Defendants-appellants Kirk Lacey and Omar Henry were convicted by jury in the United States District Court for the Southern District of New York (Kimba M. Wood, J.) on charges stemming from their involvement in a fraudulent mortgage scheme. They appeal their sentences and restitution orders but not their convictions. We hold that U.S. Sentencing Guidelines § 2B 1.1(b)(2)(A)(ii), which increases an offense by two levels if it was “committed through mass-marketing,” applies only if the audience of the mass-marketing was in some sense victimized by the scheme; because the record is unclear in this case, we remand for the district court to make additional findings. We find no error, however, in the district court’s calculation of loss amount for sentencing. Finally, we agree with the parties that the district court’s restitution calculation was erroneous. Therefore, we VACATE the sentences and restitution orders and REMAND the cases for further proceedings consistent with this opinion.
BACKGROUND
I. Facts
Because the defendants were convicted after trial, we recite the facts taking the evidence in the light most favorable to the verdict. See, e.g., United States v. Hsu, 669 F.3d 112, 114 (2d Cir.2012).
Lacey and Henry participated in a fraudulent mortgage scheme operated by MTC Real Estate, Inc. (“MTC”). The chief executive officer of MTC was co-defendant Lavette Bills, a licensed real estate broker. The contours of the scheme were simple. In the typical case, MTC purchased a property at a favorable price in a “short sale” from a financially distressed homeowner by negotiating with the homeowner’s mortgage lender. The defendants then typically resold (“flipped”) the property at a higher price to a “straw buyer” who had no intention of actually living at the property or making all of the loan payments.1
*713MTC engaged in extensive radio advertisements featuring Lavette Bills. The advertisements produced potential straw buyers, or “radio leads,” and MTC employees followed up with the leads to find buyers. The advertisements told buyers they could receive up to $50,000 for buying a house with MTC, and some straw buyers actually did receive such payments. The ads also helped MTC find financially distressed homeowners willing to sell their homes in short sales.
MTC employees worked with straw buyers to submit false mortgage applications and documentation in order to make it more likely that loans would be approved. If the mortgage was approved and the sale went forward, MTC sometimes made a few payments on the straw buyer’s behalf so that the loan did not immediately go into default, to avoid setting off a “red flag” with the lending bank. When the straw buyer did ultimately default, the lending bank typically obtained title to the property by foreclosing on its mortgage.
After a three-week jury trial, both Lacey and Henry were found guilty on December 21, 2010. Lacey was convicted of conspiracy to commit bank and wire fraud, in violation of 18 U.S.C. §§ 1343,1344, and 1349 and of substantive bank and wire fraud in violation of 18 U.S.C. §§ 1343, 1344, and 2; Henry was convicted only of conspiracy.
II. Sentencing
At sentencing, the government sought a two-level increase in both defendants’ Guidelines range under U.S.S.G. § 2Bl.l(b)(2)(A)(ii) because, it argued, the scheme “was committed through mass-marketing” within the meaning of that provision. Defendants argued that the enhancement should not apply because the radio advertisements were directed at potential property sellers and straw buyers, not at the banks who were the victims of the fraud. The district court agreed with the government, noting that “the MTC marketing campaign was critical to the success of the fraud” because the marketing was “how MTC found people with distressed properties that could be exploited.” The district court therefore held that although the mass-marketing was not directed at the victims of the fraud (that is, the banks that made the mortgage loans), the mass-marketing was still “relevant conduct” to the offense and so the enhancement should apply.
