Gene E. Beler appeals the sentence he received after pleading guilty to a series of drug offenses. Beler specifically challenges the quantity of cocaine used by the district court in calculating his base offense level under the Sentencing Guidelines. Beler maintains that the evidence supporting the district court's finding of drug quantity was unreliable and that the government failed to establish that the additional cocaine sales were part of the same course of conduct as the sales of conviction. Due to the undeveloped record below and the fact that the district court made only a conclusory finding of drug quantity in the face of inconsistent evidence, we are not convinced that the court's finding is supported by a preponderance of the reliable evidence. We therefore vacate Beler's sentence and remand for resentencing.

A five-count indictment charged Beler with various offenses relating to his distribution of cocaine to a government informant during October and November 1991. Three counts charged that Beler had distributed cocaine, one count charged possession with the intent to distribute, and the final count charged that Beler had used a firearm in relation to a drug trafficking offense. See 21 U.S.C. Sec. 841(a)(1); 18 U.S.C. Sec. 924(c). Beler pled guilty to the distribution and possession counts but went to trial on the gun charge. He was acquitted by a jury on July 31, 1992.

Neither the guilty pleas nor the trial are directly at issue in this appeal, however. Rather, Beler challenges only the sentence imposed by the district court, and in particular the court's calculation of the quantity of cocaine on which that sentence was based. Beler's three distribution convictions involved 3.3 grams of cocaine, and the possession conviction involved another 28.5 grams, yielding a total of 31.8 grams from the offenses of conviction. Yet the district court attributed an additional 592.05 grams to Beler under Sentencing Guidelines Sec. 1B1.3(a)(2), and the court accordingly sentenced Beler on the basis of 623.85 grams.

The district court relied on evidence from two sources to support its relevant conduct finding--the affidavits of Jerry Covington, the government's informant and Beler's former friend, and the trial testimony of Michael Truitt. Covington testified on behalf of the government at Beler's trial, but during his direct examination, he was unable to estimate the quantity of cocaine he had purchased from Beler over the years. Yet prior to Beler's sentencing, Covington signed an affidavit stating that he had known Beler for thirty years, that he had purchased cocaine from him "from approximately February, 1976 to approximately November 12, 1991," and that he had bought approximately "150 to 200 ounces" from Beler during that period. After Beler's counsel questioned this estimate, Covington signed a second affidavit, which stated:

The affidavit I signed on September 23, 1992, in connection with this case contains a typographical error and an inaccuracy in paragraph 3 which I wish to correct. The approximate amount of cocaine that I purchased from Gene Beler from approximately February 1976 to approximately November 12, 1991 was actually 15 to 20 ounces rather than 150 to 200 ounces.

Covington also attested to the following additional information in the second affidavit:

[A]ll of the cocaine that I purchased from Gene Beler was purchased through hand to hand buys. I paid cash or Gene Beler extended me credit when I purchased this cocaine. The purchases of cocaine from Gene Beler generally took place at his residence located at 2035 East Edwards Street, Springfield, Illinois. Some purchases occurred at taverns in the area and at my residences located at 2040 East Hamilton or 26 Royal, Springfield, Illinois. The cocaine I purchased from Gene Beler was for my own personal use. This cocaine was in powder form. It was sold to me generally in gram or on occasion in one sixteenth ounce quantities. Generally, during these drug purchases Gene Beler and I were the only persons present.

The district court relied on these affidavits to conclude that Beler had engaged in a common scheme of trafficking in cocaine between 1976 and 1991. The court credited the estimate in the second affidavit and concluded that Beler had sold fifteen ounces (425.25 grams) to Covington during this period. The district court added another six ounces (170.1 grams) in accordance with Michael Truitt's trial testimony that he had purchased between six and seven ounces of cocaine from Beler over the last three to four years. This yielded a total drug quantity of 623.85 grams, which corresponded to a Guidelines sentencing range of between 63 and 78 months. The district court sentenced Beler to 66 months in prison.

