Having identified defendant-appellant Fulvio Flete-Garcia as the architect of a massive swindle, the government charged him with a litany of fraud-based crimes. Following four days of trial, Flete-Garcia threw in the towel and entered a straight guilty plea to all 48 counts of the indictment. Prior to sentencing, though, Flete-Garcia experienced buyer's remorse and attempted to withdraw his guilty plea. The district court denied this motion, as well as sentencing-related motions for discovery and for an evidentiary hearing. It then sentenced Flete-Garcia to 132 months' imprisonment and ordered him to make restitution in the amount of $7,737,486.10. Flete-Garcia appeals, raising a gallimaufry of alleged errors. Finding his asseverational array long on perfervid rhetoric but short on substance, we affirm.

I. BACKGROUND

We briefly rehearse the background of this appeal, reserving further elaboration for our subsequent discussion of the issues. We draw the facts from the trial record, the change-of-plea colloquy, the undisputed portions of the presentence investigation report (PSI Report), and the transcript of the disposition hearing. See United States v. Arias-Mercedes , 901 F.3d 1 , 4 (1st Cir. 2018) ; United States v. Fernández-Santos , 856 F.3d 10 , 14 n.1 (1st Cir. 2017).

For over half a decade, Flete-Garcia orchestrated and operated a lucrative tax-fraud conspiracy. To further this criminal enterprise, Flete-Garcia stole personal identification information (PII) from Puerto Rico residents and used this information to prepare and file fraudulent federal income tax returns. 1 These fraudulent returns generated refund checks, which Flete-Garcia deposited (through intermediaries) for his own benefit.

Flete-Garcia's scheme involved a handful of co-conspirators. One such co-conspirator assisted in the preparation of the fraudulent tax returns, while others assisted by cashing refund checks. When a co-conspirator's accounts were frozen, Flete-Garcia simply moved on to another individual and another set of accounts.

Flete-Garcia's scheme was nothing if not ambitious. In hindsight, the government says that it has been able to account for over $7 million in funds fraudulently obtained from the Internal Revenue Service (IRS) as well as $5 million, more or less, that would have been paid but for the detection of the fraud.

The plucked chickens eventually came home to roost. On March 9, 2017, a federal grand jury sitting in the District of Massachusetts returned a 48 count superseding indictment charging Flete-Garcia with conspiracy to defraud the United States (count 1), see 18 U.S.C. § 371 ; access device fraud (counts 2 and 3), see id. § 1029; conversion of government property (counts 4 through 20), see id. § 641; aggravated identity theft (counts 21 through 37), see id. § 1028A; and money laundering (counts 38 through 48), see id. § 1956(a)(1)(B)(i). Flete-Garcia initially maintained his innocence, a jury was empaneled, and trial commenced on July 10, 2017. During the first four days of trial, the government presented the bulk of its evidence (including nineteen of twenty-three witnesses), and Flete-Garcia cross-examined nearly all of the government's witnesses. As the fourth day of trial wound down, Flete-Garcia indicated that he wished to change his plea. He told the court that no one had pressured him into this decision but, rather, he had "started thinking about [his] family."

The district court engaged in a careful change-of-plea colloquy, see Fed. R. Crim. P. 11, and Flete-Garcia admitted his guilt with respect to all 48 counts. Once the court accepted the plea, it discharged the jury.

Shortly thereafter, Flete-Garcia retained new counsel. He also wrote a pro se letter to the district court maintaining that his trial was tainted and that he wanted to "null[ ] or void" his guilty plea. The court advised both Flete-Garcia and his new lawyer that it considered this letter to be without force and that any plea-withdrawal motion should be made by counsel. Relatedly, the court postponed sentencing at Flete-Garcia's request.

Flete-Garcia's new lawyer filed a flurry of motions. These filings included a motion to withdraw Flete-Garcia's guilty plea, a motion to compel discovery, and a motion for an evidentiary hearing to determine the amount of loss. Meanwhile, the probation office prepared the PSI Report, which (when issued) recommended certain guideline calculations. The probation office began its calculations by constituting a single group comprising counts 1 through 20 and counts 38 through 48. See USSG § 3D1.2(c), (d). It then formed a second group comprising counts 21 through 37. See id. § 2B1.6. Because the offense level for the money laundering counts contained in the first group carried the highest offense level, the PSI Report calculated the guideline sentencing range (GSR) by reference to those counts. See id. § 3D1.3(a). The ensuing calculation started with a base offense level of 6, see id. § 2B1.1, and added several enhancements. These included a twenty-level enhancement for amount of loss, see id. § 2B1.1(b)(1)(K); a two-level enhancement because the offenses of conviction involved more than ten victims, see id. § 2B1.1(b)(2)(A)(i); a four-level enhancement for Flete-Garcia's leadership role, see id. § 3B1.1(a); and a two-level enhancement because the money laundering convictions implicated 18 U.S.C. § 1956 , see USSG § 2S1.1(b)(2)(B). As an offset, the PSI Report recommended a two-level reduction for acceptance of responsibility. See id. § 3E1.1(a).

These calculations yielded a total offense level of 32 which, coupled with a criminal history category of III, produced a GSR of 151-188 months. To complete the picture, the PSI Report recommended restitution in the amount of $7,737,486.10.

Flete-Garcia objected to many aspects of the PSI Report, including (as pertinent here) the enhancements for number of victims and amount of loss. He also objected to the restitution amount.

On December 18, 2017, the district court denied Flete-Garcia's motion for an evidentiary hearing concerning amount of loss. Two days later, the court convened the disposition hearing. At that time, it heard and denied Flete-Garcia's remaining motions, including his motion to withdraw his guilty plea and his motion to compel discovery. The court also heard and rejected Flete-Garcia's renewed arguments as to why an evidentiary hearing would be useful in determining amount of loss.

The district court then turned to the task of fashioning Flete-Garcia's sentence. After entertaining additional arguments from both sides, the court accepted most of the guideline calculations limned in the PSI Report. The court, however, sustained the government's objection and ruled that Flete-Garcia - who had only pleaded guilty near the end of the trial and thereafter had sought to unbuckle himself from his guilty plea - was not entitled to an offense-level reduction for acceptance of responsibility.

The district court proceeded to sentence Flete-Garcia to a downwardly variant 132-month term of immurement and ordered him to make restitution in the amount of $7,737,486.10. This timely appeal followed. In it, Flete-Garcia calumnizes his conviction and sentence on several fronts. We start with his claim that he should have been allowed to withdraw his guilty plea. Next, we deal with his claims of sentencing-related error (including those arising out of the denial of his motions for discovery and for an evidentiary hearing). We then treat with his attack on the restitution order and end with his ineffective assistance of counsel claim.