The parties also disputed the amount of loss caused by the scheme. Under U.S.S.G. § 2Bl.l(b)(l), the base offense level for various crimes resulting in financial loss is enhanced based on the amount of loss. The government argued that the loss in this case should be calculated as the total of the differences between what MTC paid for each property at short sale and the value of the mortgage loan ultimately made on each property. This calculation resulted in losses of $731,077 attributable to Henry and $536,077 attributable to Lacey; both figures result in a 14-level enhancement under § 2Bl.l(b)(l). Defendants argued principally that because the banks had received title to the properties after default, the court should use the appraised value of the property that formed the basis for the fraudulent mortgage, rather than the short-sale price. Defendants also proposed that since the loss amount was difficult to ascertain, the court should instead base its sentence on the amount of financial gain to defendants, a mechanism described in Application Note 3(B) to U.S.S.G. § 2B1.1. Defendants asserted that the evidence at trial had shown that Lacey gained approximately $15,000 through the scheme, which would *714produce a 4-level enhancement.2 The district court, however, accepted the government’s loss calculation and the corresponding 14-level enhancement for each defendant.
The court sentenced Henry principally to a term of imprisonment of one year and one day and Lacey to a term of 46 months. The district court also ordered that Henry and Lacey, along with their codefendants Bills and Peter Chevere, be jointly and severally liable for restitution in the amount of $411,161.52 to OneWest Bank, one of the victims of the scheme. Defendants jointly moved to vacate the restitution orders because they did not account for the value of the collateral OneWest received.3 This appeal followed.
DISCUSSION
We review a district court’s sentencing decision for procedural and substantive reasonableness. See United States v. Cavera, 550 F.3d 180, 189-90 (2d Cir.2008) (en banc). “A district court commits procedural error where it makes a mistake in its Guidelines calculation, does not consider the § 3553(a) factors, or rests its sentence on a clearly erroneous finding of fact.” Hsu, 669 F.3d at 120 (internal quotation marks and alterations omitted); see also Gall v. United States, 552 U.S. 38, 51, 128 S.Ct. 586, 169 L.Ed.2d 445 (2007). We review the district court’s legal determinations de novo. Hsu, 669 F.3d at 120.
Defendants make three arguments, which largely repeat those made to the district court at sentencing: that the two-level enhancement for mass-marketing was erroneously imposed; that the 14-level loss enhancement was based on an incorrect calculation of loss; and that the amount of restitution was incorrectly calculated.
I. Mass-marketing enhancement
Defendants argue first that the district court erred by applying a two-level enhancement to their sentences for an offense “committed through mass-marketing.” See U.S.S.G. § 2Bl.l(b)(2)(A)(ii). After a careful reading of the Guidelines and other relevant authority, we hold that the mass-marketing enhancement is properly applied only when the targets of the mass-marketing are also in some way victims of the scheme. Because it is not clear on the current record whether the straw buyers who were the targets of the mass-marketing in this case were in some sense victims, we will remand to the district court for further factfinding.
To interpret the Guideline, we look first to its text. Section 2Bl.l(b)(2) reads:
If the offense—
(A) (i) involved 10 or more victims; or (ii) was committed through mass-marketing, increase by 2 levels;
(B) involved 50 or more victims, increase by 4 levels; or
(C) involved 250 or more victims, increase by 6 levels.
The Guideline applies to an offense “committed through mass-marketing.” As at least one other Circuit has recognized, an offense is “committed through mass-marketing” when mass-marketing is used to recruit or deceive victims of the offense, not when mass-marketing targeted at audiences other than victims is used in connection with the fraud in some other, more tangential manner. See United States v. Miller, 588 F.3d 560, 568 (8th Cir.2009). It is not enough that a scheme may be advanced by the use of mass marketing *715techniques; a scheme is committed through mass-marketing only when the mass marketing is directed toward individuals who will be harmed by the scheme.
