I. INTRODUCTION
Defendant Blake Snowden pleaded guilty to unlawfully obtaining information from a protected computer, see 18 U.S.C. § 1030(a)(2)(C), and to intercepting emails, see id. § 2511(l)(a). In calculating his offense level under the sentencing guidelines, the district court applied a 16-level enhancement after finding that these crimes caused losses exceeding a million dollars. The resulting guideline sentencing range was 41-51 months’ imprisonment. Varying downward, the court imposed a 30-month sentence.
Defendant argues that the district court erred in calculating loss and thereby arrived at an unduly high guideline sentencing range. We discuss the court’s calculation but ultimately hold that if there was any error, it was harmless because it did not affect the sentence. The court, after thoroughly exploring the matter, explained why the sentence was proper under 18 U.S.C. § 3553 and unequivocally stated that it would impose the same sentence even if 30 months exceeded the correct guideline range. On one ground, however, we must reverse. The district court ordered restitution in the amount of $25,354. Both parties agree that this was an oversight that should be corrected on remand to $24,174. •
II. BACKGROUND
Onyx, M.D., Inc. is a physician-staffing service that specializes in locum tenens recruiting, placing physicians in hospitals and clinics for short terms. To respond quickly to requests from hospitals, Onyx relies heavily on data it maintains on individual physicians, such as one’s area of practice and willingness to work in a particular location. Onyx’s sales team develops relationships with physicians to gather this type of information.
Onyx enters all such information into a software platform called Bullhorn. “Bullhorn is a single repository for all of Onyx’s sales-related communications and information, and it contains a cumulative record of all sales and customer relations data from the beginning of Onyx in 2005 to the present.” R., Vol. 6 at 12. Sales associates have “pretty much lived and breathed *1032within [Bullhorn].” Id., Vol. 5 at 45. Onyx sees its Bullhorn database as a competitive advantage. It has never sought to market its data to others.
Defendant worked as an Onyx sales associate until it terminated his employment in August 2010. He then decided to compete with Onyx in physician placement, in large part by stealing Onyx’s proprietary data. In March 2011 an Onyx employee gave Defendant an executive’s password, which Defendant used to set up a Bullhorn account. Over the next few months Defendant logged in dozens of times, copying gigabytes of data. He was also able to intercept the emails of four Onyx executives. Every time they sent or received an email, a copy would be sent to Defendant. Almost 20,000 messages were forwarded. Defendant’s repeated and unfettered access to Bullhorn went undetected for six months. Onyx eventually noticed the hack, which the FBI traced to a computer at Defendant’s address.
Defendant told an Onyx competitor that he was “going to go after” Onyx, including by targeting its largest client. Id. at 38 (internal quotation marks omitted). But his actions did not produce their intended results. They were completely ineffectual, neither benefiting him nor harming Onyx’s business. At Defendant’s sentencing hearing Onyx’s vice president of software development acknowledged that he could not “point to any single doctor of the entire universe of doctors in Bullhorn ... that [Defendant] placed in a position after he left [Onyx].” Id., Vol. 5 at 84-85. The only impact on Onyx’s bottom line was its loss of about $25,000 in legal fees and employee time incurred in response-to the intrusion.
Under the applicable sentencing guideline, Defendant’s base offense level was 6, see USSG § 2Bl.l(a)(2) (theft offenses), subject to increase based on the amount of loss suffered by Onyx, see id. § 2Bl.l(b)(l). The district court found Onyx’s actual loss to include the $25,000 in response costs and more than $1.5 million in costs to develop the Bullhorn database. These findings increased Defendant’s offense level by 16, leading to a guideline sentencing range of 41-51 months’ imprisonment. Had the court limited its loss finding to response costs — as Defendant advocates — it would have applied only a 4-level increase, leading to a guideline range of 8-14 months.
From the 41-51 month guideline range, the court varied downward to sentence Defendant to 30 months. The court also stated that in case it had incorrectly calculated the guideline range, it would in the alternative impose the same 30-month sentence as an upward variance.
III. DISCUSSION
A. Loss Calculation
Before imposing sentence a district court must calculate the defendant’s sentencing range in accordance with the sentencing guidelines. See United States v. Todd, 515 F.3d 1128, 1130 (10th Cir.2008). The calculation requires determining (1) the offense level for the criminal activity and (2) the defendant’s criminal-history category. The offense level under USSG § 2Bl.l(b)(l), the guideline applicable to Defendant’s crimes, increases with the amount of the victim’s loss. The word loss is not defined in the guideline itself, but the commentary to the guideline provides guidance. It states that “loss is the greater of actual loss or intended loss.” Id. § 2B1.1 cmt. n.3(A). Because the government has not pursued any theory of intended loss, we need consider only actual loss. The commentary defines actual loss as “the reasonably foreseeable pecuniary harm that resulted from the offense.” Id. cmt. n.3(A)(i). Pecuniary harm is “harm *1033that is monetary or that otherwise is readily measurable in money” and “does not include emotional distress, harm to reputation, or other non-economic harm.” Id. cmt. n.3(A)(iii).