II. WITHDRAWAL OF GUILTY PLEA

Because Flete-Garcia's motion to withdraw his guilty plea was filed before the imposition of sentence, it is governed by Federal Rule of Criminal Procedure 11(d)(2)(B). Under this rule, "[a] defendant may withdraw a plea of guilty ... after the court accepts the plea, but before it imposes sentence if ... the defendant can show a fair and just reason" for its withdrawal. Despite its permissive nature, this standard "does not endow [a defendant] with an unfettered right to retract a guilty plea." United States v. Merritt , 755 F.3d 6 , 9 (1st Cir. 2014). The devoir of persuasion rests with the movant, and we review the district court's denial of such a motion solely for abuse of discretion. See id.

A court's scrutiny of a plea-withdrawal motion must take into account the totality of the relevant circumstances. See id. ; see also United States v. Caramadre , 807 F.3d 359 , 366 (1st Cir. 2015). This canvass includes consideration of whether the plea was voluntary, intelligent, and knowing when tendered; the strength of the reason(s) proffered in support of the motion to withdraw; the timing of the request; and the force of any claim of actual innocence. See United States v. Dunfee , 821 F.3d 120 , 127 (1st Cir. 2016) (per curiam); Merritt , 755 F.3d at 9 . If the defendant makes a prima facie showing of an entitlement to relief, the court must then factor into the decisional calculus the prejudice, if any, that may accrue to the government as a result of allowing the plea to be withdrawn. See Merritt , 755 F.3d at 9 .

In the district court - as here - Flete-Garcia complained that his guilty plea was neither intelligent nor knowing because "he did not understand many of the important aspects of the Rule 11 hearing." The district court rejected this conclusory plaint, finding that the plea was suitably informed and not the product of any coercion. In the court's view, it was manifest that Flete-Garcia had fully considered the decision to change his plea, understood the nature and scope of his plea, and had not made the decision in the heat of the moment. Moreover, Flete-Garcia's proffered reason for withdrawing his plea was weak, especially since his decision to plead was reached after hearing the bulk of the government's evidence and an outline of the proof that remained. See Fernández-Santos , 856 F.3d at 16-17 . Finally, the court noted that Flete-Garcia had made no claim of actual innocence.

Flete-Garcia asserts that the district court's refusal to allow him to retract his guilty plea was an abuse of discretion because he was confused about the factual basis for counts 2 and 3 - and the district court compounded his confusion by "constrain[ing]" him to "short yes or no answers." This assertion is belied by the record. The transcript of the Rule 11 hearing makes pellucid that even though Flete-Garcia's responses to some of the district court's questions warranted further inquiry, the court conducted just such an inquiry. It patiently explained and re-explained the nature of the offenses to which Flete-Garcia was pleading and recounted the implications that would follow.

An example illustrates the district court's approach. When the court asked Flete-Garcia whether he agreed to the factual basis for counts 2 and 3 (specifically, that he had possession of two lists of stolen PII), Flete-Garcia replied that he did not "know where [the government] got that from." In response, the court reminded Flete-Garcia that he did not have to plead guilty, that he could elect to resume the trial, and that he had the option of pleading guilty to some charges and continuing to contest the others. The court also restated the charges that Flete-Garcia had questioned (counts 2 and 3) and summarized the evidence underpinning those charges. At that point, Flete-Garcia reaffirmed his desire to plead guilty and admitted to the factual basis for all of the charges.

In the last analysis, Rule 11 requires a district court to ensure that the defendant both knows and understands the nature of the charges to which he is pleading. See Fed. R. Crim. P. 11(b)(1)(G). This does not mean, though, that Rule 11 is off the table simply because a defendant indicates some uncertainty about the factual basis for a proposed guilty plea. Where, as here, the court resolves such uncertainties to the defendant's expressed satisfaction through clarification and explanation, a guilty plea may qualify as voluntary, intelligent, and knowing. See United States v. Ramos-Mejía , 721 F.3d 12 , 15 (1st Cir. 2013).

This is such a case, and Flete-Garcia offers no plausible basis for concluding that he did not fully understand the charges against him. 2 At any rate, all indications are to the contrary: he was present during jury selection and nearly four full days of trial, heard the prosecutor's opening statement and the testimony of nineteen of the government's twenty-three witnesses (including some who testified about the stolen PII), and listened as the prosecutor summarized the remaining evidence at the Rule 11 hearing. Additionally, the district court explained all of the charges to Flete-Garcia as well as the consequences of changing his plea.

We find hollow Flete-Garcia's protestation that he felt constrained by the district court to respond with yes or no answers. To be sure, the district court kept a rather tight rein on the colloquy - a commendable practice given that an empaneled jury was being held in limbo. But the court did no more than was reasonably necessary to keep the proceedings on track, and we discern no error in its management of the Rule 11 hearing. Viewed objectively, the court's dialogue with Flete-Garcia adroitly balanced its obligation to ensure that the plea was voluntary, knowing, and intelligent against the need for the fair and orderly administration of the Rule 11 hearing.

The short of it is that here, as in Dunfee , Flete-Garcia "affirmatively declared under oath at a properly conducted Rule 11 hearing that he was guilty of the crimes with which he was charged." 821 F.3d at 128 . In the absence of any plausible basis for discounting them, the district court was "entitled to give weight to the defendant's statements at his change-of-plea colloquy." United States v. Santiago Miranda , 654 F.3d 130 , 138 (1st Cir. 2011). On this record, we discern nothing approaching an abuse of discretion in the district court's determination that Flete-Garcia had failed to show a fair and just reason for withdrawing his plea. 3

III. CERTAIN SENTENCING RELATED MATTERS

Flete-Garcia offers up a salmagundi of claims relating to certain matters adjudicated in connection with the sentencing hearing. We subdivide our discussion of these claims into discrete segments.

A. Enhancement for Number of Victims .

We start with Flete-Garcia's challenge to the district court's application of a two-level enhancement for crimes involving ten or more victims. See USSG § 2B1.1(b)(2)(A)(i). For this purpose, a victim is described as "any individual whose means of identification was used unlawfully or without authority." Id. § 2B1.1, app. n.4(E)(ii). Because Flete-Garcia preserved this claim of error below, we review the district court's factual findings for clear error and its application of the law (including its application of the sentencing guidelines) de novo. See United States v. Carbajal-Váldez , 874 F.3d 778 , 782-83 (1st Cir. 2017), cert. denied , --- U.S. ----, 138 S. Ct. 2586 , 201 L.Ed.2d 303 (2018) ; United States v. Nuñez , 852 F.3d 141 , 144 (1st Cir. 2017).

We have made pellucid that "[c]lear error is not an appellant-friendly standard." Carbajal-Váldez , 874 F.3d at 783 . This demanding standard is satisfied only when, "upon whole-record-review, an inquiring court 'form[s] a strong, unyielding belief that a mistake has been made.' " Nuñez , 852 F.3d at 144 (alteration in original) (quoting United States v. Cintrón-Echautegui , 604 F.3d 1 , 6 (1st Cir. 2010) ). If two plausible but competing inferences may be drawn from particular facts, a sentencing court's choice between those two competing inferences cannot be clearly erroneous. See id. at 146 ; United States v. Ruiz , 905 F.2d 499 , 508 (1st Cir. 1990).

