SELYA, Circuit Judge.

In this appeal, defendant-appellant William L. "Billy" Bernier asserts that the sentencing court erred in making a drug quantity determination. After careful consideration, we reject this assertion and affirm the defendant's sentence.

We rehearse the background facts to the extent needed to provide context. A jury found that the defendant, at the times material hereto, was a member of a marijuana distribution conspiracy. Trial testimony indicated that the ringleader, Chad Marquis, acquired marijuana in Canada and transported it to the United States. He sold some of this marijuana to the defendant. These purchases began as early as 2002.

After a time, the defendant introduced Marquis to Michael Donato and Jeff Webber. Marquis subsequently sold marijuana directly to all three men, sometimes in pairs and sometimes separately.

In addition to these business relationships, Marquis and another coconspirator, Steve Nadeau, shared a marijuana storage unit. Nadeau occasionally delivered Marquis's marijuana to the defendant and others.

On July 10, 2009, a federal grand jury indicted the defendant for conspiracy to possess with intent to distribute marijuana. See 21 U.S.C. § 841(a)(1), 846. After the jury found the defendant guilty, the district court commissioned a presentence investigation report (PSI Report). This report, as amended in response to the defendant's objections, is front and center in this appeal.

At the disposition hearing, the district court, relying on trial testimony as well as facts limned in the PSI Report, attributed 26 kilograms of marijuana to the defendant, yielding a base offense level of 18. See USSG § 2D1.1. Combined with the defendant's criminal history category (I), this offense level produced a guideline sentencing range (GSR) of 27 to 33 months. The court imposed an incarcerative sentence at the bottom of the range, 27 months. This timely appeal ensued.

Before us, the defendant argues that the sentencing court erred in calculating the drug quantity attributable to him— an error that he alleges inflated his GSR and, thus, adversely influenced his sentence. Where, as here, a sentencing court's drug quantity determination is factbound, appellate review is for clear error. See United States v. Rodríguez-Lozada, 558 F.3d 29, 42 (1st Cir.2009). Under this deferential standard, we must honor the sentencing court's findings "unless, on the whole of the record, we form a strong, unyielding belief that a mistake has been made." Cumpiano v. Banco Santander P.R., 902 F.2d 148, 152 (1st Cir.1990).

Under the sentencing guidelines for federal drug crimes, sentence length is driven in part by drug quantity. For that purpose, drug quantity need not be proven beyond a reasonable doubt but, rather, need only be supported by a preponderance of the evidence. See United States v. Rodriguez, 525 F.3d 85, 107 (1st Cir.2008).

We have recognized that "[t]he calculation of drug quantities is not an exact science." United States v. Platte, 577 F.3d 387, 392 (1st Cir.2009). Consequently, "a sentencing court charged with that responsibility need not be precise to the point of pedantry." Id. A "reasoned estimate[] based on historical data" will suffice. Id.

In the case at hand, the defendant's claim of error rests on the assertions that the court gave too much weight to the inherently unreliable testimony of his coconspirators. This error, he says, was compounded by the court's misperception of that testimony. We find these arguments unpersuasive.

The defendant's most ferocious attack is aimed at the testimony of Marquis and Donato. In measuring the force of this attack, we acknowledge that accomplices sometimes have their own agendas, trying to shift blame, minimize culpability, or accommodate prosecutors in order to better their own lot. For this reason, accomplice testimony must be viewed with special caution. See, e.g., United States v. Hernández, 109 F.3d 13, 15-16 (1st Cir. 1997); United States v. Pelletier, 845 F.2d 1126, 1128-29 (1st Cir.1988). But judges, unlike uninstructed juries, are well aware of this danger, and it is a bedrock principle that a sentencing court may find the trial testimony of coconspirators sufficiently reliable to ground a drug quantity determination. See, e.g., Platte, 577 F.3d at 393; United States v. Pierre, 484 F.3d 75, 88 (1st Cir.2007). This is merely a subset of the time-tested tenet that "credibility determinations are part of the sentencing court's basic armamentarium." Platte, 577 F.3d at 392-93. Thus, we decline the defendant's sweeping invitation to find that all coconspirator testimony is untrustworthy.

More specifically, the defendant contends that Marquis's testimony lacked credibility because on two occasions prior to trial, he failed to mention that the defendant was one of his customers. But there is no per se rule to the effect that a witness, to be believed, must tell his story in exactly the same way each and every time. Circumstances vary, and an earlier omission may or may not undermine a later account. Here, as elsewhere, we will upset a sentencing court's "credibility determination only if we have a definite and firm conviction that a mistake has been committed." United States v. González-Vélez, 587 F.3d 494, 504 (1st Cir.2009) (citation and internal quotation marks omitted). In this instance, we have no such conviction, especially since the sentencing judge presided over the trial and was in an enviable position to gauge Marquis's credibility and to separate wheat from chaff.

If more were needed—and we do not think that it is—the defendant does not identify any specific contradiction or implausibility in Marquis's testimony. Nor do Marquis's earlier statements, viewed in context, cast any substantial doubt on his veracity. Even though he did not specifically identify the defendant as a customer on those occasions, he did identify him all along as a member of the drug-trafficking ring. The minor omission of the defendant's dual status does not require discarding the baby with the bath water. Cf. United States v. Webster, 54 F.3d 1, 5 (1st Cir.1995) (upholding trial court's decision to credit testimony of witness who was "an admitted perjurer, a drug user, and a turncoat who received a substantially reduced sentence for implicating others").

