Opinion by Judge RYMER.

This appeal requires us to decide whether Application Note 4 to U.S.S.G. § 2B1.1, providing that “loss [in cases of fraud] includes any unauthorized charges made with stolen credit cards, but in no event less than $100 per card,” applies to unauthorized use of credit card numbers as well as the card itself. We hold that it does.

Kenneth Yellowe appeals his sentence after his conviction following a plea of guilty to conspiracy to possess unauthorized access devices and use unauthorized access devices to obtain goods and services in violation of 18 U.S.C. § 1029 (a)(2), (a)(3), and (b)(2). His conviction arises from a scheme to make unauthorized use of over 8,500 credit card numbers. The district court had jurisdiction pursuant to 18 U.S.C. § 3231 . This court has jurisdiction over Yellowe’s timely appeal pursuant to 28 U.S.C. § 1291 and 18 U.S.C. § 3742 . We affirm.

I

Yellowe owned a telemarketing company in Houston, Texas. In November, 1992, he spoke to Francisco Rivera about a scheme to use the credit card numbers of his old customers. Unbeknownst to Yellowe, Rivera taped that conversation, gave it to the Secret Service, and then cooperated with the Secret Service.

Yellowe’s plan was to have Rivera call in false sales orders to a “processor” — a middleman who accepts credit card charges from merchants and then processes them through banks for a fee — who would then pay Yel-lowe’s telemarketing firm. Yellowe claimed he had well over 50,000 approved credit card numbers, each of which had previously been used to make at least $500 purchases.

Rivera proposed an alternate method, claiming he had a merchant friend with a point of sale terminal and a merchant account at a bank which would allow them, for a percentage, to input the credit card numbers manually, process the charges, and deposit the sales money directly into the account themselves, without going through the processor. In reality, the terminal was supplied by the Secret Service and was not connected to a merchant account at a bank to avoid real losses.

Yellowe suggested that the merchant set up three or four different accounts at several different banks to avoid having too much activity in one account, and projected that they could make up to $300,000 in two weeks. Yellowe sent Rivera some 300 numbers to test out their plan. Rivera reported that 15 of 42 numbers (35.7%) were approved. Yel-lowe promised to send 5,000 more and suggested that Rivera get more terminals so they could process more numbers. He anticipated that Rivera could charge $200,000 in two weeks.

After Rivera told Yellowe there were four terminals working, and he had 41 approvals out of 325 numbers tried, Yellowe flew to Honolulu. At the store where the terminals were set up, Rivera introduced him to an undercover Secret Service agent. Yellowe said the scheme could continue for six weeks and that, if ten more machines were obtained, they could charge $1 million per week. Yellowe also provided the agent with a new list of 6,631 credit card numbers. He was then arrested.

At sentencing, the court heard testimony from Yellowe and a Secret Service agent. The agent testified that Yellowe could have continued the fictitious charges for four to five weeks before they would have been detected, and that 90% of the owners of the credit card numbers he contacted (36 of 40) still had the same numbers that Yellowe tried to use. Their credit limits ranged from $1,000 to $12,000, but the agent did not know how much available credit any of the customers had. Yellowe testified that he thought the loss would be much smaller. ,

The court found that the loss for purposes of determining Yellowe’s offense level was at least $700,000 but less than $800,000, such that his offense level was 16. It used Application Note 4 to U.S.S.G. § 2B1.1 to calculate the loss, finding that some 7,000 of the 8,545 cards were workable and multiplying that number by the presumed minimum loss of $100 per card. The court declined to reduce Yellowe’s offense level for partially completed conduct under § 2Xl.l(b)(2), and refused to depart downward. It then sentenced Yel-lowe within the guideline range to 24 months in prison.

II

We review the district court’s application of the Sentencing Guidelines de novo. United States v. Fagan, 996 F.2d 1009, 1017 (9th Cir.1993). The district court’s underlying factual findings are reviewed for clear error. United States v. Chapnick, 963 F.2d 224, 226 (9th Cir.1992). Its discretionary decision not to depart downward is not reviewable. United States v. Morales, 972 F.2d 1007, 1011 (9th Cir.1992), cert. denied, — U.S. -, 113 S.Ct. 1665 , 123 L.Edüd 283 (1993).

III

Yellowe argues that the district court misapplied the Guidelines by using the $100 minimum loss mandated when a credit card is used rather than determining the intended loss based on what Yellowe believed the scheme would produce. He submits that Application Note 4 applies only to use of credit cards, not numbers. We disagree.

