In these two closely related class action suits under the Fair Debt Collection Practices Act, 15 U.S.C. §§ 1692-1692p, which we have consolidated for decision, the plaintiffs complain about dunning letters sent them by the well-known Plaza Associates debt-collection agency. In both cases the district court entered summary judgment in favor of Plaza after rejecting the survey evidence prepared by the plaintiffs’ expert witness, Howard L. Gordon.
Two identical letters sent to plaintiff DeKoven state that “we have been authorized to offer you the opportunity to settle this account with a lump sum payment for 65% of the above balance due, which is equal to $2,459.22. This offer will be valid for a period of thirty-five (35) days from the date of this letter.” The letter to Kubert is similar but includes a paragraph which states — after telling the recipient that if he notifies the agency within 30 days that he “dispute[s] the validity of this debt or any portion thereof’ the agency will “obtain verification of the debt or obtain a copy of a judgment and mail you a copy of such judgment or verification”— that “you may already have satisfactory *579proof that this account is listed with us in error. If so, please send this notice back along with a copy of one of the following to support your claim: Bankruptcy Notice from the court stating case number and filing date, Satisfaction of Judgment, Proof of prior settlement, Letter from the original Creditor clearing your account.” The suits complain about the statement in the letters that the offer of settlement is valid for only 35 days and the additional statement in the Kubert letter concerning “satisfactory proof’ that the account is in error.
The plaintiffs say the first statement would be understood by many consumers to mean that this would be their last chance to settle the claim and that the terms in it would be the best they’d be offered — that in short it was a final offer— when in fact Plaza Associates had been authorized by DeKoven’s and Kubert’s creditors to settle for less; thus the offers were just the opening bid in a negotiation. Kubert complains in addition that the reference to “satisfactory proof’ of error is misleading because it implies that to “dispute” a claim a debtor must furnish “proof’ to support his position.
As an original matter one might wonder why a debtor who does not deny the validity of his debt would be heard to complain that he had failed to understand that if he turned down the debt collector’s initial offer he might be able to settle the creditor’s valid claim for even less later. But in many cases, including the ones before us, the debtor is not simply a wise guy who could afford to pay his debts in full but would prefer not to. He is someone who cannot pay them in full because he has been hit by unforeseen medical bills or lost his job unexpectedly or is otherwise under water, without wanting to cheat anyone. He might be unable to pay 65 percent of a given debt but able to pay 25 or 33 percent, and if he did pay that lesser percentage he might be able to preserve, his credit standing and avoid bankruptcy. If through poor wording of the debt collector’s letter the debtor gets the impression that the initial offer is the final one, he may pay it in full and default on other debts, or decide that his position is hopeless and declare bankruptcy.
And so while a debt collector can, if authorized by the creditor whom he is representing, make his initial offer a final one, he cannot pretend that it is final if it is not, in the hope that the debtor will think it final. See 15 U.S.C. § 1692e(10); Campuzano-Burgos v. Midland Credit Management, Inc., 550 F.3d 294, 299 (3d Cir.2008); Goswami v. American Collections Enterprise, Inc., 377 F.3d 488, 495-96 (5th Cir.2004).
The problem with implementing this rule, as we explained in Evory v. RJM Acquisitions Funding L.L.C., 505 F.3d 769, 775-76 (7th Cir.2007), is that “the settlement process would disintegrate if the debt collector had to disclose the consequences of the consumer’s rejecting his initial offer. If he says ‘We’ll give you 50 percent if you pay us by May 14, but if you don’t, we’ll probably offer you the same or even better deal later, and if you refuse that, we’ll probably give up and you’ll never have to pay a cent of the debt you owe,’ there will be no point in making offers.” We added that this “concern can be adequately addressed yet the unsophisticated consumer still be protected against receiving a false impression of his options by the debt collector’s including with the offer the following language: We are not obligated to renew this offer.’ The word ‘obligated’ is strong and even the unsophisticated consumer will realize that there is a renewal possibility but that it is not assured.”
Plaza has not included this safe-harbor language in its dunning letters, but we had *580disclaimed in Evory any suggestion “that in the absence of safe-harbor language a debt collector is per se liable for violating [the Fair Debt Collection Practices Act] if he makes the kind of settlement offer that we quoted. We see a potential for deception of the unsophisticated in those offers but we have no way of determining whether a sufficiently large segment of the unsophisticated are likely to be deceived to enable us to conclude that the statute has been violated. For that, evidence is required, the most useful sort being the kind of consumer survey described in Johnson v. Revenue Management Corp., 169 F.3d 1057, 1060-61 (7th Cir.1999).” 505 F.3d at 776; see also Hahn v. Triumph Partnerships LLC, 557 F.3d 755, 757 (7th Cir. 2009); Williams v. OSI Educational Services, Inc., 505 F.3d 675, 678 (7th Cir. 2007). (But see, for criticism of the use of survey evidence, Judge Jolly’s dissenting opinion in Gonzalez v. Kay, 577 F.3d 600, 609-11 (5th Cir.2009).)