It is by now familiar lore that the government bears the burden of proving the applicability of a sentencing enhancement. See United States v. McCormick , 773 F.3d 357 , 359 (1st Cir. 2014). "It must carry this burden by a preponderance of the evidence." Id.

In this instance, it is nose-on-the-face plain that more than ten people were affected by Flete-Garcia's scheme. After all, Flete-Garcia stole PII relating to hundreds of individuals, used the stolen PII corruptly, and pleaded guilty to no fewer than seventeen counts of aggravated identity theft. Undaunted by these ironclad facts, Flete-Garcia rests his claim of error on language in the sentencing guidelines. With respect to aggravated identity theft, the guidelines link the term of immurement to the statute of conviction. See USSG § 2B1.6(a). In this case, the statute requires a mandatory two-year term for a defendant who "knowingly transfers, possesses, or uses, without lawful authority, a means of identification of another person." 18 U.S.C. § 1028A.

But there is a rub: an application note instructs that "[i]f a sentence ... is imposed in conjunction with a sentence for an underlying offense," the sentencing court should "not apply any specific offense characteristic for the transfer, possession, or use of a means of identification" in fashioning the sentence for the underlying offense. USSG § 2B1.6, app. n.2. In other words, the application note prohibits the enhancement of a sentence for the same conduct (the transfer, possession, or use of a means of identification) that is captured by the statute.

Flete-Garcia asserts that the number of victims enhancement violates the letter (or at least the spirit) of the application note. We do not agree. Even though we have recognized that the prohibition contained in section 2B1.6 may preclude some enhancements germane to the underlying offense, see United States v. Jones , 551 F.3d 19 , 25 (1st Cir. 2008), Flete-Garcia's assertion is unfounded.

The key is whether the proposed enhancement relates to a characteristic of the offense. See United States v. Sharapka , 526 F.3d 58 , 62 (1st Cir. 2008). If so, it is precluded. See id. Otherwise, it is not precluded. See id. Flete-Garcia submits that the number of victims enhancement falls into the category of prohibited enhancements because the sole reason for its application was that he used a means of identification to perpetrate his crimes. Flete-Garcia, however, takes too myopic a view.

Here, the number of victims enhancement does not punish him simply for using a means of identification but, rather, punishes him for the breadth of his criminality, that is, for using the means of identification of ten or more individuals in the course of his criminal activity.

Section 2B1.6 prohibits an enhancement that is based on the nature of the offense (transferring, possessing, or using a means of identification). See USSG § 2B1.6, app. n.2; see also Sharapka , 526 F.3d at 62 . The guideline provision, though, is silent as to an enhancement based on the breadth of the offense. See United States v. Gonzales , 844 F.3d 929 , 932 (10th Cir. 2016) (observing that application note 2 prohibits an enhancement based on nature of crime but not one "based on the extent of such a crime" (emphasis in original)). In the case at hand, the number of victims enhancement plainly has its genesis in the breadth of Flete-Garcia's criminal activity. That is an appropriate use of the enhancement. See id. at 933 (explaining that enhancement accomplishes task of treating "criminal conduct more seriously as the number of stolen identities increases"). Such an interpretation is consistent with the structure of the enhancement itself, which ratchets up as the number of victims and the extent of the impact on them increases. Compare , e.g. , USSG § 2B1.1(b)(2)(A)(i) (increasing offense level by two for ten or more victims), with , e.g. , id. § 2B1.1(b)(2)(C) (increasing offense level by six for substantial financial hardship to twenty-five or more victims).

So, too, the district court's application of the enhancement is consistent with the text of 18 U.S.C. § 1028A. While a conviction for aggravated identity theft mandates a two-year sentence, the statute does not factor the number of victims into the calculus of punishment.

Our view of the application note aligns us with our sister circuits. All of the circuits that have considered the matter have rejected claims of sentencing error similar to the claim that Flete-Garcia advances. See , e.g. , Gonzales , 844 F.3d at 932-33 ; United States v. Ford , 784 F.3d 1386 , 1397-98 (11th Cir. 2015) ; United States v. Anderson , 532 F. App'x 373 , 378 (4th Cir. 2013) ; United States v. Lyles , 506 F. App'x 440 , 447 (6th Cir. 2012) ; United States v. Yummi , 408 F. App'x 537 , 541 (3d Cir. 2010). By the same token, it fits comfortably with our earlier cases considering the interplay between other enhancements and USSG § 2B1.6. See , e.g. , Jones , 551 F.3d at 25 ; (considering interplay between sections 2B1.1(b)(10)(B) and 2B1.6 ); Sharapka , 526 F.3d at 62 (considering interplay between sections 2B1.1(b)(10)(A)(i), (C)(i) and 2B1.6 ).

We conclude that the statute and the guideline provision, read together, present no barrier to the application of the number of victims enhancement in this case. As applied here, the enhancement punished Flete-Garcia for the overall breadth of his criminal activity - a factor not captured by the statute of conviction and, thus, not foreclosed by application note 2.

That ends this aspect of the matter. Because we hold that the district court did not clearly err in imposing a two-level enhancement for the presence of ten or more victims, Flete-Garcia's claim of error founders.

B. Enhancement for Amount of Loss .

Flete-Garcia posits that the district court erred in calculating the amount of loss attributable to the offenses of conviction. In this regard, he notes that the district court accepted the government's calculation of both the amounts refunded on fraudulent tax-return claims and the amounts sought (though not paid) on other fraudulent refund claims. In a nutshell, he argues that these loss calculations were errant: they were based on the flawed testimony of witnesses, unreliable information, and unsound auditing methods.

We begin with bedrock. "In a fraud case resulting in financial loss, the defendant's guideline sentencing range is determined in part" by the amount of loss. United States v. Naphaeng , 906 F.3d 173 , 179 (1st Cir. 2018), cert. denied , --- U.S. ----, 139 S. Ct. 1233 , 203 L.Ed.2d 246 (2019). For this purpose, "loss is the greater of actual loss or intended loss." USSG § 2B1.1, app. n.3(A). Since intended loss normally subsumes actual loss, intended loss is often the greater of the two. So it is here, 4 and intended loss is defined as "the pecuniary harm that the defendant purposely sought to inflict," whether or not achievable. Id. § 2B1.1, app. n.3(A)(ii).

The government must prove the amount of loss by a preponderance of the evidence. See United States v. Curran , 525 F.3d 74 , 78 (1st Cir. 2008). In arriving at a loss figure, a sentencing court is free to consider both losses stemming directly from the conduct underlying the offenses of conviction and losses stemming from any relevant conduct (including uncharged conduct). See id. ; United States v. Ranney , 298 F.3d 74 , 80-81 (1st Cir. 2002) ; United States v. Sklar , 920 F.2d 107 , 110 (1st Cir. 1990).