Beler appeals, arguing that the affidavits were unreliable because they contradicted Covington's trial testimony, and that Truitt's estimate did not withstand cross-examination. Beler further argues that the government failed to establish that these additional sales were part of the "same course of conduct or common scheme or plan" as the offenses of conviction. See U.S.S.G. Sec. 1B1.3(a)(2).

A district court's calculation of the quantity of drugs involved in an offense is a finding of fact, which we review under a clearly erroneous standard. United States v. Montgomery, 14 F.3d 1189, 1196 (7th Cir.1994); United States v. Cedano-Rojas, 999 F.2d 1175, 1179 (7th Cir.1993). We must affirm the court's drug quantity finding unless " 'on the entire evidence [we are] left with the definite and firm conviction that a mistake has been committed.' " United States v. McMillen, 8 F.3d 1246, 1250 (7th Cir.1993) (quoting United States v. Duarte, 950 F.2d 1255, 1262 (7th Cir.1991), cert. denied, --- U.S. ----, 113 S. Ct. 174, 121 L. Ed. 2d 120 (1992)), cert. denied, --- U.S. ----, 114 S. Ct. 1649, 128 L. Ed. 2d 368 (1994); see also United States v. Sykes, 7 F.3d 1331, 1335 (7th Cir.1993). The government bears the burden of establishing the quantity of drugs by a preponderance of the evidence. McMillen, 8 F.3d at 1250; United States v. Jackson, 983 F.2d 757, 771 (7th Cir.1993).

In calculating a defendant's base offense level under the Guidelines, the sentencing court must consider types and quantities of drugs not specified in the counts of conviction but that were "part of the same course of conduct or common scheme or plan" as the convicted offenses. U.S.S.G. Sec. 1B1.3(a)(2); see also U.S.S.G. Sec. 3D1.2(d). "The defendant need not have been either charged with or convicted of carrying out these other acts." United States v. Thomas, 969 F.2d 352, 355 (7th Cir.), cert. denied, --- U.S. ----, 113 S. Ct. 274, 121 L. Ed. 2d 202 (1992); see also Duarte, 950 F.2d at 1263.1 This "relevant conduct" or "aggregation" rule provides the government tremendous leverage in drug cases. See United States v. Sepulveda, 15 F.3d 1161, 1198 (1st Cir.1993). We observed in Duarte that the government's advantage "derives from the arbitrage-like quality of criminal trials--prosecutors must prove beyond a reasonable doubt all elements of the crimes charged in the indictment, but, once having garnered a conviction, need only prove by a preponderance of the evidence facts alleged at sentencing." 950 F.2d at 1263; see also United States v. Ebbole, 917 F.2d 1495, 1501 (7th Cir.1990) (relevant conduct rule "invite[s] the prosecutor to indict for less serious offenses which are easy to prove and then expand them in the probation office." (internal quotation omitted)).