This reading is bolstered by the surrounding text. Cf. Rowland v. Cal. Men’s Colony, Unit II Men’s Advisory Council, 506 U.S. 194, 199-200, 113 S.Ct. 716, 121 L.Ed.2d 656 (1993) (noting that the Dictionary Act, 1 U.S.C. § 1, directs a court to look to “context,” which includes “the text of the Act of Congress surrounding the word at issue, or the texts of other related congressional Acts”). All the other subsections of § 2Bl.l(b)(2) base enhancements on the number of victims. Indeed, the mass-marketing enhancement is posed as an alternative to the smallest number of victims in an escalating series of adjustments based on rising numbers of victims. The pattern thus strongly suggests that the enhancement scheme is designed to measure the scope of the wrong by the number of victims, and that the use of mass-marketing is relevant even when the number of actual victims is small, because fraudulent mass-marketing creates a large number of potential victims. Given this context, it is logical to interpret § 2Bl.l(b)(2)(A)(ii) as applying only when the mass-marketing is directed at individuals who may be victimized by the scheme.
Finally, the Guidelines’ definition of “victim” supports our reading. As relevant for this case, a “victim” is defined as “any person who sustained any part of the actual loss determined under subsection (b)(1),” U.S.S.G. § 2B1.1 app. note 1, while “actual loss” is in turn defined as “the reasonably foreseeable pecuniary harm that resulted from the offense,” id. § 2B1.1 app. note 3(A)(i). If a mortgage fraud scheme predictably results in pecuniary harm to unwitting, deceived straw buyers, the straw buyers have sustained “actual loss” and are therefore “victims” within the meaning of the Guidelines. They are therefore properly considered under the mass-marketing enhancement.
Returning to the facts of the instant case, it is not clear on the present record whether at least some of the consumers who were the targets of mass marketing were in some sense victimized, notwithstanding that the main thrust of the fraud was directed at banks. To the extent that any straw buyer was in on the scheme or received the promised $50,000 payment, such a buyer could not be seen as a victim. But some straw buyers testified that their credit scores were ruined. Others testified that they intended in good faith to purchase the property and pay the mortgage, and that Bills misled them into believing that they would be able to pay the mortgages on the properties through rental income. One straw buyer had to retain an attorney to deal with the legal consequences of foreclosure. Thus, there is evidence that at least some straw buyers were harmed by the scheme.
Our Court has not previously interpreted the mass-marketing enhancement, although two other Circuits, the Fifth and Eighth, have. Our reading of the rule is consistent with that of the Eighth Circuit. In United States v. Miller, a case very similar to this one, the Eighth Circuit upheld the district court’s rejection of an enhancement under § 2B1.1(b)(2)(A)(ii) for a defendant convicted of wire fraud. 588 F.3d at 568. Although the Miller defendant “engage[d] in mass-marketing to consumers via television commercials,” the court noted that his offense “involv[ed] fraud on financial institutions, not consumers,” and so the targets of the mass marketing (the consumers) were not the victims of the fraud (the banks). Id. The Eighth Circuit’s holding is consistent with our own reading of the Guideline, although we note that the court in Miller apparently *716was not faced with evidence that some or all of the consumers were also injured or defrauded. Id. at 567-68.
By contrast, in a pair of medical device fraud cases, United States v. Mauskar, 557 F.3d 219, 232-33 (5th Cir.2009) and United States v. Isiwele, 635 F.3d 196, 203-05 (5th Cir.2011), the Fifth Circuit upheld application of the mass-marketing enhancement. In those cases, mass-marketing techniques were directed at Medicare and Medicaid recipients to induce them either to visit a doctor to obtain a prescription for motorized wheelchairs that they did not need, Mauskar, 557 F.3d at 224, 233, or to provide a fraudster with their billing information, which was used to file fraudulent claims for such wheelchairs that were neither needed nor provided, Isiwele, 635 F.3d at 198. In these cases, although those persons did not sustain any actual financial loss under the Guidelines definition, see U.S.S.G. § 2B1.1 app. note 3(A)(i), and were thus not “victims” as defined by the Guidelines, see U.S.S.G. § 2B1.1 app. note 1, they were deceived into ordering unneeded and in some cases unprovided goods or services. They avoided financial loss only because the government as their insurer ultimately bore the cost of the deception. We need not decide whether the enhancement would properly apply in situations more closely analogous to the Fifth Circuit cases. We note, however, that a plausible argument can be made that the deceived patients were victimized by the scheme.