The sentencing court has considerable discretion in determining what evidence should be regarded as reliable in making findings as to the amount of loss. See Sklar , 920 F.2d at 110 . Such evidence may come from "virtually any dependable [source of] information." Id. It is common ground that a sentencing court's loss calculations are entitled to "appropriate deference," given the court's "unique position to assess the evidence and estimate the loss based upon that evidence." USSG § 2B1.1, app. n.3(C). Similarly, "credibility determinations lie within the domain of the district court," and "[o]nly rarely - and in the most urgent circumstances - will we, from the vista of a sterile appellate record, meddle in such matters." United States v. St. Cyr , 977 F.2d 698 , 706 (1st Cir. 1992).

We add, moreover, that a loss calculation need not be precise: the sentencing court need only make a reasonable estimate of the range of loss. See Curran , 525 F.3d at 78 . This latitude comports with the way in which amount of loss relates to a defendant's sentence: the extent of the "loss" enhancement corresponds to a range of loss amounts. See USSG § 2B1.1(b)(1). Here, the district court's loss calculation totaled approximately $12.7 million, leaving a cushion of more than $3 million over the minimum amount ($9.5 million) needed to trigger the twenty-level enhancement. See id. § 2B1.1(b)(1)(K), (L) (directing twenty-level enhancement for loss exceeding $9.5 million and up to $25 million). Thus, the same enhancement results whether the loss falls at the bottom of the range, at the top of the range, or somewhere in the middle.

Flete-Garcia pleaded guilty to converting seventeen particular tax-refund checks (totaling $125,756) for his own use. At sentencing, the government contended that this was the tip of the iceberg, and that it constituted a mere fraction of the overall harm wrought by Flete-Garcia. The district court agreed: it was "satisfied" that "$7.7 million in checks" had been "paid out" by the IRS in response to fraudulent returns instigated by Flete-Garcia. The court was also satisfied that approximately $5 million in "blocked" refunds (that is, refunds claimed but not paid) qualified as intended loss. In making these findings, the court credited the trial testimony of Flete-Garcia's co-conspirators and a testifying IRS agent, stating that it believed the witnesses' descriptions of the scheme and accepted the IRS agent's method of calculating the losses stemming from that scheme.

In reviewing these determinations, we start by taking a closer look at the $7.7 million in losses stemming from fraudulent tax-refund claims actually paid out by the IRS. To cash the checks that he received, Flete-Garcia engaged a pair of co-conspirators, Carmen Guzman and Rocio Dominguez (both of whom operated businesses that provided check-cashing services). Guzman and Dominguez testified that they began handling checks for Flete-Garcia in 2007 and 2009, respectively. Flete-Garcia's arrangements with the two women were similar: he would give tax-refund checks to one of them, she would deposit the checks in one of her accounts, and would then give Flete-Garcia a sum equal to the aggregate face amount of the checks, less an agreed "cut." Flete-Garcia did not appear as a named payee on any of the stacks of checks (sometimes as many as forty at a time) that he delivered to Guzman and Dominguez. Guzman testified that she never took a fraudulent tax-refund check from anyone other than Flete-Garcia. Dominguez, though, testified that she accepted fraudulent tax-refund checks not only from Flete-Garcia but also from another fraudster. She added that she could "count ... in [her] hand" the total number of fraudulent checks she received from this other fraudster.

According to both Guzman and Dominguez, their complicity in Flete-Garcia's scheme continued until their accounts were frozen (in or around 2012). The freezing of these accounts did not deter Flete-Garcia: he simply recruited a third check-cashing co-conspirator, Dubin Gonzalez-Pabon (Gonzalez). Gonzalez testified that he, along with another person, picked up tax-refund checks from Flete-Garcia at a barbershop and gave Flete-Garcia the bulk of the proceeds. At the same time, Gonzalez was involved in another tax-fraud racket (which the PSI Report classified as a "related case").

In preparation for trial, an IRS agent, Tuan Nguyen, surveyed all of the co-conspirators' bank accounts for evidence of deposited tax-refund checks. While Flete-Garcia's guilty plea pretermitted Nguyen's scheduled trial testimony, Nguyen nonetheless identified approximately 1,400 fraudulently obtained checks and catalogued them in summary charts submitted at sentencing. These summary charts identified approximately $7.7 million in monies actually paid out by the IRS in the form of fraudulently obtained tax-refund checks - and each check corresponded to a tax return actually filed.

At sentencing, Flete-Garcia assailed the government's loss calculations by pointing to purported anomalies in the evidence. In his view, these anomalies compelled the conclusion that "much of the data and documents used by the Government and the Probation Office were flawed and contained information that did not support the loss figures." Moreover, he argued that "the checks could not have been printed and sent from the U.S. Treasury in the manner described by the Government."

The district court considered all of Flete-Garcia's arguments. In the end, the court had no difficulty in attributing the claimed $7.7 million in actual loss to Flete-Garcia. The court credited the government's explanation of Nguyen's methodology and supportedly found his summary charts reliable. Noting that checks identified by the government "match[ed] tax returns that were actually filed with the IRS," the court concluded that the actual loss figure - $7.7 million - was "real money, actually paid" by the IRS and was attributable to Flete-Garcia. In reaching this conclusion, the court deemed it highly significant that the monies paid out were traceable to checks in the co-conspirators' accounts. As the court put it, "[t]hat's what those witnesses said" and "I believed them."

The record makes manifest that the district court did not blindly accept the government's reconstruction of the pertinent events. To the contrary, the court gave due consideration to Flete-Garcia's compendium of purported evidentiary anomalies. The court observed that Flete-Garcia raised "some things around the margin," but found that none of the matters mentioned by Flete-Garcia shook its faith in the credibility of the government's witnesses.

In this venue, Flete-Garcia argues that the district court's findings were riddled with error. He continues to insist that the evidence on which the district court relied was faulty because, among other things, the checks (or at least some of them) were altered; the mailing addresses for some checks did not match the government's theory of the case; multiple checks were sometimes issued for the same taxpayer in a single tax year; the amounts of a few refund checks were not entirely congruent with the underlying tax returns; the checks cashed with Dominguez's help were not altogether consistent with her testimony; and the government failed to offer any co-conspirator testimony pertaining to e-tax returns. These arguments, though multifarious, boil down to - in his words - an all-out assault on "the reliability of the cooperating witnesses, as well as the documentary evidence."

But for the most part, Flete-Garcia is firing blanks. To begin, his arguments give unduly short shrift to "the time-tested tenet that 'credibility determinations are part of the sentencing court's basic armamentarium.' " United States v. Bernier , 660 F.3d 543 , 546 (1st Cir. 2011) (quoting United States v. Platte , 577 F.3d 387 , 392-93 (1st Cir. 2009) ). We will overturn such determinations "only if we have a definite and firm conviction that a mistake has been committed," id. (quoting United States v. González-Vélez , 587 F.3d 494 , 504 (1st Cir. 2009) ), and the record in this case gives rise to nothing resembling such a conviction. After all, "the sentencing judge presided over the trial and was in an enviable position to gauge [the witnesses'] credibility and to separate wheat from chaff." Id. Flete-Garcia has identified nothing that leads us to believe that "a reasonable factfinder would not credit" the government's witnesses and evidence. Anderson v. City of Bessemer City , 470 U.S. 564 , 575, 105 S.Ct. 1504 , 84 L.Ed.2d 518 (1985).

