ORDER
The memorandum disposition filed July 8, 2002, is hereby withdrawn.
MEMORANDUM*
Arturo Campos (“Campos”), Ignacio Felix (“Felix”), Alex Manon (“Manon”) and Michael Vasquez (“Vasquez”) (collectively “Defendants”) pleaded guilty to conspiring to engage in drug trafficking on behalf of the Mexican Mafia in violation of the Racketeer Influenced and Corrupt Organizations Act, 18 U.S.C. § 1962(d). At sentencing, the district court held each defendant accountable for his own drug dealing as well as the drug trafficking of his co-defendants that occurred while he participated in the conspiracy. In addition, the district court attributed to each defendant the drug activity of Jesse Detevis (“Detevis”), an indicted co-conspirator and member of the Mexican Mafia who testified on behalf of the government.
Defendants raise various sentencing related claims on appeal, principally alleging that the district court based its drug quantity determinations on insufficient and unreliable evidence. For the most part we agree. We have jurisdiction pursuant to 28 U.S.C. § 1291, and we vacate each defendant’s sentence and remand for resentencing.
DISCUSSION
I Appropriate Standard of Proof to Determine Drug Quantity
A. Standard of Review
We review the constitutionality of a sentence de novo. United States v. Johansson, 249 F.3d 848, 853 (9th Cir.2001).
B. Analysis
Vasquez claims that the district court’s consideration of drug quantity at sentencing had a disproportionate effect on his sentence, and therefore the district court was required to compel the government to prove drug quantity by clear and convincing evidence rather than a preponderance of the evidence. See, e.g., United States v. Mezas de Jesus, 217 F.3d 638, 642 (9th Cir.2000) (requiring proof of sentencing factor by clear and convincing evidence when factor has “extremely disproportionate effect on sentence relative to the offense of conviction”) (quotations omitted). We disagree.
We have consistently upheld the use of the preponderance standard at sentencing to calculate a defendant’s base offense level for drug offenses. See, e.g., United States v. Harrison-Philpot, 978 F.2d 1520, 1523 (9th Cir.1992); United States v. Sanchez, 967 F.2d 1383, 1384-85 (9th Cir. 1992). Indeed, we do not regard the increase in Vasquez’s offense level due to his additional drug activity as an “enhancement.” See Harrison-Philpot, 978 F.2d at 1523 (holding that a six-fold increase in sentence “involve[d] no ‘enhancement’ of sentence” because “Harrison-Philpot was charged and convicted of conspiracy; the extent of the conspiracy caused the tremendous increase in her sentence”). Thus, the district court did not err in requiring the government to prove drug quantity attributable to Vasquez by a preponderance of the evidence.
II Detevis’s Reliability
A. Standard of Review
We review for abuse of discretion the district court’s determination that evi*607dence is sufficiently reliable to be considered at sentencing. United States v. Blitz, 151 F.3d 1002,1009 (9th Cir.1998).
B. Analysis
Each defendant, expressly or implicitly, attacks Detevis’s reliability and the district court’s reliance on his testimony. The district court noted Detevis’s shortcomings as a witness: “Detevis attempted to testify truthfully. However, his memory was suspect as to certain topics ... and his ability to specify quantities of drug dealing was limited.” Still, the court found Detevis’s testimony rehable:
I didn’t find that he [Detevis] was upright. ... That doesn’t mean he’s not credible. I was the one who had to make that determination. Listening to him, watching him, listening to the cross-examination, knowing that he had a motive to he to curry favor with the government, I knew that.... He was a careful witness. He was a responsive witness. He didn’t display any particular effort to bury anybody.... I based my opinion on those findings of credibility-
Watching Detevis testify and respond to cross-examination afforded the district court the opportunity to assess his credibility. Aware of Detevis’s motive for bias and his memory shortcomings, the district court stih found him rehable. We cannot say the district court abused its discretion in relying on Detevis’s testimony.
Ill The District Court’s Relevant Conduct Determinations
A. Standard of Review
We review for clear error the district court’s determination that a coconspirator’s conduct is relevant for sentencing purposes. See United States v. Kahlon, 38 F.3d 467, 470 (9th Cir.1994).