It is also not clear on this record whether the defendants’ activities are properly deemed “mass-marketing” under the relevant Guideline. The Sentencing Commission has defined “mass-marketing,” and this Court must give the Commission’s interpretation of its own Guideline “ ‘controlling weight unless it is plainly erroneous or inconsistent with the regulation’ ” or violates the Constitution or a federal statute. Stinson v. United States, 508 U.S. 36, 45, 113 S.Ct. 1913, 123 L.Ed.2d 598 (1993), quoting Bowles v. Seminole Rock & Sand Co., 325 U.S. 410, 414, 65 S.Ct. 1215, 89 L.Ed. 1700 (1945). “Mass-marketing” is defined by Application Note 4(A) to § 2B1.1:
For purposes of subsection (b)(2), “mass-marketing” means a plan, program, promotion, or campaign that is conducted through solicitation by telephone, mail, the Internet, or other means to induce a large number of persons to (i) purchase goods or services; (ii) participate in a contest or sweepstakes; or (iii) invest for financial profit. “Mass-marketing” includes, for example, a telemarketing campaign that solicits a large number of individuals to purchase fraudulent life insurance policies.
U.S.S.G. § 2B1.1 app. note 4. The three categories enumerated in the definition all describe common frauds in which fraudsters use mass media to attract victims to buy into fraudulent schemes that will separate the victims from their money. The categories, moreover, are apparently intended to be jointly exhaustive — the application note defines what mass-marketing “means,” not merely what it “includes.”
Defendants’ behavior clearly does not fall into either of the first two categories. First, neither the straw buyers nor the initial homeowners were solicited to “purchase goods or services,” since real estate is not “goods.” See Black’s Law Dictionary (9th ed.2009) (defining “goods” as “[tjangible or movable personal property other than money,” and citing U.C.C. § 2-105(1)). Second, this case does not involve a contest or sweepstakes. It is not clear, however, whether the radio and television ads fall into the third category, because the record does not clearly establish whether the straw buyers were invited to *717“invest for financial profit.” As defendants now describe the scheme, the straw buyers were never genuinely solicited to invest in property; rather, they were recruited to lend their names to a paper transaction in which they simply purchased and flipped the property in a way that purportedly would earn them a risk-free fee. Moreover, defendants’ scheme does not fit the typical mold in which advertisements or sales calls use false representations induce victims to invest money in high-risk or nonexistent ventures, leading to the loss of their investments.
Nevertheless, the record contains evidence that could be found to meet the Sentencing Commission’s definition. At least some advertisements and follow-up calls advised the targeted consumers that they could obtain a financial profit. Moreover, at least some straw buyers put their own credit at risk, which might be deemed an investment. Indeed, at trial, one witness testified that after hearing the various radio and television advertisements, a number of potential buyers called MTC and specifically expressed an interest in buying houses for investment purposes.
On remand, therefore, the district court should consider two questions: first, whether the defendants engaged in “mass-marketing” within the meaning of the relevant Guideline, as interpreted by the commentary; and second, if the defendants did engage in “mass-marketing,” whether the consumers who were the target of that mass-marketing were also in some sense victims of the overall criminal scheme, i.e., whether they were injured by the scheme.4
We add one final observation. The application of the mass-marketing enhancement presents significant issues of interpretation about which reasonable people can disagree, as evidenced by Judge Straub’s thoughtful dissent. The Sentencing Commission can easily clarify whether it intends the enhancement to apply whenever techniques of mass solicitation are employed in some way in connection with a scheme, or whether it is intended to apply in a narrower category of eonsumer/investmentylottery frauds in which a fraudulent scheme is marketed to large numbers of potential victims. We urge the Commission to do so.