It would serve no useful purpose to dissect Flete-Garcia's wide-ranging critique, epithet by epithet. A few examples, though, will help to illustrate why those criticisms are nowhere near as potent as Flete-Garcia would have us believe.

Flete-Garcia asserts that Dominguez testified that she wrote by hand on every check that he gave her after she verified that it was really an IRS tax-refund check and, thus, the checks in evidence from her accounts (which were devoid of her handwriting) were unreliable. But Flete-Garcia is painting with too broad a brush: Dominguez testified that she followed this practice only "in the beginning" of her criminal partnership with Flete-Garcia. Her testimony further suggested that her business practices evolved in response to the burgeoning volume of checks that he delivered, thus paving the way for a reasonable inference that Dominguez had abandoned the time-consuming practice of verifying each check by the time the government's investigation commenced.

Another example makes a similar point. Flete-Garcia harps on the fact that the government's charts included some instances of multiple refund checks issued for the same social security number in the same year. He complains that the government offered no explanation for how this phenomenon could have occurred when the IRS screens for duplicate social security numbers. But as one of the government's witnesses explained, a refund check could get issued for a bogus social security number if the error was "not accounted for in time." We think that this testimony, although specifically referencing fictitious social security numbers rather than duplications, is sufficient to ground a reasonable inference that fraud detection is not always automatic. So viewed, the presence of duplicate social security numbers did not render the government's charts wholly unreliable.

A third example deals with Flete-Garcia's complaint about the government's failure to offer particularized evidence about e-tax returns. Although the government's exhibits did include some e-tax returns, nothing in the record furnishes a basis for a founded claim that some special sort of proof was required with respect to those returns. The testimony that the government offered to validate the exhibits perforce validate the e-tax returns as well.

To sum up, the district court had wide discretion to "evaluate virtually any dependable information" and determine the probative value of such information with respect to issues material to sentencing. United States v. Bradley , 917 F.2d 601 , 605 (1st Cir. 1990). Although some of the discrepancies identified by Flete-Garcia might plausibly suggest problems with a few of the items on which the district court relied, such an inference was by no means compelled. And as we have said, "when there are two plausible views of the record, the sentencing court's adoption of one such view cannot be clearly erroneous." St. Cyr , 977 F.2d at 706 . We conclude, therefore, that the district court's finding of $7.7 million in losses actually paid was not clearly erroneous.

This brings us to the approximately $5 million in intended loss attributed to "blocked" refunds. The facts undergirding this finding came largely from the testimony of yet another co-conspirator: Juan Santiago. Santiago recalled that Flete-Garcia gave him a computer, the PII, a list of addresses, and a list of employer identification numbers. Santiago then was instructed how to plug this information into tax-preparation software and how to prepare tagalong W-2 forms (ostensibly issued by "any number of companies"). He worked his part of the scheme over the course of several years, but on two occasions sold PII to a person who turned out to be a government cooperator.

With the lists of PII in evidence, an IRS agent, Richard Adams, described the methodology that he used to identify fraudulently filed tax-refund claims deriving from these lists. In general terms, Adams compared each filed tax return corresponding to a name in the PII to independent information obtained from an IRS database. When he spotted a mismatch, he classified the return as fraudulent. He then prepared summary charts, which catalogued approximately $5 million in bogus "blocked" claims.

The sentencing court was intimately familiar with this testimony: it not only presided over the trial but also presided over the change-of-plea hearing (at which Flete-Garcia admitted his guilt with respect to, inter alia, access device fraud arising out of his possession of the two lists of PII). The court made an express finding that Santiago "was a truthful witness." Similarly, it found Adams' methodology (and, thus, his summary charts) reliable.

Even so, the court fretted over whether all $5 million could fairly be attributed to Flete-Garcia given that Santiago had admitted to doing some "freelancing." In the end, though, the court accepted the government's explanation that Santiago's freelancing was limited to the sale of the lists of PII and of checks not attributable to Flete-Garcia. Relatedly, the court found that Santiago consistently used the lists to advance Flete-Garcia's scheme by "prepar[ing] returns that were rejected by the IRS." Given these findings - and taking into account that the government had not attributed all the losses from Santiago's activities to Flete-Garcia - the court concluded that the government had carried its burden of proving by preponderant evidence an intended loss of $5 million referable to the "blocked" refunds.

Flete-Garcia tries to sidestep this hurdle, taking issue with both Santiago's veracity and the related amount of loss. To bolster his arguments, he points out that the summary charts contain addresses across the United States. He says that when Santiago sold the PII, he did not sell the accompanying list of addresses and, in any event, the geographic diversity of addresses contradicts Santiago's testimony that the tax refunds obtained were mostly directed to post office boxes in Massachusetts and Louisiana. He adds that the W-2 forms reflected putative employers different than those identified by Santiago.

These arguments lack force. Santiago's testimony, found credible by the district court, forges a direct link between Flete-Garcia and the PII. To seal the deal, the record shows with conspicuous clarity that the IRS used the PII from the lists - lists that Flete-Garcia pleaded guilty to possessing - to identify the suspect tax returns. While Santiago admitted to freelancing, the court reasonably credited the government's explanation that Santiago's separate activities were not included in the loss calculation performed for Flete-Garcia. Viewing the record as a whole, the sentencing court had more than enough reliable information to ground its finding, by a preponderance of the evidence, that the $5 million loss was attributable to Flete-Garcia.

We add a coda. In this case, the sentencing court was operating with a substantial cushion. The court found the aggregate amount of loss to be approximately $12.7 million, even though all that was needed to justify the twenty-level enhancement was a loss amount of $9.5 million. It follows that Flete-Garcia cannot demonstrate reversible error with respect to the amount-of-loss enhancement simply by nibbling around the edges of the district court's findings: he must persuade us that at least $3.2 million of the loss attributed to him by the court was the product of clear error. We are not so persuaded.

Mathematics is an exact science. Engineering is an exact science. But calculating amount of loss under the sentencing guidelines is far from an exact science. In fraud cases, amount of loss is meant to be a proxy for the harm (both actual and intended) inflicted by the fraudster's nefarious activities. All that is required is a reasonable estimate of the amount of loss. See Curran , 525 F.3d at 78 . Here, the district court's findings, which spell out how and why the amount of loss attributable to Flete-Garcia comfortably exceed the $9.5 million threshold, easily pass muster. Clear error is clearly absent.

C. Discovery .

Three days before his scheduled sentencing, Flete-Garcia moved to compel production of materials regarding the IRS agent, Jamie Clarke, who had overseen his case. Flete-Garcia averred that, after he had pleaded guilty, Clarke was the subject of an assault allegation reported in the Boston Globe and that everything pertaining to the allegation was material to his case. The government opposed the motion, and the district court denied it.