The district court did not identify any pecuniary harm other than $25,000 in response costs. It acknowledged, and did not dispute, Defendant’s claim at sentencing that “his conduct ... did nothing to diminish the value of Onyx’s database and did not damage or disrupt Onyx’s business ..., and there is no evidence he placed Onyx’s candidates in a hospital position or placed another physician in any of Onyx’s client’s hospital openings.” R., Vol. 6 at 19-20. But the court included as “actual loss” the $1.5 million cost to Onyx of creating the Bullhorn database. It relied on the provision in the commentary to § 2B1.1 stating that the court “need only make a reasonable estimate of the loss,” and may take into account, “[i]n the case of proprietary information (e.g., trade secrets), the cost of developing that information.” USSG § 2B1.1 cmt. n.3(C)(ii). Based on that loss calculation and other undisputed findings under the guidelines, the court set Defendant’s sentencing range at 41-51 months. If, as Defendant insists, the loss did not include the cost of creating the database, his sentencing range would have been 8?14 months.
We have some sympathy with the district court’s calculation of the offense level. Development costs seem to us to be a reasonable measure of the severity of offenses such as Defendant’s. He appropriated to his own use information that cost Onyx $1.5 million to develop. One’s sense of outrage at the nonconsensual copying of a proprietary database is likely to be proportional to the time, effort, and expense incurred in creating the database. Indeed, this measure of severity has been adopted in the statute under which Defendant was convicted. Whether a violation of 18 U.S.C. § 1030(a)(2)(C) is a misdemeanor or a felony can depend entirely on whether “the value of the information obtained exceeds $5,000,” 18 U.S.C. § 1030(c)(2)(B)(iii); and the cost of development is a reasonable (if rough) measure of value.
Nevertheless, the guideline requires proof of loss, and the guideline commentary does not depart from that approach. To be sure, the loss caused by a theft will typically be the value of the property stolen. Thus, the note relied on by the district court allows the “estimate of the loss ” to “tak[e] into account, as appropriate ..., the cost of developing” the stolen proprietary information. USSG § 2B1.1 cmt. n.3(C) (emphasis added). But we do not read that language in the commentary as contradicting the guideline by substituting cost of development for loss in calculating the offense level; and the evidence before the district court was that Onyx did not suffer any business loss from Defendant’s acts. Perhaps one could assess Onyx’s loss by treating Defendant’s theft as an involuntary sale of the data by Onyx and estimating the price that Onyx would reasonably demand if it had to share its database; but neither the district court nor the government has suggested this approach or what that reasonable price would be.
All this is to say that we would be reluctant to affirm Defendant’s sentence on the ground that his offense level (and, hence, his guideline sentencing range) was correctly calculated. “Loss” is the key, and the development cost was not adequately tied to any loss actually suffered by Onyx.
B. Harmless Error
Nevertheless, we affirm Defendant’s sentence because the district court *1034would have imposed the same sentence even if the guideline range had been much lower than what it calculated. The court, was not required to impose a sentence within the guideline range. See United States v. Booker, 543 U.S. 220, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005). It could decide to vary from that range and impose a harsher or more lenient sentence if persuaded that doing so better took into account the sentencing factors set forth in 18 U.S.C. § 3553(a).1 Indeed, the court did not impose a sentence within the guideline range it calculated. Based on its assessment of the § 3553 factors, it varied downward from the range by 11 months. And, more importantly, it made clear that those factors would cause it to impose the same sentence even if it had miscalculated the guideline range as higher than it should have been.
This is the rare case in which any error in the loss calculation was harmless, because “the error did not affect the district court’s selection of the sentence imposed.” United States v. Glover, 413 F.3d 1206, 1210 (10th Cir.2005) (internal quotation marks omitted). The district court devoted a great deal of time and reflection to Defendant’s sentence; and we have no reason to doubt its repeated assertion that it would impose the same sentence even if Defendant’s true guideline range was below what the court calculated.