II. Loss amount
Defendants also argue that the district court erroneously calculated the amount of loss attributable to the fraud, resulting in an inaccurately high Guideline recommendation. In reviewing the district court’s loss calculation, we “must determine whether the trial court’s method of calculating the amount of loss was legally acceptable.” United States v. Rutkoske, 506 F.3d 170, 178 (2d Cir.2007) (internal quotation marks and brackets omitted). We review legal conclusions, such as interpretations of the Guidelines, de novo and findings of fact for clear error. United States v. Turk, 626 F.3d 743, 747 (2d Cir.2010).
Defendants raise four arguments against the district court’s loss amount calculation. We find none of them persuasive.
Defendants first argue that the district court erred by failing to calculate both the intended and actual loss from the scheme. This argument misapprehends the district court’s duty at sentencing. Guidelines § 2Bl.l(b)(l) provides for stepped, cumu*718lative two-level enhancements based on the amount of loss attributable to a scheme. Application Note 3(A) to that section (which, under Stinson, is binding unless an unreasonable interpretation of the Guideline or contrary to law, see 508 U.S. at 45, 113 S.Ct. 1913) defines “loss” as “the greater of actual loss or intended loss.” U.S.S.G. § 2B1.1 app. note 3(A)(i). “ ‘Actual loss’ means the reasonably foreseeable pecuniary harm that resulted from the offense,” id., while “‘[i]ntended loss’ (I) means the pecuniary harm that was intended to result from the offense; and (II) includes intended pecuniary harm that would have been impossible or unlikely to occur (e.g., as in a government sting operation, or an insurance fraud in which the claim exceeded the insured value),” id. app. note 3(A)(ii).
Given that the court must apply the greater of the actual or intended loss amount, defendants have not explained how the district court’s alleged failure to calculate actual loss could have prejudiced them. Either the actual loss would have been less than the intended loss, and therefore irrelevant, or the actual loss would have been greater than the intended loss, in which case the court’s failure redounded to defendants’ benefit. Without deciding whether the court’s procedure might in some abstract sense constitute “error,” we note that procedural error in sentencing is subject to harmless error analysis, see United States v. Jass, 569 F.3d 47, 68-69 (2d Cir.2009), and any such error would certainly be harmless here.
Second, and relatedly, defendants argue that at least in a case involving fraudulent loans secured by collateral, the sentencing court must use the actual loss rather than the intended loss. This contention is directly contrary to the Guidelines commentary, which, as set forth above, defines loss as the greater of the actual or intended loss. See U.S.S.G. § 2B1.1 app. note 3(A)(i). Nothing in the text or application notes gives the slightest indication that a special exception applies in mortgage fraud cases. Defendants argue, however, that our decision in Turk and the Fifth Circuit’s decision in United States v. Goss, 549 F.3d 1013, 1016-19 (5th Cir.2008), establish such an exception. Defendants misread those cases.
In Turk, the defendant falsely told individual investors that she would record mortgages against property she owned in order to secure large loans, but in fact she did not record the mortgages, leaving the investors’ loans unsecured. 626 F.3d at 745. She also took out bank loans that were secured by recorded mortgages in the same property. Id. When the properties were later sold in bankruptcy, the unsecured investors lost virtually all of their money. Id. at 745-46. As a result, the investors’ actual losses were nearly the full value of the loans, and the key question on appeal was whether the losses to the investors were foreseeable. Id. at 748-51. The value of the property was irrelevant: either the investors’ unrecorded mortgages constituted no interest in the property at all, or that interest was essentially worthless. Id. at 748-49. Thus, it was clear that the actual loss was the full loan value, which was necessarily at least as great as the intended loss. The court therefore did not have occasion to consider the intended loss. Id. at 748 n. 3. Turk therefore does not establish defendants’ proposed rule.