We review the denial of a motion to compel discovery in a criminal case for abuse of discretion. See United States v. Caro-Muñiz , 406 F.3d 22 , 29 (1st Cir. 2005). Typically, the government is obliged "to disclose evidence in its possession that is favorable to the accused and material to guilt or punishment." United States v. Prochilo , 629 F.3d 264 , 268 (1st Cir. 2011) (citing Brady v. Maryland , 373 U.S. 83 , 87, 83 S.Ct. 1194 , 10 L.Ed.2d 215 (1963) ). To prevail on a motion to compel, the defendant must show a likelihood of prejudice stemming from the government's nondisclosure. See United States v. Rosario-Peralta , 199 F.3d 552 , 559 (1st Cir. 1999). Thus, when a criminal defendant seeks discovery, he must "be able to articulate with some specificity what evidence he hopes to find in the requested materials, why he thinks the materials contain this evidence, and finally, why this evidence would be both favorable to him and material." Prochilo , 629 F.3d at 269 . 5

Clarke did not testify at Flete-Garcia's trial. Nevertheless, Flete-Garcia says that the requested information was material because Clarke was present at trial, answered some questions for the district court (out of the jury's earshot), served as the lead case agent, and had access to all of the information that the government used to calculate the loss amounts.

The district court denied Flete-Garcia's motion, determining that the government's investigation of Clarke was irrelevant and immaterial because that investigation did not bear on Flete-Garcia's case. The court noted that Clarke neither compiled the summary charts that reflected the loss computations nor testified at trial. And the assault allegation - though serious - was wholly unrelated to Flete-Garcia's case.

The government concedes that Flete-Garcia's lawyer spoke with the prosecutor before sentencing and asked for "all information regarding prior complaints against Special Agent Clarke and all information regarding the current investigation of Special Agent Clarke." Thus, the question reduces to whether the district court abused its discretion in failing to find that the requested information was likely to be both favorable to Flete-Garcia and material to his case.

We discern no abuse of discretion in the district court's denial of the motion to compel discovery. Flete-Garcia submits that "the allegations against ... Clarke indicate a life unraveling" and that (given the other questions raised about the reliability of the government's evidence) "the integrity of the agent who was responsible for that evidence is material." But the allegation against Clarke was simply that: a bare allegation. More importantly, Flete-Garcia's quest for discovery had all the earmarks of an old-fashioned fishing expedition. He never explained - apart from rank speculation - how information about that allegation might have altered the course of the sentencing proceeding or otherwise affected his case. Where, as here, a government agent is alleged to have committed misconduct unrelated to an earlier investigation that he supervised, such an allegation, without more, does not render the earlier investigation suspect. Cf. United States v. Nelson-Rodriguez , 319 F.3d 12 , 35 (1st Cir. 2003) (finding no Brady violation when "government failed to produce ... prior investigations which contained allegations that [confidential informant] was involved in ... murders" because his "alleged role in [those] murders was a collateral matter").

To say more on this point would be to paint the lily. A defendant ordinarily must make some showing of prejudice before an appellate court will step in and overrule a district court's reasoned decision to deny discovery, see Rosario-Peralta , 199 F.3d at 559 , and Flete-Garcia has made no such showing here. Flete-Garcia's theory of materiality is based entirely on conjecture and, in such circumstances, a district court's refusal to compel production of requested information is not an abuse of discretion. See United States v. Goris , 876 F.3d 40 , 45 (1st Cir. 2017), cert. denied , --- U.S. ----, 138 S. Ct. 2011 , 201 L.Ed.2d 267 (2018).

D. Evidentiary Hearing .

At sentencing, Flete-Garcia moved for an evidentiary hearing, suggesting that one was needed to reconcile the parties' competing narratives about the amount of loss. The district court denied his motion, and Flete-Garcia assigns error.

A criminal defendant, facing the imposition of sentence, is not entitled to an evidentiary hearing on demand. See United States v. DeCologero , 821 F.2d 39 , 44 (1st Cir. 1987) (explaining that "hearings cannot be convened at the whim of a suitor, made available like popsicles in July, just because a passerby would like to have one"). The decision as to whether to hold such a hearing "is left to the sound discretion of the district court." United States v. Brown , 621 F.3d 48 , 57 (1st Cir. 2010). At sentencing, evidentiary hearings are the exception, not the rule - and an order denying an evidentiary hearing is reviewed only for abuse of discretion. See United States v. Shattuck , 961 F.2d 1012 , 1014-15 (1st Cir. 1992).

In gauging the need for an evidentiary hearing, we consider whether the defendant made "a sufficient threshold showing that material facts [were] in doubt or in dispute." United States v. McAndrews , 12 F.3d 273 , 280 (1st Cir. 1993) (quoting United States v. Panitz , 907 F.2d 1267 , 1273 (1st Cir. 1990) ). Before the district court, Flete-Garcia claimed that a hearing was needed because "the evidence in this case does not line up with the sworn testimony, and there are evidentiary anomalies that do not comport with the [g]overnment's claims at trial." In his view, "[w]itnesses need[ed] to be subjected to cross-examination on the critical problems with the evidence that underlies the loss calculation."

The district court disagreed, denying the motion and holding that Flete-Garcia did "not identif[y] disputed issues of material fact." The court noted, however, that it would be open to reconsidering its decision depending on what transpired at the disposition hearing.

During the disposition hearing, Flete-Garcia once again tried to convince the district court of the desirability of an evidentiary hearing. He emphasized the evidentiary anomalies to which he earlier had referred by, for example, pointing to some instances in which there was a mismatch between the issue date and the deposit date of particular checks. The district court acknowledged that the date discrepancy was a "fair point[ ]," but concluded that no additional testimony was necessary because "the vast majority, if not all of" the dates matched.

On appeal, Flete-Garcia maintains that there were "clear problems with the evidence" and that, without a hearing, he was "hamstrung in exploring these problems." In support, he relies on a guideline provision instructing that "the parties shall be given an adequate opportunity to present information to the court" regarding a factor that is "reasonably in dispute." USSG § 6A1.3(a).

Flete-Garcia's reliance on section 6A1.3 is mislaid. In terms, this provision is meant to ensure that, at sentencing, a defendant is "given an adequate opportunity" to be heard. United States v. Gerante , 891 F.2d 364 , 367 (1st Cir. 1989) (quoting USSG § 6A1.3(a) ). But an "adequate opportunity" is not always synonymous with an evidentiary hearing, see id. , and Flete-Garcia had a more-than-adequate opportunity to be heard and to question the evidence germane to sentencing.

For one thing, much of the evidence relating to loss was presented at trial, and Flete-Garcia had the opportunity to cross-examine the witnesses who presented it (including Adams, Guzman, Dominguez, Gonzalez, and Santiago). For another thing, Flete-Garcia's counsel was able to argue extensively at the disposition hearing about perceived problems with the government's evidence. In the circumstances of this case, no more was exigible to satisfy the "adequate opportunity" requirement.