The government's burden at sentencing is further eased by the fact that the Federal Rules of Evidence do not apply, and the court is thus free to consider a wide range of information, including hearsay evidence, that may have been inadmissible at the defendant's trial. United States v. Corbin, 998 F.2d 1377, 1385 (7th Cir.1993), cert. denied, --- U.S. ----, 114 S. Ct. 1124, 127 L. Ed. 2d 432 (1994); United States v. Johnson, 997 F.2d 248, 254 (7th Cir.1993). The Third Circuit recently noted, however, that a number of commentators have begun to criticize these relaxed evidentiary standards due to "the critical role of district court fact finding under the Guidelines sentencing regime." United States v. Miele, 989 F.2d 659, 663 n. 5 (3d Cir.1993) (citing commentary). Indeed, this court too has emphasized that the district court's factual findings at sentencing have a significant impact on the length of the sentence ultimately imposed, and that is particularly true in drug cases, as section 2D1.1(a)(3) sets a drug offender's base offense level in accordance with the type and quantity of illegal substance involved. Duarte, 950 F.2d at 1262; United States v. Buggs, 904 F.2d 1070, 1080 (7th Cir.1990); see also Sepulveda, 15 F.3d at 1196-1197 ("In drug-trafficking cases under the sentencing guidelines, sentences are largely quantity-driven"); United States v. Paulino, 996 F.2d 1541, 1545 (3d Cir.), cert. denied, --- U.S. ----, 114 S. Ct. 449, 618, 126 L. Ed. 2d 382, 582 (1993); United States v. Simmons, 964 F.2d 763, 773 (8th Cir.), cert. denied, --- U.S. ----, 113 S. Ct. 632, 121 L. Ed. 2d 563 (1992). Despite these concerns, however, the aggregation rule, with its lesser burden of proof and relaxed evidentiary standards, is mandated by the Guidelines themselves, and we have applied the rule faithfully in a long line of cases. See, e.g., Montgomery, 14 F.3d at 1196-98; United States v. Rivera, 6 F.3d 431, 445 (7th Cir.1993), cert. denied, --- U.S. ----, 114 S. Ct. 1098, 127 L. Ed. 2d 411 (1994); Cedano-Rojas, 999 F.2d at 1180-81; Thomas, 969 F.2d at 355-56; United States v. Nunez, 958 F.2d 196, 198-99 (7th Cir.), cert. denied, --- U.S. ----, 113 S. Ct. 168, 121 L. Ed. 2d 115 (1992); Duarte, 950 F.2d at 1263 (compiling additional cases).

The rule, however, is not without limits (see Duarte, 950 F.2d at 1263), and because its application so favors the government, we must be scrupulous to ensure that the government has adhered to those limits. The rule's primary limitation is of course that any act used to enhance a defendant's sentence must be "part of the same course of conduct or common scheme or plan as the offense of conviction." U.S.S.G. Sec. 1B1.3(a)(2). We first explained in United States v. White, 888 F.2d 490, 500 (7th Cir.1989), and we have reiterated in subsequent cases, that section 1B1.3(a)(2) should not be applied to "[o]ffenses [that are] of the same kind, but not encompassed in the same course of conduct or plan" as the convicted offenses. (Emphasis in original); see, e.g., Sykes, 7 F.3d at 1336; United States v. Pollard, 965 F.2d 283, 287 (7th Cir.1992); Duarte, 950 F.2d at 1264; see also United States v. Hahn, 960 F.2d 903, 910-11 (9th Cir.1992). We have thus required sentencing courts to "explicitly state and support, either at the sentencing hearing or (preferably) in a written statement of reasons, [the] finding that the unconvicted activities bore the necessary relation to the convicted offense." Duarte, 950 F.2d at 1263; see also Jackson, 983 F.2d at 771-72; Thomas, 969 F.2d at 355; Pollard, 965 F.2d at 288.