B. Analysis
Felix and Manon attack the district court’s relevant conduct determinations under the United States Sentencing Guidelines Manual (“U.S.S.G.”). See U.S.S.G. § IB 1.3(a) (1998). Specifically, Felix argues that the drug activity of Campos, Manon, and Detevis was neither within the scope of his agreement nor reasonably foreseeable and, thus, not properly attributed to him at sentencing. Manon makes the same argument about Campos, Felix, and Detevis. We agree.1
A defendant is not always accountable under the Sentencing Guidelines for the acts of his co-conspirators. “The sentencing court may not rely simply upon the total amount involved in the drug conspiracy, but must undertake an individualized evaluation of the amount for which [the defendant] is accountable under the Guidelines.” United States v. Garcia-Sanchez, 189 F.Sd 1143, 1147 (9th Gir.1999) (alteration in original) (quotations omitted). A conspirator may be judged only “on the basis of the quantity of drugs which he reasonably foresaw or which fell within ‘the scope’ of his particular agreement with the conspirators.” United States v. Vgeri, 51 F.3d 876, 881 (9th Cir. 1995) (quotations omitted) (emphasis added); see also U.S.S.G. § 1B 1.3(a)(1)(B)
Several illustrations from the Sentencing Guidelines provide guidance. Defendant R recruits Defendant S to distribute 500 grams of cocaine. Defendant S knows that Defendant R is the prime figure in a much larger cocaine trafficking operation, *608but so long as Defendant S’s agreement and conduct is limited to the distribution of that 500 grams, he is not accountable for the larger quantity distributed by Defendant R. U.S.S.G. § 1B1.3 cmt. n. 2(c)(7). Similarly, “[a] street-level dealer who shares a supplier with other dealers and knows of the other dealers’ sales but operates independently is only accountable for his own sales. But if the street-level dealer pools his resources and profits with the other sellers, he is accountable for all the drug sales.” Garciar-Sanchez, 189 F.3d at 1148 (citing U.S.S.G. § 1B1.3 cmt. n. 2(c)(6)) (quotations omitted); see also U.S.S.G § 1B1.3 cmt. n. 2(c)(4) (stating that child pornography retailers who purchase contraband from the same wholesaler and are aware of each other’s criminal activity, but otherwise operate independently, are responsible only for the quantity each purchased because “the scope of his jointly undertaken criminal activity is limited to that amount.”).
Here, the district court did not identify the particular scope of each defendant’s agreement. In their guilty pleas, Campos, Felix, and Manon each stated that he agreed to assist the criminal enterprise (i.e., the Mexican Mafia), knowing and agreeing that other members of the enterprise would commit illegal acts involving narcotics trafficking. Campos, Felix, and Manon knew that Detevis was a member of the Mexican Mafia who engaged in widespread drug dealing; he served as a hub to each defendant’s spoke. The government, however, provided no evidence that Campos, Felix, or Manon knew one another or knowingly fostered each others’ activities. In these respects, the three codefendants are similarly situated to the recruited and street-level drug dealers in the above illustrations. Campos’s, Felix’s, and Manon’s mere awareness that others in the Mexican Mafia would distribute drugs is not particularized enough to permit attribution of the others’ conduct. Without evidence that Campos, Felix, and Manon aided and abetted or could have reasonably foreseen each others’ and Detevis’s conduct, the district court erred in attributing the drug sales of the codefendants to one another. See Garciar-Sanchez, 189 F.3d at 1148 (“[I]f [defendant] had no interest in, agreement regarding, or involvement with the sales of the others, then the district court erred in sentencing [him] for the sales of other conspiracy members.”).2
We therefore vacate Campos’s, Felix’s, and Manon’s sentences and remand for resentencing. If on remand the district court holds additional hearings during which the government proves by a preponderance of the evidence that (1) Felix, Manon, and Detevis’s drug dealing was within the scope of Campos’s agreement, (2) Campos, Manon, and Detevis’s drug dealing was within the scope of Felix’s agreement, or (3) Campos, Felix, and Detevis’s dealing was within the scope of Manon’s agreement, then the district court may resentence Campos, Manon, and Felix accordingly. As the record stands, howev*609er, there is insufficient evidence to attribute to Campos, Felix, and Manon the drug activity of their codefendants.
