ORDER
Last time this case was before us, the district court had sentenced Cruz Saenz to 293 months’ imprisonment for transporting drug money, but we vacated and remanded because there was no evidentiary basis to support the district court’s denial of a minor role reduction pursuant to U.S.S.G. § 3B1.2. United States v. Saenz, 623 F.3d 461, 467 (7th Cir.2010) (hereinafter “Saenz I ”). We then explained that applying the minor role reduction requires comparing the culpability of the defendant to that of the average member of the conspiracy. See id. at 468. On remand, after a hear*545ing, the district court sentenced Saenz to 252 months’ imprisonment and again denied the minor role reduction, but the district court did not compare Saenz’s culpability to that of the average member of the conspiracy, which was error. We acknowledge that determining the culpability of an “average” member of a complex conspiracy is not straightforward, and that Saenz I did not provide sufficient guidance on this score, which we hope to rectify in this order. But because of the error, and because it is not clear that Saenz would have received the same 252-month sentence had the minor role reduction been applied, we vacate his sentence and remand.
I. BACKGROUND
As this is the second appeal in this case and the issue on appeal is narrow, we summarize the facts only briefly. Saenz and his co-defendants were involved in a large-scale cocaine distribution network based in Mexico. Those based in Mexico obtained cocaine and arranged cocaine shipments out of Mexico and into the United States. Others in Texas arranged cocaine deliveries into Indianapolis. Upon delivery, another co-defendant passed the cocaine onto suppliers in the region and also arranged for truck drivers to take cocaine proceeds back to Mexico and Texas. Saenz was one of those truck drivers, and on one occasion attempted to deliver about $500,000 in cocaine proceeds to Texas, when he was arrested. Saenz was convicted of conspiring to distribute more than five kilograms of cocaine in violation of 21 U.S.C. § 846, and he was sentenced to 293 months’ imprisonment. During the sentencing proceeding, the district court found that Saenz was “far more than a courier” and denied him a minor role reduction under U.S.S.G. § 3B1.2.
On appeal, we vacated the sentence and remanded, principally because the district court’s finding that Saenz was “far more than a courier” lacked an evidentiary basis. See Saenz I, 623 F.3d at 467. We added:
[T]he minor participant determination is heavily fact-dependent. A drug courier should neither automatically receive nor automatically be precluded from receiving a role reduction. Instead, the controlling standard is whether the defendant is substantially less culpable than the average participant in the offense. When assessing whether a defendant is substantially less culpable such that he should receive a role reduction, a defendant’s role should be compared to that of the average member of the conspiracy, not with the leaders. The question here is whether this one-time courier is less blameworthy than the average defendants in the conspiracy.
Id. at 468 (citations and quotation marks omitted). We noted that Saenz was “certainly less culpable” than the Mexico-based co-defendants who coordinated the cocaine shipments from Mexico, that he was less culpable than the co-defendant in charge of the Indianapolis hub of the conspiracy, that he never touched any drugs or participated in any negotiations, and that he had no decision-making authority beyond choosing the route he could take from Indianapolis to Texas. Id.
On remand, the district court again denied the minor role reduction. In doing so, it said: “The task that the Court faces in trying to figure out the average member of the conspiracy is a rather daunting task, and so it seems logical to look at other people in the conspiracy who were charged with similar conduct as Mr. Saenz.” It then “narrow[ed] the scope of inquiry to the other eourier[s]” in the conspiracy, noted that “none of the other four couriers got any reduction for being a minor participant,” and concluded that Saenz was therefore also undeserving of this reduc*546tion. The district court did not conduct any analysis concerning the culpability of the average member of the conspiracy. It found that, without the minor role reduction, the advisory Guidelines range was 240 months to 293 months, and went on to consider “whether the additional 53 months” on top of the 240-month statutory minimum was warranted. It concluded that a sentence of 252 months was sufficient, as it was “one year more [than the statutory minimum] in light of the perjury that the defendant committed” at trial. Saenz appealed.