A related limitation is that the government must prove a defendant's additional conduct by reliable evidence. We have held that a criminal defendant "has a due process right to be sentenced on the basis of reliable information" (United States v. Campbell, 985 F.2d 341, 348 (7th Cir.1993); see also United States v. Westbrook, 986 F.2d 180, 182 (7th Cir.1993); United States v. Atkinson, 979 F.2d 1219, 1224 (7th Cir.1992)), and the Guidelines themselves reflect our concern with reliability. Section 6A1.3(a) authorizes a court to "consider relevant information without regard to its admissibility under the rules of evidence applicable at trial, provided that the information has sufficient indicia of reliability to support its probable accuracy." (Emphasis added); see United States v. Atkinson, 15 F.3d 715, 720 (7th Cir.1994); United States v. Soria, 965 F.2d 436, 443 (7th Cir.1992); United States v. Miller, 891 F.2d 1265, 1270 (7th Cir.1989). Where the drugs actually purchased or seized from the defendant do not necessarily reflect the scope of his illegal operation, the district court must "approximate the quantity of the controlled substance" at issue, and the court may consider, for example, "the price generally obtained for the controlled substance, financial or other records, similar transactions in controlled substances by the defendant, and the size or capability of any laboratory involved." U.S.S.G. Sec. 2D1.1 application note 12; McMillen, 8 F.3d at 1250; United States v. Cagle, 922 F.2d 404, 407 (7th Cir.1991). As section 6A1.3(a) indicates, however, the information underlying the court's approximation must possess sufficient indicia of reliability. United States v. Ortiz, 993 F.2d 204, 207 (10th Cir.1993); see also United States v. Kozinski, 16 F.3d 795, 818-19 (7th Cir.1994). Unreliable allegations must not be considered. U.S.S.G. Sec. 6A1.3 commentary; see also Ortiz, 993 F.2d at 207. The Third Circuit has instructed that section 6A1.3(a)'s reliability standard must be rigorously applied (Miele, 989 F.2d at 664), and we certainly agree. See also Sepulveda, 15 F.3d at 1198; Paulino, 996 F.2d at 1545 ("the sentencing court must carefully scrutinize the government's proof to ensure that its estimates are supported by a preponderance of the evidence").2 The record in this case leaves us unconvinced that the district court conducted a sufficiently searching inquiry into the government's evidence to ensure its probable accuracy. The sentence imposed on the basis of that evidence therefore cannot stand.

The evidence that allegedly supports the district court's finding of drug quantity is twofold. First, there is the Covington affidavit, or more accurately, the two Covington affidavits. Our duty to scrutinize the government's evidence of drug quantity compels some comment on the rather troubling error in Covington's first affidavit, an error that resulted in an estimate of drug quantity ten times higher than the corrected estimate. The government characterized the discrepancy as a typographical error, and the district court seemed to agree. But the record suggests that the court reached that conclusion without any further inquiry into the matter. Covington was never required to explain under oath the discrepancy between his two estimates. Indeed, neither party presented any live testimony in the sentencing phase of the case. The government's typographical error explanation is certainly plausible, given that the corrected numbers merely removed zeros from the originals. Yet even so, the error reflects badly on the reliability of the affidavits as a whole. We must assume, after all, that Covington reviewed and signed the initial affidavit after the typographical error had been made. If he were truly sure of the number of ounces purchased from Beler, we would expect him to have caught the error.

But there was other evidence suggesting that Covington was not so sure. In addition to his affidavits, Covington testified on the government's behalf at Beler's trial. He referred generally to a series of cocaine transactions with Beler, but when asked on direct examination to approximate the amount of cocaine he had purchased from Beler over the years, Covington was unable to do so, responding: "It would be really hard to pin down to an exact amount, sir, because there was a lot of money spent, lot of damage done." (July 28, 1992 Tr. at 15.) Beler called this testimony to the district court's attention at sentencing, but the court apparently believed it had no impact on the reliability of the estimates in the second affidavit, as the court concluded that Covington's sworn affidavit was uncontroverted. (Nov. 30, 1992 Tr. at 10.) We have some difficulty with the court's conclusory finding. Although Beler failed to come forward with his own alternative estimate of drug quantity, he did succeed in discrediting Covington's estimate by pointing to Covington's inability to make a similar estimate of drug quantity at trial. Because the affidavits offered only bare estimates with no explanation of how Covington may have arrived at those figures, we think Beler's reliance on the trial testimony was sufficient to raise a question about the reliability of Covington's affidavits. We thus believe that the district court should have further explored the factual basis for the estimate before accepting it as uncontroverted.3

Covington's drug quantity estimate was also not the only variance between the affidavits and the trial testimony. For instance, Covington attested in the affidavits to almost a sixteen-year course of dealing that began in February 1976 and ended finally with Beler's arrest in November 1991. Yet Covington indicated at trial that although he had been acquainted with Beler as early as 1976, he had not begun to purchase cocaine from the defendant until 1983 or 1984. (July 28, 1992 Tr. at 6-7).4 The affidavits thus added approximately seven years to the relevant course of dealing.5 Although the district court heard the evidence at trial, it made no attempt to reconcile the affidavits' description of the course of dealing with that inconsistent testimony. The court instead relied on the "uncontroverted" affidavits to conclude by a preponderance of the evidence that Beler had been selling drugs to Covington over a sixteen-year period.