Before the proceeding at which it imposed sentence the court had conducted a six-hour evidentiary hearing (in two sessions) and had received briefs and proposed findings and conclusions from the parties. The court said that it had devoted many hours to reviewing these submissions- and the evidence. After reciting its evidentiary findings at length, it declared that Defendant’s offense level was 21, but it added that if the proper offense level were found lower on appeal, it would vary upward. (The court also stated that it would depart upward but, as the government concedes, the court did not give proper notice that it might depart, so we consider only the variance.) In the court’s words:
And I want to say that in the event that on appeal the Tenth Circuit may find that I have incorrectly calculated the loss for purposes of United States Sentencing Guideline Section 2B1.1, I want to make sure that.it is clear that in the alternative, I find that the sentencing guidelines do not adequately account for the seriousness of the offense, to provide adequate deterrence, or to sufficiently protect the public and innocent victims from the infliction of a further harm at the defendant’s hands for the following reasons:
[The court discusses upward departure.] [A]pplying the 18 United States Code Section 3553(a) factors, in particular the seriousness of this offense, providing adequate deterrence, sufficiently protecting the public and innocent victims, the Court finds that an upward variance in this case would also be appropriate.
*1035Now, this is only if my calculation under 2B1.1 was not appropriate. I find that the guideline range provided by an offense level 21 is a more appropriate guideline range.
R., Vol. 6 at 32-33.
The court proceeded to set forth jn detail its justifications for its sentence under the § 3553 factors, including that Defendant’s “conduct was egregious”; “[h]e knew what he was doing was wrong”; he assaulted the personal privacy of the victims; he gained an unfair competitive advantage over Onyx; and he does not respect the law. Id. at 33-39. It stated its inclination to sentence Defendant to 41 months’ imprisonment, the bottom of the guideline range, and noted also that it would upwardly vary to 41 months if it had incorrectly calculated that range. See id. at 39-40. But after counsel pleaded on behalf of Defendant, the court agreed to vary downward to 30 months, even after expressing its general view that “[t]he guidelines are definitely much lower than they should be” with “white collar fraud criminals.” Id. at 44. Again it indicated that if the properly calculated guideline range was less than 30 months, it would vary upward to 30 months:
Before I get to the special conditions, I want to clarify that my variance [downward to 30 months] is based on the assumption that my calculation of the guideline range under 2B1.1 is correct, so I vary downward from that. If that is incorrect, based on the reasons I previously stated for an upward departure or variance, my inclination would be wherever that guideline range is, that it is 30 months that is a sufficient but not greater than necessary sentence based on the factors that I enumerated.
Id. at 62.
By varying downward from the guideline range it had calculated and stating that it would vary upward to the same sentence if the correct guideline range was below the 30-month sentence, the court made utterly clear that the specifics of Defendant’s offense overrode the intricacies of the guideline calculations.
Defendant argues that we should not recognize the validity of the district court’s alternative upward variance for two reasons. First, he contends that the court did not state that it would impose a 30-month sentence by varying upward, only that it would find such a variance “appropriate.” Aplt. Br. at 24. This is an artificial reading of the district court’s position. We have no doubt about what the court meant. Its language was as precise as one could expect in the circumstances. It suffices that the court clearly indicated that it would impose the same 30-month sentence either as a downward variance if the guideline range was 41-51 months or as an upward variance if the range was lower. It could hardly have been more forceful in conveying that it had decided what the proper sentence should be, whatever the guidelines might advise.
Second, citing to United States v. Pena-Hermosillo, 522 F.3d 1108, 1117-18 (10th Cir.2008), Defendant argues that because the district court failed to give proper notice that it might depart upward, its decision to vary was “tainted by legal error.” Aplt. Br. at 26. We see no taint. The decisions to vary and to depart were independent. Pena-Hermosillo does not help Defendant. In that case we refused to rely on the district court’s statement that it would impose the same sentence even if the guideline range was different from its calculation, but that was because the court “offer[ed] no more than a perfunctory explanation for its alternative holding.” Pena-Hermosillo, 522 F.3d at 1118. Such a variance, we said, required more explanation. See id. at 1117. The *1036court’s explanation here, however, was far from perfunctory.
We recognize the importance of the sentencing court’s properly calculating the guideline sentencing range. That range provides an anchor for the court in evaluating the effects of the § 3553(a) factors. It might have been better if the district court had explicitly stated that it would have varied upward to impose the same sentence even if it had been informed that the correct guideline range was 8-14 months. But in the context of the presentations at sentencing, that was implicit in the court’s statement that it would impose the same sentence even if the guideline range had been set too high. And that statement was not some boilerplate remark to avoid reversal. The court had devoted uncommon attention to Defendant’s sentence. And the court’s thoughtful independence from the guidelines was demonstrated by its setting a sentence below what it considered the correct guideline range even though its general view was that the guidelines were too lenient on white-collar criminals. In short, we are confident that if the district court erred in its guideline calculation (a matter we need not decide), that error did not affect Defendant’s sentence. To put it another way, a remand to correct the offense-level calculation would be a waste of everyone’s time and effort because Defendant would receive the same sentence.
IY. CONCLUSION
We REMAND for entry of restitution in the amount of $24,174 and otherwise AFFIRM Defendant’s sentence.