The offense level for access device fraud offenses is determined under § 2F1.1; it varies depending on the amount of “loss.” Application Note 7, which indicates that “if an intended loss that the defendant was attempting to inflict can be determined, this figure will be used if it is greater than the actual loss,” U.S.S.G. § 2F1.1, comment, (n. 7), cross references the commentary to § 2B1.1, which applies to theft offenses. Note 4 to § 2B1.1 provides that:

U.S.S.G. § 2B1.1, comment, (n. 4). Other commentary to § 2F1.1 indicates that “loss need not be determined with precision. The court need only make a reasonable estimate of the loss, given the available information.” U.S.S.G. § 2F1.1, comment, (n. 8).

Nothing in § 2F1.1 or § 2B1.1 suggests that loss having to do with unautho rized charges made "with stolen credit card numbers should be treated differently from unauthorized charges made with the plastic itself. In both cases, the value of the unauthorized use exceeds the intrinsic value of the device; to determine loss based on the value of the card or the number alone would understate the severity of the offense. Loss is caused in precisely the same way, by unauthorized use of the credit card number. The distinction Yellowe urges is therefore one without a difference.

We have previously applied § 2B1.1, comment. (n. 4) to misuse of numbers, albeit without focusing on the argument Yellowe makes here. United States v. Burns, 894 F.2d 334 (9th Cir.1990) (per curiam). We now explicitly hold that loss under Application Note 4 to § 2B1.1 includes any unauthorized charges made with stolen credit card numbers, as well as cards.

This loss is a presumed loss, setting a floor beneath which neither “actual” nor “intended” loss may fall. As the district court did not err in applying Note 4 to unauthorized use of numbers, and there is no dispute about the number of stolen numbers, Yel-lowe’s subjective intent is immaterial.

Yellowe argues, however, that even if the court correctly calculated his offense level under Application Note 4, it should have gone with his estimate of the likely return (12.6% or 35.7%, depending on which conversation with Rivera is used), rather than the agent’s (90%). He contends that because determining “intended loss” is the goal of the guideline, the focus should be on his knowledge and intentions. As he says he believed that only 12.6% of the cards would be valid, the “loss” could have been no more than $107,000. This would produce only a six-level increase instead of ten.

We cannot say the district court clearly erred in relying on the agent’s testimony. Application Note 4 says nothing about probabilities; rather, it presumes $100 per unauthorized credit card number. For the district court to multiply less than all the stolen numbers by $100 therefore gave Yellowe the benefit of the doubt. Cf. United States v. Lorenzo, 995 F.2d 1448, 1460 (9th Cir.) (“amount of loss that the conspirators intended to inflict does not have to be realistic”), cert. denied, — U.S.-, 114 S.Ct. 225 , 126 L.Ed.2d 180 (1993).

IV

Yellowe contends that the terminals operation was thought up by law enforcement, and real terminals linked to real accounts with real credit and real banks were missing. He accordingly argues that he should have been entitled to a three-level reduction pursuant to § 2Xl.l(b)(2), which requires a decrease unless “the defendant or a co-conspirator completed all the acts the conspirators believed necessary on their part for the successful completion of the substantive offense or the circumstances demonstrate that the conspirators were about to complete all such acts but for apprehension or interruption by some similar event beyond their control.” Section 2Xl.l(b)(2) specifically applies to partially completed frauds. U.S.S.G. § 2F1.1, comment, (n. 9).

Yellowe, however, had the unauthorized access devices and was about to use them to make charges and get money. He had the equipment and personnel. Only two things stopped him: being arrested, and the fact that the terminals were not connected to a bank. The commentary says that “no reduction ... is warranted” in cases where “the substantive offense was substantially completed or was interrupted or prevented on the verge of completion by the intercession of law enforcement authorities.” U.S.S.G. § 2X1.1, comment, (backg’d.). The district court consequently did not err in declining to grant a three-level reduction.

V

Yellowe argues that the court failed to consider the mitigating factors he argued individually — the sting, that his conduct was aberrant, and his otherwise good record— cumulatively. See United States v. Cook, 938 F.2d 149, 153 (9th Cir.1991). The government argues that he never asked the court to consider these circumstances together, and has therefore waived the issue.

Even if Yellowe has not waived the point, the district court acknowledged that it had the discretion to depart downward but chose not to. This discretionary decision is not reviewable. Morales, 972 F.2d at 1011 .

AFFIRMED.

. Application Note 3 to § 2B 1.1 is to the same effect, providing that "loss need not be determined with precision, and may be inferred from any reasonably reliable information available, including the scope of the operation.” U.S.S.G. § 2B1.1, comment, (n. 3). . The legislative history also suggests that credit card number fraud should be treated similarly to credit card fraud. See S.Rep. No. 368, 98th Cong., 2d Sess. 2 (1984), reprinted in 1984 U.S.C.C.A.N. 3182, 3647, 3648 (noting growing problem in counterfeit credit cards and unauthorized use of account numbers); H.R.Rep. No. 894, 98th Cong., 2d Sess. 4 (1984), reprinted in 1984 U.S.C.C.A.N. 3689, 3689 (same).