This brings us to the defendant's claim that the coconspirators' trial testimony, even if not inherently unreliable, did not fairly support the lower court's drug quantity determination. This claim focuses on the absence of definite numbers and emphasizes that, in the context of a conspiracy, a sentencing court "must determine the specific quantity of drugs for which the defendant is personally responsible." United States v. Rivera Calderón, 578 F.3d 78, 100 (1st Cir.2009) (citations and alteration omitted).

The defendant's criticism is misplaced. The court below recognized its obligation to make an individualized determination that the defendant was personally accountable for 26 kilograms out of the much larger store of marijuana handled by the conspiracy as a whole.

Relatedly, the defendant identifies several instances in which the testimony noted by the court was inexact. Given the wavering contours of this testimony, the defendant insists that the court's ultimate drug quantity determination was little more than a guess. Some examples will serve to put this argument into perspective. First, Marquis estimated the frequency of his interactions with the defendant during the critical period as "nine to twelve times." Second, Donato testified that he and the defendant purchased marijuana from Marquis "probably four to six times." In light of these and other similar imprecisions, the defendant maintains that the district court lacked a sound basis to attribute 26 kilograms of marijuana to him. We do not agree.

It is well settled that a sentencing court's selection from among plausible alternative scenarios or divergent inferences presented by the record cannot be clearly erroneous. See Platte, 577 F.3d at 393-94; United States v. Ruiz, 905 F.2d 499, 508 (1st Cir.1990). Given the ranges in the testimony concerning the frequency and volume of the drug exchanges, the district court wisely adopted the approach recommended in the amended PSI Report: it used throughout conservative estimates of the number of interactions and low-end estimates of volume. As we explain below, the court's bottom-line drug quantity determination, based squarely on these assessments, does not come close to clear error.

The sentencing court derived its drug quantity determination from four sources. First, it attributed 16 kilograms of marijuana to the defendant's purchases from Marquis. Second, it attributed five kilograms of marijuana to deliveries made by Marquis to Webber in the defendant's presence. Third, it attributed four kilograms of marijuana to joint purchases made by Donato and the defendant. Fourth, it attributed .73 kilograms of marijuana to purchases made by the defendant from Donato. The first three sub-parts of the court's drug quantity determination survive scrutiny and the fourth does not matter.

To begin, Marquis testified that he first sold marijuana to the defendant around 2002, initially in quarter-pound quantities, progressively increasing in size to pound increments, with these sales occurring every month or so. Then, for a couple of years prior to 2007, Marquis sold the defendant bags of marijuana containing anywhere from one to five pounds every month or every couple of months, and he sold as much as five pounds of marijuana on almost nine to twelve occasions.

Faced with this sales history, the district court limited its calculation of direct purchases by the defendant from Marquis to the last couple of years, effectively ignoring the previous three-year record of transactions between the two men. It exercised further restraint by disregarding all the five-pound transactions and using three pounds as the norm, assuming that Marquis sold the defendant that much marijuana every other month for two years. These consignments totaled 36 pounds of marijuana (roughly 16 kilograms). This estimate, which resolved virtually every ambiguity in Marquis's testimony in favor of the defendant, was reasonable and sufficiently supported by the record.

The same is true of the district court's quantification of the sales that Marquis made to the tandem of Webber and the defendant. Marquis testified that he delivered from one to five pounds of marijuana to this duo on six to ten occasions and that the average sale was two to three pounds. Once again, the court used a series of conservative estimates, positing six transactions of two pounds each. The end product—a finding that Webber and the defendant together purchased twelve pounds of marijuana (roughly five kilograms)—was reasonable and sufficiently supported by the record.

The third component of the district court's drug quantity determination likewise passes muster. Donato testified that the defendant introduced him to Marquis as a potential buyer. Following this introduction, Donato and the defendant together purchased around two pounds of marijuana from Marquis on four to six occasions. The court took a modest view of this testimony, positing four sales of two pounds each, for a total of eight pounds of marijuana (roughly four kilograms).1 This finding was reasonable and sufficiently supported by the record.

The defendant fairs no better with respect to the last sub-part. This portion of the court's findings attributes to the defendant.73 kilograms of marijuana bought from Donato before the defendant introduced Donato to Marquis. On this point, the defendant's chief complaint is that these purchases occurred outside the charged conspiracy. This complaint is unavailing.

Even if we assume that the conspiracy did not start until after these transactions took place, the findings previously upheld confirm that no less than 25 kilograms of marijuana were properly attributed to the defendant. See supra note 1. That amount, without more, suffices to place him well above the 20-kilogram threshold for a base offense level of 18. Accordingly, any error in including the .73 kilograms was harmless. See Calderón, 578 F.3d at 105.

We offer one final, more general, observation. Transactions between drug wholesalers and drug retailers are often conducted surreptitiously and with little if any documentation. The need for secrecy is manifest, and the keeping of records can place the participants in jeopardy. Determining drug quantities after the fact is, therefore, likely to require a careful sorting of anecdotal information and the exercise of sound judgment.

This case constitutes a paradigmatic example. The court below took a measured approach, evaluated the testimony carefully, and erred, if at all, on the side of caution. At every turn, the court used conservative figures and low-end estimates. Consequently, we conclude without serious question that the court's ultimate drug quantity determination was not clearly erroneous.

We need go no further. For the reasons elucidated above, we reject the defendant's claim of sentencing error.

Affirmed.