While the Goss court appears at one point to equate deduction of the value of collateral with an “actual loss” calculation,5 *719the case holds simply that a sentencing court must deduct the value of real-property collateral from loss and cannot ignore that value, as the sentencing court had done in that case. See 549 F.3d at 1015— 17. Here, of course, the sentencing court did reduce the loss amount by what it found to be the value of the collateral: The court deducted the price at which defendants had purchased each property in the short sale transactions. It is clear, then, that Goss does not alter the general rule that the greater of actual or intended loss is the appropriate measure of loss for purposes of the Guidelines, and equally clear that the district court’s calculation here was fully consistent with Goss.
Third, defendants argue that the district court’s calculation of intended loss was erroneous because it failed to account for defendants’ subjective expectations and intent. Defendants rely principally on United States v. Confredo, in which this Court held that a defendant must be permitted “to persuade the sentencing judge that the loss he intended was less than the face amount of the loans.” 528 F.3d 143, 152 (2d Cir.2008). Initially, we note that Conpedo dealt with fraud in unsecured loans, id. at 151, so the district court in that case had no occasion to deduct the value of retained security from the amount of loss. At any rate, the district court in this case did not prevent Henry or Lacey from introducing evidence that he subjectively intended a lesser quantum of loss. To the extent that defendants argued that they intended or expected a lesser loss, however, the district court was entitled to find them not credible. While defendants are entitled to present evidence of their intentions, Conpedo in no way limits the role of objective evidence of intended loss. As the First Circuit noted in United States v. McCoy, which approved a similar loss calculation on similar facts, the term “intended loss” may fairly be read to encompass a defendant’s reasonable expectation of loss. 508 F.3d 74, 79 (1st Cir.2007). The difference between the short-sale price and the mortgage amount constitutes objective evidence of the amount that a reasonable defendant might expect a bank would lose in the transaction.
Fourth, defendants contend that the district court erroneously evaluated the collateral. Instead of relying on the short-sale price, defendants argue that the court should have valued each property according to appraisals submitted to the lender banks when the straw buyers purchased the properties and obtained mortgages. Because the short-sale prices were affected by the sellers’ distressed circumstances, defendants argue, those prices do not represent the true market value of the properties, and so should not be used in calculating the loss intended in each transaction.
In support of this argument, defendants point to Application Note 3(E)(ii) to U.S.S.G. § 2B1.1, which states that “[i]n a case involving collateral pledged or otherwise provided by the defendant,”
[l]oss shall be reduced by ... the amount the victim has recovered at the time of sentencing from disposition of the collateral, or if the collateral has not been disposed of by that time, the fair market value of the collateral at the time of sentencing.
Defendants argue that the short-sale price is not “the fair market value of the collateral at the time of sentencing,” since it represents a fire-sale price.
As with any finding of fact, this Court reviews the district court’s loss determination for clear error, United States v. Uddin, 551 F.3d 176, 180 (2d Cir.2009), and we find none here. The sentencing court is only required to make a “reasonable estimate of the loss.” U.S.S.G. § 2B1.1 *720app. note 3(C); see also United States v. Coppola, 671 F.3d 220, 249-50 (2d Cir.2012). Furthermore, because the sentencing court “is in a unique position to assess the evidence and estimate the loss based upon that evidence,” the sentencing court’s “loss determination is entitled to appropriate deference.” U.S.S.G. § 2B1.1 app. note 3(C). That observation is certainly applicable where, as here, the sentencing judge also presided over a weeks-long trial and heard a great deal of live testimony.
As an initial matter, we agree with several of our sister Circuits that although Application Note 3(E)(ii) “accurately describes the calculation of actual loss,” the note “cannot be mechanically followed where intended loss is higher,” since the larger intended amount is a better “measure for the defendant’s culpability” than is the actual loss. McCoy, 508 F.3d at 79, citing United States v. McCormac, 309 F.3d 623, 628-29 (9th Cir.2002) and United States v. Williams, 292 F.3d 681, 686 (10th Cir.2002); see also United States v. Innarelli, 524 F.3d 286, 290-91 (1st Cir.2008). Thus, a sentencing court need not apply the fair market value as an offset in calculations of intended loss; it need only offset the loss amount by however much it finds the defendant did not intend loss. In the case of a loan secured by an interest in real property, the sentencing court may — given appropriate evidence — draw the inference that the intended loss should include an offset for the value of the property. But that is because it would be unlikely for even a nefarious defendant to intend the improbable result that real property be destroyed or otherwise rendered valueless.