II. ANALYSIS
In Saenz I, we instructed the district court to compare Saenz’s culpability with that of the average member of the conspiracy, but nothing from the transcript suggests that the district court did so. See, e.g., United States v. Diaz-Rios, 706 F.3d 795, 799 (7th Cir.2013) (“[W]e cannot tell whether the district court compared Diaz-Rios’s role in the offense against those of average participants, as it should have.”). The district court instead looked simply at whether comparable couriers in the conspiracy also received minor role reductions. Though this approach is appealing (and it might be relevant to making the separate determination of whether the sentence should be adjusted so as to avoid unwarranted sentencing disparities under 18 U.S.C. § 3553(a)(6)), merely looking at whether other couriers in the conspiracy received a minor role reduction does not necessarily demonstrate how Saenz’s culpability is comparable to that of the average member of the conspiracy. Absent a finding as to whether the couriers themselves represent average members of the conspiracy, comparing Saenz to other couriers simply does not indicate how Saenz compares to the average member of the conspiracy. The other couriers may not have received a minor role reduction because they did not advance that argument, or perhaps certain evidence was not part of the record when the other couriers were sentenced. If there were any findings relevant to analyzing the culpability of the average member of the conspiracy in the sentencing proceedings of the other couriers, we do not know to what extent the district court relied on them. In sum, we have no direct or indirect indication that the district court compared Saenz’s culpability to that of the average member of the conspiracy, and so we conclude that the district court’s minor role determination was error.
Sentencing errors do not warrant remand if the error is harmless, but here we cannot say that the error is harmless. See United States v. Abbas, 560 F.3d 660, 667 (7th Cir.2009) (“To prove harmless error, the government must be able to show that the Guidelines error did not affect the district court’s selection of the sentence imposed.” (quotation marks and citation omitted)). The district court placed significant weight on the Guidelines range, specifically discussing what sentence within the 240-to-293 month range was appropriate (e.g., asking “whether the additional 53 months” on top of the 240-month statutory minimum was warranted). A minor role reduction, however, would have resulted in a flat Guidelines range of 240 months, see U.S.S.G. § 5Gl.l(b) (“Where a statutorily required minimum sentence [240 months here] is greater than the maximum of the applicable guideline range [135 to 168 months with the minor role reduction here, see Saenz I, 623 F.3d at 466], the statutorily required minimum sentence shall be the guideline sentence.”), putting the 252-month sentence above the applicable Guidelines range. Given the importance of the Guidelines range in the district court’s analysis, we cannot be convinced that “the sentence the judge imposefd] *547[would have been] identical.” United States v. Hill, 645 F.3d 900, 906 (7th Cir.2011) (quotation marks and citation omitted); see, e.g., United States v. Love, 680 F.3d 994, 998 (7th Cir.2012) (“Even though Love received a sentence that was significantly below the guidelines range, the range on which his sentence was based was erroneously calculated. Such an error is not harmless because it is impossible to know whether the district court would have imposed the same sentence had it not committed this procedural error.”). The government correctly observes that the district court attributed the 12-month addition to the 240-month minimum to Saenz’s perjury, but this finding was not made until after the court expressly satisfied itself that it was working within the Guidelines range. Had the minor role reduction been applied, the district court might not have found that the reprehensibility of the perjury was enough to justify going above the resulting flat Guidelines range of 240 months. Therefore we must vacate Saenz’s sentence and remand.
We note that the district court’s observation that “trying to figure out the average member of the conspiracy is a rather daunting task” was understandable, especially given the relative lack of guidance provided in Saenz I. So we attempt to provide additional guidance below as to how the minor role analysis might be done in this case, though we emphasize that the following is merely illustrative and should not be taken as a mandatory blueprint. Innumerable alternative approaches could also pass muster.
