did not accurately reflect reality, and he did his part to further the scheme.

The defense’s contrary interpretation—that Krezja merely signed documents prepared by others and was unaware of Page 17any broader deception-was certainly one the jury could have accepted. But because it was not the only reasonable interpretation of the evidence, the district court properly denied Krezja’s Rule 29 motion as to the record falsification object of the conspiracy.² United States v. Robinson , 165 F.4th 1043, 1051 (7th Cir. 2026) (“We will overturn a district court’s Rule 29 denial only if no rational trier of fact could have found the defendant guilty, a burden for defendants that we have described as nearly insurmountable.”) (citation modified).

A person can be punished as a principal under 18 U.S.C. § 2 if he 'commits an offense against the United States or aids, abets, counsels, commands, induces or procures its commission.' The indictment charges Krezja with aiding and abetting Gembara's embezzlement, abstraction, purloining, and willful misapplication of over $2.8 million from WFBS. The government focused solely on an embezzlement theory before the district court; we similarly limit our discussion on appeal. To convict Krezja of this offense, the government needed to show that Gembara embezzled funds by proving: (1) Gembara was a bank employee; (2) WFBS was a federally insured bank; and (3) Gembara used his “position to embezzle the bank’s funds . . . with the intent to injure or defraud the bank.” United States v. Parker , 716 F.3d 999, 1008 (7th Cir. 2013); see also 18 U.S.C. § 656. Only the third requirement is in dispute.

Page 18“Embezzlement is the fraudulent appropriation of property by a person to whom such property has been entrusted, or into whose hands it has lawfully come.” United States v. Bailey , 734 F.2d 296, 303 (7th Cir. 1984) (citation modified). “The elements of embezzlement are: (1) a trust or fiduciary relationship, (2) that the property claimed embezzled is embraced within the meaning of the statute, (3) that it came into the possession or care of accused by virtue of his employment, (4) it is property of another, (5) that his dealing therewith constituted a fraudulent conversion or appropriation of same to his own use, and (6) such was with the intent to deprive the owner thereof.” Id .

According to Krezja, embezzlement requires that the embezzler (here, Gembara) profit or benefit from the allegedly embezzled funds (Krezja’s loans). In Krezja’s view, because the government did not present evidence of any benefit Gembara received from his loans, it failed to prove that Gembara embezzled those funds.

In response, the government contends that the law does not require it to prove that Gembara benefitted from the allegedly embezzled funds. According to the government, the law’s requirement that entrusted property be converted to the embezzler’s “own use” means only “a use different than that for which they were entrusted with the property,” not the embezzler’s own consumption or benefit. In other words, the government says, “conversion to one’s ‘own use’ describes not a benefit to the embezzler, but rather [a] change in character of the property that lawfully came into the embezzler’s possession and care in the first instance.” The government argues that it proved embezzlement under this definition Page 19because the evidence showed that Gembara used the funds entrusted to him for purposes other than those intended.

Although both sides present persuasive arguments, we need not define the bounds of the law to decide this case. This is because our only task is to determine whether the evidence was sufficient to support Krezja’s conviction. We believe it was under Krezja’s own standard.

At trial, the government presented testimony from an agent from the FDIC’s Office of the Inspector General who analyzed Krezja's bank accounts. The agent testified that, between 2013 and 2017, WFBS disbursed $480,000 to Krezja, and that Krezja used at least some of the money for personal expenses such as trips and his car payments. The jury also heard from the government’s former forensic investigator who testified that Krezja remitted payment to Gembara’s insurance agency in 2014 and 2015.

Given the timing of the payments, a reasonable juror could have inferred that Krezja used his loan disbursement to make payments to Gembara's insurance company for Gembara's benefit. From this, a juror could conclude that Krezja aided and abetted Gembara’s embezzlement. That is enough to sustain Krezja’s conviction.