We are troubled by the lack of any inquiry below into this potentially crucial variance, and we are unwilling to assume that the length of the course of dealing had no impact on Covington's estimate of drug quantity. As we have mentioned, the affidavit offers no insight into how Covington arrived at the fifteen to twenty ounce figure, but given the length of the course of dealing, it is reasonable to assume that Covington may have estimated his weekly or monthly cocaine purchases and then multiplied by the number of weeks or months that he had been involved with Beler. That is an acceptable method of estimating drug quantity. See, e.g., United States v. Welch, 945 F.2d 1378, 1384 (7th Cir.1991) (total quantity based on informant's testimony that he had sold perhaps one to two ounces of cocaine a month for the defendant over approximately an eight-month period), cert. denied, --- U.S. ----, 112 S. Ct. 1235, 117 L. Ed. 2d 469 (1992); United States v. Miller, 891 F.2d 1265, 1269 (7th Cir.1989) (quantity calculated on the basis of one ounce every three weeks over the course of a conspiracy). Given Covington's inability to provide an estimate at trial, it is not unlikely that he would have employed such a methodology in making the approximation here. If he did, the resulting estimate may be seriously flawed because Covington may have erroneously added seven years to the relevant course of dealing. Of course, we are only speculating because, given the lack of any inquiry below into the basis for Covington's estimate, we have no way of knowing what impact the seven years may have had. But this unexplained discrepancy between the "uncontroverted" affidavits and Covington's own trial testimony certainly adds to the level of our discomfort with the district court's finding of drug quantity.

One other aspect of the Covington affidavits gives us pause. Covington purports in the first affidavit to have known Beler for approximately thirty years, but he testified at trial that he had known the defendant for "ten to twelve years or longer." (July 28, 1992 Tr. at 5.) When pressed, he stated that he had known Beler since he had been about fourteen or fifteen years old, and Covington was thirty-seven at the time of his testimony. (Id.) The inconsistencies on this relatively minor point do not instill confidence in Covington's ability to estimate the quantity of drugs he had purchased from Beler over what, according to the government, was almost a sixteen-year course of dealing.

Our discomfort with Covington's affidavits stems in part from these inconsistencies, but we are also aware that Covington was a cocaine addict during most, if not all, of the relevant time period. During cross-examination at trial, Covington admitted that he had been addicted to cocaine in 1991 when he became a government informant and that his addiction had spanned approximately ten years. (July 28, 1992 Tr. at 41, 44.) He also indicated in his direct testimony that his cocaine use had been "heavy" from approximately 1983 to 1989. (Id. at 15.) Indeed, in explaining at trial why he was unable to estimate the volume of his purchases from Beler, Covington said that "there was a lot of money spent, [a] lot of damage done." (Id.) Despite this addiction to cocaine and despite the passage of time, Covington was able to come up with an estimate post-trial. In these circumstances, we believe the district court should have subjected any information provided by Covington to special scrutiny in light of his dual status as a cocaine addict and government informant. Such a heightened standard of scrutiny is consistent with the rule adopted by a number of other circuits when considering a drug quantity estimate provided by an informant who has a history of drug addiction. See Miele, 989 F.2d at 666-67 ("Because of the questionable reliability of an addict-informant, we think it is crucial that a district court receive with caution and scrutinize with care drug quantity or other precise information provided by such a witness before basing a sentencing determination on that information"); Simmons, 964 F.2d at 776 (rejecting drug quantity testimony of an informant who had lied under oath at trial and whose memory was impaired by "a history of cocaine addiction"); United States v. Robison, 904 F.2d 365, 371-72 (6th Cir.1990) (rejecting drug-quantity estimate of heavy drug user who admitted that relevant time period was "very hazy" due to her drug use), cert. denied, 498 U.S. 946, 111 S. Ct. 360, 112 L. Ed. 2d 323 (1990). As our previous discussion indicates, the district court in this case did not subject Covington's affidavits to the searching scrutiny we must require; instead, the court merely accepted the affidavits at face value. In light of the inconsistencies we have identified between Covington's affidavits and his trial testimony, as well as Covington's status as an informant who had been addicted to cocaine, we find that the court clearly erred in doing so.