We acknowledge that some of the questions raised by Flete-Garcia were nuanced. But in McAndrews , we recognized that "we have consistently abjured mandatory evidentiary hearings in a wide variety of ... delicate [and] nuanced situations." 12 F.3d at 279 . Here, the "evidentiary anomalies" mentioned by Flete-Garcia dealt mainly with peripheral matters and did not cast any significant shadow over the government's proof. Moreover, the district court was in an excellent position to weigh these alleged anomalies and gauge their import even without an evidentiary hearing. As we previously have noted, the district court had presided over the trial and was intimately familiar with the evidence before it. We must, therefore, accord substantial deference to its determination that an evidentiary hearing would not have been productive. See id. at 280 ; Shattuck , 961 F.2d at 1015 ; cf. United States v. Cannons Eng'g Corp. , 899 F.2d 79 , 94 (1st Cir. 1990) (warning that "[d]istrict courts are busy places and makework hearings are to be avoided"). Giving the district court its due, we discern no abuse of the court's broad discretion in its denial of Flete-Garcia's motion for an evidentiary hearing.

E. The Due Process Claim .

Flete-Garcia asserts, for the first time on appeal, that his due process rights were violated by the district court's reliance on false and misleading evidence. 6 "The plain error hurdle is high." United States v. Hunnewell , 891 F.2d 955 , 956 (1st Cir. 1989). To prevail under plain error review, an appellant must show "(1) that an error occurred (2) which was clear or obvious and which not only (3) affected the defendant's substantial rights, but also (4) seriously impaired the fairness, integrity, or public reputation of judicial proceedings." United States v. Duarte , 246 F.3d 56 , 60 (1st Cir. 2001).

The Due Process Clause protects a defendant by, inter alia, safeguarding against a sentence predicated on information that is "false or materially incorrect." United States v. Curran , 926 F.2d 59 , 61 (1st Cir. 1991). We have cautioned, however, that "the due process right at sentencing is not as robust as the due process right at trial." United States v. Stile , 845 F.3d 425 , 430 (1st Cir. 2017). Even so, "due process ... requires that the defendant be given an adequate opportunity to refute information relied on at sentencing." Id. (quoting United States v. Wilfred Am. Educ. Corp. , 953 F.2d 717 , 722 (1st Cir. 1992) ). Relatedly, due process demands that a sentencing court "consider all the available evidence, including conflicting evidence" to "assure itself that a piece of proof is sufficiently reliable." United States v. Tavano , 12 F.3d 301 , 305 (1st Cir. 1993).

Here, Flete-Garcia was unquestionably on fair notice of all of the facts that the government deemed relevant to the imposition of sentence. Cf. United States v. Berzon 941 F.2d 8 , 19-20 (1st Cir. 1991) (holding that due process requires that defendant be given fair notice of conduct and facts that will inform the sentencing court's determinations). Many of them had been explored during the nearly four days of trial, and others had surfaced either at the Rule 11 hearing or in the PSI Report. Some of these facts may have been arguable, but none of them was plainly false. Moreover, Flete-Garcia was given wide latitude in his attempts to poke holes in the government's factual mosaic, and he was able to develop his argument that some other fraud, independent of his own criminality, was afoot.

For aught that appears, Flete-Garcia received all the process that was due. His claim of error therefore stumbles at the first step of the plain error inquiry.

IV. RESTITUTION

Flete-Garcia's next plaint builds on one of his earlier plaints. He points to arguments that he marshalled in reproving the sentencing court's findings with respect to amount of loss, see supra Part III(B), and asserts that those arguments "make clear that the ... restitution order is flawed." This bareboned assertion does not get him very far.

The district court imposed the restitution order pursuant to the Mandatory Victims Restitution Act (MVRA), 18 U.S.C. § 3663A. "[T]he MVRA mandates that a defendant convicted of certain federal crimes, including those 'committed by fraud or deceit,' must make restitution to victims commensurate with ... actual losses." Naphaeng , 906 F.3d at 179 (quoting 18 U.S.C. § 3663A(c)(1)(A)(ii) ). Although restitution typically bears a relationship to amount of loss, the two are conceptually distinct. In fraud cases, the amount of loss is an integer in the sentencing calculus: it is computed in order to establish the defendant's offense level and, thus, his GSR. Viewed in context, then, amount of loss is "the primary metric by which 'the seriousness of the offense and the defendant's relative culpability' are measured." United States v. Alphas , 785 F.3d 775 , 782 (1st Cir. 2015) (quoting USSG § 2B1.1, cmt. (bckg'd)).

Restitution is a horse of a different hue, serving a wholly different purpose. It "is designed to compensate the victim, not to punish the offender." Naphaeng , 906 F.3d at 179 . Moreover, restitution deals exclusively with losses actually sustained and (unlike amount of loss calculations) makes no provision for intended loss. See Alphas , 785 F.3d at 786 . To that end, the MVRA directs that restitution shall be made "to each victim in the full amount of each victim's losses as determined by the court." 18 U.S.C. § 3664 (f)(1)(A). And whereas loss calculations require only a reasonable estimate of the range of loss, see Curran , 525 F.3d at 78 , the entire amount of restitution must be supported - albeit only by "a modicum of reliable evidence," United States v. Vaknin , 112 F.3d 579 , 587 (1st Cir. 1997).

We generally review preserved challenges to restitution orders for abuse of discretion. See Naphaeng , 906 F.3d at 179 . Unpreserved challenges are, at best, reviewed for plain error. See United States v. Sánchez-Maldonado , 737 F.3d 826 , 828 (1st Cir. 2013). Here, Flete-Garcia objected to the restitution section in the PSI Report, but he did not renew this objection either in his sentencing memorandum or at the disposition hearing. We thus treat his challenge to the restitution order as forfeited and review it only for plain error. 7 See Duarte , 246 F.3d at 60 .

Flete-Garcia's argument with respect to restitution is nebulous. His opening brief contends, in conclusory terms, that "[t]he problems" he had identified with respect to the government's loss calculations "make clear that the evidence relied upon by the [district] court for the restitution order is flawed and thus insufficient to support the restitution order." His reply brief reiterates that "the loss figure [is] incorrect" and that "[t]he loss calculation is what drove the restitution amount." In effect, then, he tries to copy and paste his loss-calculation arguments - arguments that we already have rejected, see supra Part III(B) - in support of his restitution claim. He offers nothing that indicates an awareness of the differing criteria for loss calculations and restitution calculations, nor does he attempt to tailor his amount of loss arguments to the restitution context. At the end of the day, Flete-Garcia has left "the court to do counsel's work, create the ossature for the argument, and put flesh on its bones." United States v. Zannino , 895 F.2d 1 , 17 (1st Cir. 1990).