C. Evidentiary Challenges 3

We evaluate challenges to the admissibility of evidence in light of all the evidence before the jury and review a district Page 20court’s decision to admit or exclude evidence for abuse of discretion. Ajayi , 808 F.3d at 1125. Because district courts enjoy “broad discretion to control the admission of evidence” we will deem a court’s evidentiary decision erroneous “only if no reasonable person could take the view adopted by the trial court.' Id . at 1121 (citation modified). Furthermore, even if we find that the district court committed an evidentiary error, we will reverse and order a new trial only if the error was not harmless. United States v. Simon , 727 F.3d 682, 696 (7th Cir. 2013).

Krezja challenges the district court’s exclusion of evidence that he successfully repaid prior loans to WFBS. According to Krezja, this evidence was central to his good-faith theory because it demonstrated that his historical relationship with WFBS involved legitimate lending transactions rather than criminal diversion of funds. He maintains that, by excluding this evidence, the district court prevented the jury from receiving the full story of his longstanding relationship with WFBS.

The district court excluded the proposed evidence based on its belief that the prior loans were insufficiently connected to the charged conduct and the evidence risked creating Page 21collateral mini-trials concerning unrelated transactions. The court did not abuse its discretion in reaching that conclusion.

In challenging the evidence, the government identified several material differences between the repaid loans and the charged loans. Namely, Krezja made zero payments on the charged loans but made at least some payments on the earlier set. The earlier projects were completed and sold whereas the charged properties sat unfinished for years. The charged loans vastly exceeded their approved amounts; the earlier loans did not. And finally, the charged loans ran through the entire conspiracy period while the earlier loans had only slight overlap with it . These dissimilarities diminish the repaid loans’ relevance to this case and undercut the premise that the earlier repayments could show Krezja’s intent as to the charged loans.

What's more, evidence that Krezja repaid earlier loans could have led the jury to acquit him based on the notion that one who honored past debts would not defraud a bank on new ones-an improper propensity inference, as the court acknowledged. The district court did not err by trying to guard against such propensity reasoning based on marginally relevant conduct. See United States v. Rogers , 587 F.3d 816, 822 (7th Cir. 2009) (quoting Fed. R. Evid. 403 advisory committee’s note) (“Evidence poses a danger of ‘unfair prejudice’ if it has 'an undue tendency to suggest [a] decision on an improper basis, commonly, though not necessarily, an emotional one.’”).

The incompleteness of the bank’s records independently supported the exclusion. Krezja planned to present a pattern of prior payments to show that his conduct during the conspiracy mirrored his behavior in the past. However, without Page 22complete payment records—which WFBS’s retention policy had rendered unavailable—he may not have been able to present that narrative accurately. Given that reality, it was reasonable for the district court to conclude that evidence of Krezja’s prior repayments could have confused the jury and caused them to speculate in order to fill the gaps.

Krezja responds that this reasoning conflates general-intent and specific-intent crimes, and that courts have upheld the admission of other-act evidence offered to negate a defendant’s specific intent. It is true that other-act evidence is admissible to rebut a defendant’s claimed lack of specific intent if it proves 'motive, opportunity, intent, preparation, plan, knowledge, identity, absence of mistake, or lack of accident.” Fed. R. Evid. 404(b)(2). But the district court’s ruling did not rest on a categorical rejection of other-act evidence’s potential use in disputing specific intent. It instead rested on the substantial dissimilarity between the two sets of loans and the unreliability of the surviving repayment records-both permissible grounds. The court also declined to admit the evidence on the grounds that, in the form available, it could generate confusion disproportionate to the evidence’s probative value. See Fed. R. Evid. 403. This too was not an abuse of discretion.

What’s more, even had the prior-loan evidence been admitted, it would not have proven Krezja lacked intent to defraud WFBS. An intent to eventually repay a loan does not negate the intent to defraud required under § 656; the relevant inquiry is whether the defendant intended to injure or defraud the bank in the interim, not whether he ultimately meant to make the bank whole. United States v. Radziszewski , 474 F.3d 480, 485 (7th Cir. 2007), as amended on denial of Page 23reh’g (May 14, 2007) (“It is well settled that a defendant’s ultimate intention to pay off a debt obtained fraudulently is irrelevant to the intent to obtain the money through deceptive means.”).