Our rejection of the drug quantity estimate in Covington's second affidavit does not bar Covington from providing drug quantity information on remand. But we must require that the district court scrutinize that information to ensure that it possesses "sufficient indicia of reliability to support its probable accuracy." U.S.S.G. Sec. 6A1.3(a). In light of the questions raised by the current record and of the importance of the drug quantity determination to Beler's sentence, we believe that a hearing would aid the district court's resolution of this disputed issue on remand. See U.S.S.G. Sec. 6A1.3 commentary ("An evidentiary hearing may sometimes be the only reliable way to resolve disputed issues"); see also United States v. Roberts, 14 F.3d 502, 521 (10th Cir.1993); Robison, 904 F.2d at 371-72.

The district court also credited Michael Truitt's direct testimony that he had purchased six to seven ounces of cocaine from Beler during the last three to four years. Truitt testified that he had purchased cocaine from Beler "over ten times," although he could not remember how many times more than ten. (July 29, 1992 Tr. at 85.) Truitt conceded on cross-examination, however, that the six to seven ounce figure was only a "rough estimate" and that his memory was not all that good. (Id. at 89.) Beler's counsel then attempted to elicit more specific information about Truitt's purchases. Truitt testified that less than half the time (approximately four times), he had purchased quarter ounces from Beler. (Id. at 90-91.) A quarter ounce equals approximately seven grams, meaning that Truitt could account through quarter-ounce purchases for approximately twenty-eight grams of the six to seven ounces he allegedly had purchased. The remaining purchases (approximately six) involved "eightballs," which are one eighth of an ounce, or approximately three and one-half grams, of cocaine. (Id. at 91.) The eightball purchases accounted for an additional twenty-one grams, bringing the total to approximately forty-nine grams. Yet the six to seven ounces to which Truitt testified on direct examination corresponds to between 168 and 196 grams of cocaine, leaving more than 100 unaccounted-for grams. The government explains the discrepancy by emphasizing that Truitt had referred to "over ten" purchases, whereas counsel's cross-examination had assumed ten or close to ten purchases. That is true, but it does not sufficiently explain a total quantity estimate that was more than three times the amount that Truitt could specifically account for on cross-examination. Moreover, Truitt could provide no indication of how many additional purchases he had made, and he admitted that his memory was not very good.

Again, the district court failed to mention the potentially inconsistent cross-examination testimony, indicating merely that the six to seven ounce figure had come from Truitt's "courtroom testimony." (Nov. 30, 1992 Tr. at 11.) In Duarte, we explained that where a sentencing court "relies upon one of two contradictory statements offered by a single witness, it should directly address the contradiction and explain why it credits one statement rather than the other." 950 F.2d at 1266; see also Miele, 989 F.2d at 664 (refusing to uphold a sentence imposed on the basis of the higher of two inconsistent estimates where the district court failed to address the inconsistency); cf. United States v. Walton, 908 F.2d 1289, 1302 (6th Cir.) ("when choosing between a number of plausible estimates of drug quantity, none of which is more likely than not the correct quantity, a court must err on the side of caution"), cert. denied, 498 U.S. 906, 989, 990, 111 S. Ct. 273, 530, 532, 112 L. Ed. 2d 229, 541, 542 (1990). We believe the Duarte principle applies here where the witness makes a conclusory estimate in his direct testimony and then is unable to support that figure on cross-examination. Before the court relies on the higher estimate, it must provide some explanation for its failure to credit the inconsistent statement on cross-examination. Cf. United States v. Casas, 999 F.2d 1225, 1230 (8th Cir.1993) (district court found witness' estimate credible despite forceful cross-examination), cert. denied, --- U.S. ----, 114 S. Ct. 894, 127 L. Ed. 2d 86 (1994). Absent such an explanation, we cannot conclude that a preponderance of the evidence supports the court's finding. See Sepulveda, 15 F.3d at 1198-1199. We thus direct that the district court revisit Truitt's testimony on remand.