However the district court wishes to determine whether the minor role reduction applies, it must make some explicit or implicit finding concerning the culpability of the average member of the conspiracy. In doing so, the court might start by broadly summarizing the levels of involvement of the various groups of defendants in the conspiracy. See, e.g., United States v. McGee, 408 F.3d 966, 987 (7th Cir.2005) (conducting similar type of summary analysis); United States v. Hunte, 196 F.3d 687, 694 (7th Cir.1999) (same); see also United States v. Mendoza, 457 F.3d 726, 729-30 (7th Cir.2006) (“[A]n examination of each codefendant’s total role in the criminal offense provides a much more thorough insight into their responsibility as well as position in the conspiracy and, indeed, is required under the guidelines.”); Diaz-Rios, 706 F.3d at 798-99 (culpability may depend on factors such as the relationship with the leaders of the conspiracy, length and extent of involvement, potential financial gain, and knowledge of the conspiracy). The district court need not exhaustively detail the culpability of each and every co-defendant. See United States v. Stephenson, 53 F.3d 836, 850 (7th Cir.1995) (“The district court need not articulate its comparison [of culpability] on a person by person basis with each member of the conspiracy.”). For instance, four of the conspirators (Fierro-Mendez, Martinez, “Alex,” and “Lalo”) were based in Mexico and oversaw the entire distribution network, likely making them the most culpable in the conspiracy. Next most culpable might be Danny Delgado, who handled the United States point of contact in El Paso, Texas and ensured that cocaine would move onto Indianapolis, and Manual Mascorro, who oversaw the Indianapolis hub and was the point of contact for the regional distributors, which included Quinton Dews. Three other co-defendants, Pedro Mendoza, Carrera Camilo, and Ramiro Gonzalez, helped Mascorro offload cocaine shipments, weigh and package cocaine, and count drug proceeds, so they were likely less culpable than Mascorro. Last we have the couriers, whose culpability may vary depending on the nature of their *548roles. See, e.g., Mendoza, 457 F.3d at 729-30.
Next the district court should determine what might represent the culpability of the average member of the conspiracy and then compare it to Saenz’s culpability. See United States v. Gallardo, 497 F.3d 727, 741 (7th Cir.2007) (“David’s role should be compared to that of the average member of the conspiracy... .”); United States v. Leiskunas, 656 F.3d 732, 739 (7th Cir.2011) (“Rather than relying on a determination of whether Leiskunas’s act was necessary or repeated, the court should have evaluated Leiskunas’s role in context of the other participants in the scheme, keeping in mind that a minor player is substantially less culpable than the average participant, not the leaders.”). Mathematical precision is not required. For example, if the involvement of the couriers was particularly extensive and made their culpability comparable to the distributors’, then the couriers might indeed represent an average member of the conspiracy, and Saenz would likely not be entitled to a minor role reduction. See, e.g., McGee, 408 F.3d at 987 (“This was a wide-ranging drug conspiracy, and King, Keith McGee, and Smith all played roles that could be fairly described as average.”); United States v. Castillo, 148 F.3d 770, 776 (7th Cir.1998) (“In this five person conspiracy, three members, Castillo, Gonzales, and Manza-nares, had approximately the same level of involvement. The district court concluded that they were average members of this conspiracy.”). On the other hand, if most of the conspirators’ level of involvement substantially exceeded that of Saenz, then Saenz might qualify for a minor role reduction.
We repeat that the above examples are illustrative and not the only way the district court could conduct the analysis. For instance, it may not be necessary to specifically describe what the “average member of the conspiracy” looks like, if the district court’s comparison of the defendant’s culpability to those of the other members of the conspiracy leaves us confident that such a finding was at least implicitly made. See, e.g., Diaz-Rios, 706 F.3d at 799 (“The court should have looked at his role in the conspiracy as a whole.... ”); United States v. Osborne, 931 F.2d 1139, 1159 (7th Cir.1991) (concluding that defendant was not “substantially less culpable than the conspiracy’s remaining participants” without affirmatively stating what the average member of the conspiracy looked like). And we do not at all suggest that the district court’s analysis needs to be as lengthy as the above illustration.
We also note that none of the above discussion should be construed as factual findings that the district court must adopt, and in that sense we reject Saenz’s suggestion that Saenz I’s brief discussion about Saenz’s relative culpability is “law of the case” that is binding upon the district court. Much of the above is hypothetical and greatly simplified for purposes of illustration. It is the district court that is most intimately familiar with the facts and should, in the first instance, make the factual findings necessary to support a minor role determination. McGee, 408 F.3d at 987 (“[T]he role in the offense adjustment is a fact-intensive inquiry that the district court is best suited to address in the first instance, especially after becoming intimately acquainted with the roles of the members of a drug conspiracy....”).
III. CONCLUSION
For the above-stated reasons, we VACATE Saenz’s sentence and we REMAND *549for further proceedings consistent with this order.