Flete-Garcia's failure to tailor his arguments to the restitution context drains them of any force. As we have explained, there are significant differences between the proper method of calculating loss and the proper method of calculating restitution - a fact that is made starkly apparent by the substantial dollar differential between the district court's loss calculation (approximately $12.7 million) and its restitution award (approximately $7.7 million). Flete-Garcia could have shaped his restitution argument accordingly. For example, he could have argued (perhaps successfully) that the handful of checks Dominguez admitted to cashing for another individual could not be included in computing the restitution order because that other fraudster's victims were not "directly and proximately harmed" by Flete-Garcia's malefactions. 18 U.S.C. § 3663A(a)(2). But Flete-Garcia did not advance that argument, either in the district court or on appeal, much less make any attempt to quantify the monetary value of the handful of checks. Given his default, there is no reason why we should make such an argument for him. See Zannino , 895 F.2d at 17 (warning that "a litigant has an obligation 'to spell out its arguments squarely and distinctly,' or else forever hold its peace" (quoting Rivera-Gomez v. de Castro , 843 F.2d 631 , 635 (1st Cir. 1988) )); cf. Paterson-Leitch Co., Inc. v. Mass. Mun. Wholesale Elec. Co. , 840 F.2d 985 , 989 (1st Cir. 1988) ("Courts, like the Deity, are most frequently moved to help those who help themselves.").

Cognizant that a "court's calculation of restitution is not held to standards of scientific precision," Sánchez-Maldonado , 737 F.3d at 828 , we discern no plain error in the district court's restitution order. The government introduced testimony from multiple witnesses and detailed charts substantiating the $7.7 million paid out by the IRS. The charts delineated particular amounts of money, actually expended, based on each co-conspirator's bank records. The district court concluded that this evidence laid a solid foundation for its restitution order.

In the circumstances of this case, the standard of review is dispositive. As said, Flete-Garcia's challenges to the restitution order are reviewable only for plain error - and "[t]he proponent of plain error must carry the devoir of persuasion as to each of the four elements that collectively comprise the plain error standard." United States v. Bramley , 847 F.3d 1 , 5 (1st Cir. 2017). Even if we assume for the sake of argument that there may have been some error (perhaps an obvious error) in the district court's restitution calculation, Flete-Garcia has made no argument regarding the extent to which any such error affected that calculation. He has not shown, for example, that the restitution order is off-target by, say, $100,000, or $50,000, or $5. Instead, he makes generalized allegations that inaccuracies permeate the restitution order - and he leaves it to us to do the juridical equivalent of an archeological dig and monetize his allegations.

We have admonished before that parties act at their peril in leaving "the court to do counsel's work," Zannino , 895 F.2d at 17 , and we are reluctant to reward such tactics. The third element of the plain error inquiry requires that an appellant demonstrate "a reasonable probability that, but for the error, the outcome would have been different." Bramley , 847 F.3d at 7 . Flete-Garcia has made no such demonstration but, rather, leaves us to work out how the claimed errors - even if they apply in the restitution context - may affect the amount of restitution that the district court ordered. This causal approach to appellate advocacy - throwing generalized scraps of information on the table and hoping that the panel will make a meal of them - falls well short of what the third element of plain error demands.

Given Flete-Garcia's utter failure to carry his burden with respect to the third element of the plain error inquiry, the fourth element also cuts against him. We think it evident that the claimed errors, if left uncorrected, will not "seriously impair[ ] the fairness, integrity, or public reputation of judicial proceedings." Duarte , 246 F.3d at 60 .

The short of it is that we find no plain error. Therefore, we affirm the district court's restitution order.

V. INEFFECTIVE ASSISTANCE OF COUNSEL

In a final attempt to improve his lot, Flete-Garcia insists that the representation provided by his trial counsel was ineffective. See U.S. Const. amend. VI ; see also Strickland v. Washington , 466 U.S. 668 , 687, 104 S.Ct. 2052 , 80 L.Ed.2d 674 (1984). He offers vignettes illustrating his counsel's allegedly deficient performance - ignoring him, withholding discovery materials from him, and failing to make certain inquiries when cross-examining witnesses - and says that, as a result of this feckless representation, he felt that he had no plausible alternative but to plead guilty. This claim of error was never adjudicated in the district court: although Flete-Garcia made disparaging remarks about his trial counsel's shortcomings and suggested that ineffective assistance contributed to his decision to plead guilty, he never advanced a Sixth Amendment claim. Consequently, no effort was made to develop a record suitable for the adjudication of such a claim.

"We have held with a regularity bordering on the monotonous that fact-specific claims of ineffective assistance of counsel cannot make their debut on direct review of criminal convictions, but, rather, must originally be presented to, and acted upon by, the trial court." United States v. Mala , 7 F.3d 1058 , 1063 (1st Cir. 1993). "In adopting this prudential praxis, we have reasoned that 'such claims typically require the resolution of factual issues that cannot efficaciously be addressed in the first instance by an appellate tribunal.' " United States v. Santana-Dones , 920 F.3d 70 , 82 (1st Cir. 2019) (quoting Mala , 7 F.3d at 1063 ). In particular, "questions about whether counsel's challenged decisions were mistakes of a constitutional magnitude or simply reasonable strategic choices that did not pan out" normally can be answered only with the benefit of a developed record. United States v. Mercedes-De La Cruz , 787 F.3d 61 , 66 (1st Cir. 2015).

To be sure, the Mala rule is not ironclad. On rare occasions, we have considered the merits of ineffective assistance of counsel claims raised for the first time on direct review. See , e.g. , United States v. Natanel , 938 F.2d 302 , 309 (1st Cir. 1991). We have limited this exception, though, to cases in which "the critical facts are not in dispute and the record is sufficiently developed to allow reasoned consideration of the [ineffective assistance] claim." Mala , 7 F.3d at 1063 .

This case falls squarely within the general rule, not within the long-odds exception to it. The record is tenebrous as to critical issues, such as the content of the discovery (if any) Flete-Garcia received from trial counsel. Nor does the record indicate either why trial counsel elected not to cross-examine Adams or why he eschewed particular lines of inquiry when cross-examining other government witnesses. There are both tactical and strategic considerations that may have informed trial counsel's decisions in this case, but the record sheds no light on trial counsel's thinking. Consequently, the record does not tell us whether trial counsel's decisions, when made, were calculated stratagems or amateurish blunders. Last - but far from least - the record affords us no insight into what discussions took place between Flete-Garcia and his trial counsel regarding his decision to change his plea. Lacking a developed record that illuminates these critical areas, we have no principled choice but to conclude that the ineffective assistance of counsel claim is prematurely raised. See United States v. Miller , 911 F.3d 638 , 646 (1st Cir. 2018) ; United States v. Negrón-Narváez , 403 F.3d 33 , 41 (1st Cir. 2005). Accordingly, we dismiss the ineffective assistance of counsel claim; without prejudice, however, to Flete-Garcia's right, if he so elects, to raise it through a petition for post-conviction relief under 28 U.S.C. § 2255 .

VI. CONCLUSION

We need go no further. 8 The record before us withstands Flete-Garcia's myriad claims of error. For aught that appears, he was fairly tried, lawfully convicted, and appropriately sentenced in proceedings patiently conducted by an able trial judge. Consequently, we affirm the judgment of the district court; without prejudice, however, to Flete-Garcia's right to raise his ineffective assistance of counsel claim in a collateral proceeding pursuant to 28 U.S.C. § 2255 .

So ordered .