Krezja next challenges the admission of testimony and evidence about examinations the OCC conducted. Krezja does so on two grounds. The first has to do with a document that Billy Lyons’s OCC team prepared, which we will call the OCC Memo. Krezja argues that the admission of the OCC Memo violated the Sixth Amendment's Confrontation Clause because he did not have an opportunity to cross-examine its authors. He cross-examined only Lyons. As for the second ground, Krezja argues that the OCC Memo and Lyons’s testimony amounted to investigative hearsay, institutional conclusions, and quasi-expert opinions about wrongdoing at WFBS, not evidence of any crime by Krezja. Further to this point, he contends that Lyons’s testimony carried undue prejudicial force because jurors were likely to assign excessive weight to it because of Lyons’s position with the OCC. We address each argument in turn, beginning with Krezja’s constitutional challenge.

We must determine the appropriate standard of review before addressing the merits of Krezja’s Confrontation Clause challenge. Krezja contends that we should review his challenge de novo because he appropriately raised the issue before the district court. The government contends that we must review the challenge only for plain error because defense Page 24counsel did not clearly object to the evidence on constitutional grounds below. Krezja has the better argument.

At sidebar before the government questioned Lyons about the OCC Memo, defense counsel told the court: “I can’t crossexamine the people that wrote this or contributed to it.” That statement put the district court and the government on notice that counsel objected to the evidence, at least in part, based on an inability to confront the document’s authors. Contrary to the government’s suggestion on appeal, the defense was not required to cite the Sixth Amendment or explicitly invoke Confrontation Clause caselaw to preserve a confrontation challenge; the substance of the objection controls. See United States v. Cunningham , 462 F.3d 708, 712 (7th Cir. 2006) (recognizing that preservation of an objection is determined based on the wording and timing rather than whether counsel invokes a specific constitutional provision). We therefore review Krezja’s Confrontation Clause challenge de novo.

Based on our independent review, however, Krezja’s challenge fails on the merits. Krezja is correct that the Confrontation Clause is usually satisfied when a document’s author is made available for cross-examination. However, that requirement does not demand that the testifying witness have personally drafted the document or performed every underlying observation. A supervisor who oversaw the preparation of the document, vouches for its contents as the product of his team’s work, and is subject to meaningful cross-examination may stand in as the author for Confrontation Clause purposes. Accord United States v. Sokolow , 91 F.3d 396, 404–05 (3d Cir. 1996) (Confrontation Clause satisfied where a supervisor who personally audited many of the underlying claims and testified that an exhibit represented the results of his team’s Page 25processing was cross-examined at trial, even though he had not prepared the exhibit himself).

That standard is met here. The OCC Memo was a single, coordinated work product compiling the OCC examination team's findings about Krezja's loans. The Memo was prepared under Lyons’s direct supervision. He led the OCC examination team and personally reviewed and approved the OCC Memo’s contents before it was sent to WFBS’s Board of Directors. And he testified at trial that the OCC Memo reflected his team’s findings concerning Krezja’s loans. As the district court noted, given Lyons’s involvement in the examination, review, and approval process, it was “reasonable to consider him an author of the Memorandum, such that the testimonial statements in the Memorandum are his own.' That Lyons was available for and subject to cross-examination, then, is enough to satisfy the Confrontation Clause.

Krezja raises an additional concern that his cross-examination of Lyons was incomplete because his counsel did not ask Lyons about the document's preparation. This concern goes to the scope of the cross-examination Krezja’s team chose to conduct, not to the adequacy of the confrontation opportunity. Lyons was present, sworn, and answerable to questions about the OCC Memo’s authorship, its content, and his role in producing it. Because Lyons was an appropriate witness to present the OCC Memo, its admission did not violate the Confrontation Clause.