It is clear, as the government asserts, that the scope of Beler's cocaine dealings extended beyond the 31.8 grams for which he was convicted. Yet the fact that Beler made other cocaine sales does not corroborate the specific estimates of drug quantity provided either by Covington or Truitt. The relevant question under section 2D1.1(a)(3) is not whether the defendant engaged in other sales, but rather, how much cocaine was involved. As the Third Circuit explained in Miele, "a determination that [the defendant's] drug activity was substantial does not translate readily into a specific drug quantity finding, which is the ultimate issue for sentencing purposes." 989 F.2d at 668; see also Roberts, 14 F.3d at 522; Ortiz, 993 F.2d at 208 ("the relevant issue is not whether Defendant distributed marijuana, but rather the quantity of marijuana that Defendant distributed").

In addition to the issue of drug quantity, the district court should also more specifically address on remand whether the sales to Covington and Truitt were "part of the same course of conduct or common scheme or plan as the offense[s] of conviction." U.S.S.G. Sec. 1B1.3(a)(2). The district court found that the additional sales were all part of a common scheme, but that finding too was based primarily on Covington's affidavits, which we are requiring the court to revisit on remand.

Whether conduct is included in a single course of conduct or common scheme or plan depends upon the "similarity, regularity, and temporal proximity of the incidents in question." Montgomery, 14 F.3d at 1198 (quoting Cedano-Rojas, 999 F.2d at 1180); see also Sykes, 7 F.3d at 1336. It would seem that most, if not all, of the additional sales here may meet the similarity test, but we are less sure of their regularity and temporal proximity. On the basis of Covington's affidavits, the district court found a course of dealing that extended for almost sixteen years, one of the longest such periods that we have encountered. Yet the undeveloped record below provides little or no insight into the regularity of the sales making up that alleged course of dealing. Indeed, the evidence relating to those sales is so general that we know very little about when or how regularly they occurred. One thing we do know is that the sales to Covington were not necessarily continuous. They seemed to begin in 1983 or 1984 and to have continued on a consistent basis until 1989, when, according to both Covington and Beler, there was a break in their relationship. We cannot be sure how long the break lasted, but that could be significant, as it could impact on whether there was a single course of dealing or common scheme that included the sales of conviction. If the earlier course of dealing ended in 1989, any cocaine sales prior to that date would not be linked to the 1991 sales for which Beler was convicted. Under those circumstances, the earlier sales should not be included as relevant conduct under section 1B1.3(a)(2). The district court should consider this issue, as well as the relationship of the Truitt sales to the sales of conviction, in making a relevant conduct finding on remand.

We are mindful that by remanding for resentencing after a hearing and further fact finding, we impose a burden on the district courts far greater than that they experienced pre-Guidelines. This of course is regrettable in light of the increasingly crowded dockets that our district judges face. Yet today's decision is necessitated by the fact-intensive nature of sentencing decisions under the Guidelines. The record below leaves us unconvinced that the conclusory drug quantity estimates of either Covington or Truitt were reliable. The district court's equally conclusory finding based on those estimates does nothing to assuage our concerns. Gene Beler simply has too much at stake for us to be satisfied with a conclusory factual finding based on potentially unreliable evidence. We therefore vacate Beler's sentence and remand for resentencing consistent with this opinion.

VACATED AND REMANDED.