IV The District Court’s Drug Quantity Calculations
A. Standard of Review
We review de novo the method adopted by the district court to estimate relevant drug quantity. United States v. Scheele, 231 F.3d 492, 497 (9th Cir.2000). The determination of drug quantity is a factual finding reviewed for clear error. United States v. Asagba, 77 F.3d 324, 325 (9th Cir.1996). Though drug quantity need be established only by a preponderance of the evidence for sentencing purposes, we require that the district court base its findings solely on evidence possessing “ ‘sufficient indicia of reliability to support its probable accuracy.’ ” United States v. August, 86 F.3d 151, 154 (9th Cir.1996) (quoting U.S.S.G. § 6A1.3(a)). Where, as here, no drugs were seized, the district court “shall approximate the quantity of the controlled substance,” U.S.S.G. § 2D1.1 cmt. n. 12, but we require the court’s estimate to “err on the side of caution.” August, 86 F.3d at 154.
B. Analysis
Each defendant contests the drug quantity attributed specifically to him and the amount attributed to Detevis. We address the district court’s findings for each defendant in turn.
1. Campos
The district court estimated that Campos distributed 151 grams of methamphetamine (“meth”). Evidence of Campos’s drug activity came entirely from Detevis’s testimony. Detevis specifically remembered occasions where Campos sold two ounces of meth (56.7 grams), purchased one-sixteenth ounce of meth (1.75 grams) from Detevis, and delivered to Detevis one-half ounce of meth (14 grams) that Campos picked up from another dealer. Detevis testified that he saw Campos deal drugs five or six other times but could not remember any details. While this evidence supports a finding that Campos dealt roughly 71 grams of meth, we cannot determine how the district court settled on 151 grams.3 Our ability to review a defendant’s drug sentence depends on the district court’s clearly setting forth its method of calculation. We cannot discern any method from the record before us. Thus, while we appreciate the difficulty faced by the district court in estimating Campos’s drug activity, we hold that the court erred by attributing 151 grams of meth to Campos. We vacate the district court’s meth determination above 71 grams and remand for resentencing.
2. Felix
The district court estimated that Felix distributed 151 grams of meth and one-eighth ounce of crack cocaine. Evidence of Felix’s drug activity came entirely from Detevis’s testimony. Detevis remembered one occasion where he and Felix went to Perez’s house to sell one ounce of meth (28.35 grams). Detevis testified also that Felix told him about selling one ounce of meth (28.35 grams) to Moreno, and then gave Detevis $450 which Felix allegedly said was from the Moreno sale. Under our review for clear error, this testimony would support a finding that Felix dealt 56.7 grams of meth.
*61056.7 grams, however, is not 151 grams, and the record lacks evidence with sufficient indicia of reliability to support the nearly 100 gram difference. Detevis testified he was with Felix on “more than ten” other occasions when Felix sold drugs. Though Detevis did not specify how many of the “more than ten” sales involved meth,3 he estimated that Felix sold anywhere from one-sixteenth to one-eighth ounce of meth on those occasions. Even if the district court attributed to Felix ten additional sales of meth — -perhaps a speculative finding on this record— and conservatively estimated that each involved one-sixteenth ounce of meth, only another ten-sixteenths of an ounce of meth (17.72 grams) would result. When added to the 56.7 grams discussed above, approximately 74.5 grams of meth would be attributable to Felix. Hence, the district court clearly erred in attributing 151 grams of meth to Felix. We vacate the district court’s meth determination above 56.7 grams and remand for resentencing.4
3. Manon
The district court estimated that Manon distributed one pound of meth. Evidence of Manon’s drug activity came entirely from Detevis’s testimony. Detevis testified that he sold Manon one pound of meth in April 1998. Detevis remembered another instance when he saw Ma-non sell one-quarter pound of meth. In addition, the government played a tape recording of Detevis, Manon, and Woody, in which Manon bragged about selling meth “like hot cakes.” From this evidence, the district court estimated that Manon distributed one pound of meth during his participation in the conspiracy. This finding was not clearly erroneous.
4. Vasquez
The district court estimated that Vasquez distributed eight grams of crack cocaine. Evidence of Vasquez’s drug activity came entirely from Detevis’s testimony. Detevis testified that he saw Vasquez sell crack cocaine more than ten times. Detevis bought two grams of crack from Vasquez on three or four of those occasions, purchasing a total of six to eight grams. Detevis specifically recalled another instance where he saw Vasquez sell two grams of crack. Detevis further testified that each of the ten transactions involved amounts ranging from two grams to one-half ounce. From this testimony, the district court estimated that Vasquez distributed eight grams of crack cocaine. This finding was not clearly erroneous.