Krezja also challenges the district court’s admission of the OCC Memo on relevance and hearsay grounds. Although his arguments are not altogether clear on this point, he seems to Page 26suggest that the jury may have attributed undue weight to the evidence and improperly viewed Lyons and his team as experts on industry practices and financial irregularities.

Although we acknowledge that the OCC evidence carried institutional authority and presented a danger that jurors would equate regulatory oversight with proof of criminality, we cannot conclude that the evidence in this case crossed the line or was inadmissible. The government's case relied, in part, on showing that bank insiders concealed delinquent, nonperforming loans from regulators. Through this lens, Lyons’s testimony and other evidence about the OCC examinations were not merely background matters. They showed how the scheme went undetected because insiders preserved the appearance of a viable loan portfolio, and how the fraud was eventually discovered through an OCC examination. Without some explanation of the regulatory process, the jury may have struggled to understand how the recreated documents, missing delinquency information, and manipulated accounting entries were relevant to the case. The district court’s decision to admit the evidence was reasonable on these facts.⁴

Krezja’s final challenge concerns the district court’s admission of evidence pertaining to WFBS’s collapse and the FDIC’s consequent losses. Specifically, he maintains that the court should have prohibited the government from presenting testimony by the FDIC’s chief of accounting operations and documents quantifying the FDIC's total loss of approximately $140 million from WFBS’s closure. Krezja maintains that this testimony about accounting procedures, WFBS’s failure, and the bank’s closure process was unnecessary given the charges in this case. And, he says, the documentary evidence unfairly aggregated losses for all the bank’s improprieties and risked misleading the jury about his personal culpability. According to Krezja, this evidence improperly bolstered the government’s case and encouraged the jury to punish him for the bank's failure and the FDIC's corresponding losses rather than determine whether the government proved the charged offenses beyond a reasonable doubt.

Here again, Krezja's argument is reasonable. Evidence concerning institutional collapse and great loss to the government (and therefore taxpayers) carries emotional force extending beyond the narrow elements of the charged crimes. Jurors may be tempted to assign blame broadly once they learn that a bank failed and regulators intervened.

But the challenged evidence also possessed substantial probative value. The government’s theory depended on proof that the loans were not functioning as legitimate commercial transactions and that bank insiders concealed that reality from regulators as the financial condition of the bank's Page 28portfolio deteriorated. Evidence concerning mounting losses and regulatory intervention helped explain why bank employees manipulated records and continued disguising delinquent loans. Under these circumstances, we cannot say the court abused its discretion in concluding that the probative value of the evidence was not substantially outweighed by unfair prejudice.

We must end our review of the district court’s evidentiary decisions by assessing whether any error was harmless. We conclude that it was.

“The test for harmless error is whether, in the mind of the average juror, the prosecution’s case would have been significantly less persuasive had the improper evidence been excluded.” United States v. Gomez , 763 F.3d 845, 863 (7th Cir. 2014) (en banc) (citation modified). The government introduced extensive evidence supporting Krezja's intent to defraud. The evidence showed that between 2013 and 2017, Krezja received $480,000 in disbursements that he used partly for personal expenses, including international travel and car payments, rather than construction costs; possessed a backdated promissory note bearing his signature; possessed a 2011 mortgage interest statement falsely reporting over $87,000 in interest payments he never made; and submitted a false affidavit to federal investigators. Against that record, we cannot say the admission of evidence that Krezja repaid earlier loans or the exclusion of the OCC and FDIC-related evidence would have created a reasonable probability of a different verdict. Simon , 727 F.3d at 696.

Page 29The line separating an improper lending relationship from criminality was not obvious in this case. But the jury heard extensive evidence that Krezja’s projects ceased functioning as viable developments while bank insiders continued authorizing advances, concealing delinquency, manipulating records, and creating documents to preserve the appearance of legitimacy. Viewed in the light most favorable to the government, that evidence permitted a rational jury to conclude that Krezja knowingly participated in a scheme to defraud WFBS and aided and abetted embezzlement. Further, on this record, we cannot say that no reasonable person could agree with the district court's evidentiary decisions. The district court’s judgment is therefore AFFIRMED .