Here, the district court’s method was a reasonable estimate of the intended loss. The district court was • entitled to find that the short-sale prices, rather than the appraisals made for the mortgages, were an appropriate offset for at least two reasons: First, the short-sale prices were negotiated, not fraudulent; second, evidence showed that the appraisals at the time of the fraudulent mortgages may not have been reliable.
First, at sentencing, the government argued that the trial evidence supported the view that the short-sale prices represented a negotiated, arm’s-length price. For example, Mosheh Flowers, an MTC employee, testified that the short-sale prices were themselves based on third-party appraisals of the properties. He also testified that the sellers sometimes rejected MTC’s initial offers, requiring MTC to increase the price. While a fact-finder would be entitled to take into account the distressed circumstances of “underwater” property owners in deciding whether a short-sale price accurately reflects the fair market value of the property, no rule of law disqualifies such a sale as evidence of the fair market value. It is hardly clear error for a sentencing judge to conclude that a price negotiated by a willing buyer and a willing seller is better evidence of the property’s value than an appraisal by a purported expert.
Second, there was evidence that the appraisals that MTC submitted to obtain the mortgages may have been unreliable. For example, Flowers testified that MTC employees sometimes paid off appraisers in order to persuade them to raise the assessed value of the properties, increasing the mortgage value and thus MTC’s fraudulent proceeds. In the case of one of the properties, 1236 Tinton Avenue, the bank found MTC’s appraisal was inaccurate and required a new assessment which revised the value downward. The appraisals also left out the short-sale price that MTC had just paid to obtain the property, even though the appraisal forms required the appraiser to include all transactions on the *721property within the last 12 months. The district court was entitled to assess the credibility of the appraisals, and to make its own determination of the fair market value of the properties.
We need not decide whether the district court’s method would be appropriate in every similar fraud case. Here, based on the record before the court, we find no error in its loss calculation.
III. Restitution
The government concedes that the district court erred by failing to credit any of the value of the collateral in formulating its restitution orders. We agree, and will remand the case for a recalculation of the restitution amount. We note that unlike the loss calculation for the purposes of sentencing, which may incorporate a merely intended loss in order to punish a culpable defendant, restitution is designed to make the victim whole, see Innarelli, 524 F.3d at 293-94, and must therefore be based only on the actual loss caused by the scheme. See 18 U.S.C. § 3663A(b)(l); see also United States v. Marino, 654 F.3d 310, 319-20 (2d Cir.2011) (“[Restitution is authorized ... only for the victim’s actual loss.” (internal citation and quotation marks omitted)). To determine restitution, the district court will therefore have to make a new loss calculation based only on the banks’ actual losses.
CONCLUSION
For the foregoing reasons, the sentences and restitution orders are VACATED. The cases are REMANDED to the district court for further proceedings consistent with this opinion.
. An MTC employee testified at trial that MTC permitted sale to a legitimate buyer if one could be found, but he also testified that he could not recall a case in which a legitimate buyer had actually been used.
. Henry did not (and does not now) specify how much he gained from the scheme.
. It does not appear from the docket that the district court ever ruled on this motion.
. We note that defendants have not challenged the substantive reasonableness of their sentences. On remand, whether nor not the mass-marketing enhancement technically applies, the district court remains free to impose whatever sentence it finds is the lowest sentence necessary to accomplish the purposes of sentencing, after considering the factors set forth in 18 U.S.C. § 3553(a), including the correct Guidelines range.
. See Goss, 549 F.3d at 1018 ("[W]hether to deduct collateral' — whether to employ an actual or an intended-loss calculation — will depend upon the specific facts at hand.”).