5. Detevis
The district court held that “Detevis surely dealt in at least 500 grams of crack cocaine and 500 grams of meth” during the period of the Defendants’ participation in the drug conspiracy. Defendants objected at sentencing that the district court spun these quantities from thin air. The district court responded that the amount attributed to Detevis “was not based on speculation or surmise or guess. I reached that conclusion based upon his testimony and the circumstances of his involvement, my knowledge of the tapes and the transcripts and the testimony at the hearing.”
The Defendants’ plea agreements did not hint at their involvement with specific quantities of drugs. The only potentially *611relevant evidence on this issue was Detevis’s testimony. Detevis testified only to quantities distributed by the other defendants, and though Detevis participated in several of these transactions, the district court did not indicate which of these transactions it relied upon to estimate Detevis’s drug activity.
The government fails to clarify matters. Conceding in its briefs that Detevis never testified specifically about the amount of drugs he distributed, the government gamely maintains that a preponderance of the evidence demonstrated the 500 grams of meth and 500 grams of crack cocaine attributed to him. The government notes that Detevis’s testimony “reveals that the over-all conspiracy was distributing significant amounts of drugs.” The government points also to Detevis’s plea agreement in which he admitted to “numerous conversations” with members and associates of the Mexican Mafia concerning narcotics distribution and to selling 248 grams of crack cocaine on one occasion. Though the transaction involving 243 grams is not attributable to Defendants because it occurred before they were involved with the conspiracy, the government urges that this sale “providefs] evidence of the scale of his drug-dealing activities” while Defendants were involved. We disagree.
We have readily recognized the need and the appropriateness of estimating drug quantity in cases where none is seized. See, e.g., Scheele, 231 F.3d at 498 (estimation by multiplying estimated number of transactions by estimated average amount per transaction); August, 86 F.3d at 154 (estimation based on lab capacity); United States v. Basinger, 60 F.3d 1400, 1409-10 (9th Cir.1995) (approximation from two empty ephedrine containers); United States v. Williams, 989 F.2d 1061, 1073 (9th Cir.1993) (capacity of glassware and precursor chemicals). Here, however, the district court gave no indication of the bases for its calculations. Even conceding the sweep of Detevis’s drug dealing, the evidence gives no assurance that the quantities attributed to him are not off by a factor of two or ten or twenty. Moreover, to the extent the district court estimated Detevis’s drug activity based on the Defendants’ trafficking, attributing his conduct to each defendant risks punishing them twice for the same transactions.
In short, the evidence relied upon by the district court lacks the “sufficient indicia of reliability” to ensure the accuracy of its estimation of Detevis’s drug activity. We therefore vacate Defendants’ sentences and remand for resentencing without consideration of the 500 grams of meth and 500 grams of crack cocaine previously attributed to Detevis. On remand, the district court is free to reconsider the quantity of drugs trafficked by Detevis so long as its findings are supported by evidence with sufficient indicia of reliability.
VACATED and REMANDED for re-sentencing.
This disposition is not appropriate for publication and may not be cited to or by the courts of this circuit except as may be provided by Ninth Circuit Rule 36-3.
. Campos filed a Notice of Joinder which adopted Felix's and Manon’s arguments concerning the relevant conduct determinations. Thus, he benefits from our ruling on this issue.
. Even if Campos’s, Felix’s, or Detevis’s conduct were properly attributable to Manon, we would vacate his sentence and remand for resentencing. The district court determined that Manon participated in the conspiracy from June to July 1998, whereas Campos participated from June to August 1998, and Felix participated from June 1998 to February 1999. The district court attributed Campos’s and Felix’s entire drug activity to Ma-non; the court should have attributed only those portions that occurred during Manon's involvement with the enterprise — i.e., June to July 1998. See U.S.S.G. § 1B1.3 cmt. n. 2. For the same reason, we would vacate Campos’s sentence because the district court held him responsible for Felix’s entire drug activity despite the fact that Campos's involvement in the conspiracy ended several months before Felix's (August 1998 vs. February 1999).
. Even if the district court credited Detevis’s testimony by attributing another five transactions to Campos, and conservatively estimated that each involved one-sixteenth ounce of meth, only 8.75 grams more meth would result.
. Detevis testified that some of the sales involved crack cocaine.
. Detevis testified he and Felix distributed one-eighth of an ounce of crack cocaine once a week for several months. Thus, the district court’s conservative finding that Felix distributed a total of one-eighth of an ounce of crack cocaine is not clearly erroneous and need not be reconsidered on remand.