In the present cases the plaintiffs’ expert did conduct a survey. But both judges considered it inadmissible under the standards governing the admission of survey evidence (a form of expert evidence) in federal court. See, e.g., Muha v. Encore Receivable Management, Inc., 558 F.3d 623, 625-26 (7th Cir.2009); Peaceable Planet, Inc. v. Ty, Inc., 362 F.3d 986, 992 (7th Cir.2004); United States v. Curtin, 588 F.3d 993, 997-98 (9th Cir.2009); Vail Associates, Inc. v. Vend-Tel-Co., Ltd., 516 F.3d 853, 864 n. 8 (10th Cir.2008); see generally Fed.R.Evid. 702.
The survey staff interviewed 160 shoppers at a mall in a Chicago suburb. Half were shown the letter to Hubert; the other half — the members of the control group — were shown the letter minus the “valid for a period” and “satisfactory proof’ paragraphs. There was no need to show either group the letters to DeKoven, since they differed materially from the letter to Hubert only in lacking the reference to “satisfactory proof.”
After the survey respondents read the letter (either the survey letter or the control letter, depending on which group a respondent had been placed in), they were first asked questions about the letter orally, then given orally two answers to choose between, and finally handed a card with the answers printed on it and asked to pick one of them. The cards also contained a third answer option, which had not been presented orally: “DON’T KNOW/NOT SURE.” The critical question, asked of the respondents in both groups, was what the respondent thought would happen if he or she didn’t accept the offer in the letter — would it be renewed or extended, or was this the last chance to get a discount off the balance owed?
Of the respondents in the survey group, 59 percent thought the offer was final, 26 percent thought that it would be renewed or extended, and 15 percent didn’t know or weren’t sure. The corresponding percentages in the control group were 24 percent, 10 percent, and 66 percent. These statistics may seem strongly to support the hypothesis that the letter to Hubert was misleading. And likewise if we treat the “don’t know/not sure” respondents as having correctly interpreted the letter to leave uncertain whether the offer was final or would be renewed; for there was nothing misleading about the letter if interpreted so because, as we said, a debt collector is not required to reveal his negotiating strategy. Then 76 percent of the respondents in the control group (the 10 percent who thought the offer nonfinal and the 66 percent who didn’t know whether it would be renewed) correctly interpreted the letter they were given to read, compared to only 41 percent of the respondents who read the real letter to Hubert.
*581But as the district judges found, the members of the control group may well have been confused by the omission from the cropped letter of any reference to a deadline. How could an offer be extended if it had no deadline? The survey should have included in the letter shown the control group the safe-harbor language in Evory (“We are not obligated to renew this offer”) or some variant thereof. If the 66 percent of the respondents who answered “don’t know/not sure” are excluded, on the ground that they may well have been confused by the cropped letter, then the results of the survey do not support the plaintiffs’ claim. Of the 34 percent of the respondents in the control group who either thought the offer was final (24 percent) or thought it would be renewed (10 percent), 71 percent (24 percent divided by 34 percent) thought the offer was final. In the survey group, if the don’t know/not sure 15 percent is set aside, a slightly lower percentage of respondents — 69 percent (59 percent, the total who thought it final, divided by 85 percent [59 percent plus 26 percent], the sum of those who thought it final and those who thought it would be renewed) — thought the offer final. The implication is that the control letter was more confusing than the actual survey letter — that by adding the “valid for a period” paragraph that the plaintiffs claim confuses consumers, Plaza Associates reduced consumer confusion!
We decline to draw this inference, however, because the control letter was no good and may have confused respondents besides those who answered “don’t know/ not sure.” Therefore the survey was no good, as the judges found. It was no good for another reason: if the don’t know/not sure respondents are eliminated, the control group shrinks to 27 persons. Determining the minimum sample size from which reliable extrapolations can be made to the sampled population is tricky. Floyd J. Fowler, Jr., Survey Research Methods 45 (4th ed.2008). But 27 is too small a sample, especially when one considers the mismatch between the population to be sampled — people who receive dunning letters from debt collectors' — and the sample, which consisted of mall patrons none of whom, for all one knows, may ever have received such a letter. The sample drawn by the plaintiffs’ expert is what is called a “convenience” sample — convenient to the sampler — as distinct from a “representative” sample — representative of the population sampled.
The survey was bad for still another reason: the omission of the “don’t know/ not sure” option in the oral questioning. The omission was likely to make respondents guess — though the control letter was so confusing that most of the respondents in the control group may have grasped that option as a drowning man grasps at a straw. Cf. Sir Thomas More, A Dialogue of Comfort against Tribulation, ch. 3 (1534).
A properly designed control group is vital in a survey intended to reveal whether a debt collector is confusing debtors. Cf. Free v. Peters, 12 F.3d 700, 705-06 (7th Cir.1993); Penney v. Praxair, Inc., 116 F.3d 330, 333-34 (8th Cir.1997); United States v. Aguilar, 883 F.2d 662, 706-08 (9th Cir.1989). The debt collector can’t be blamed if consumers don’t understand his dunning letter unless he should have added or subtracted something to make it clearer. The plaintiff thus has “to show that the additional language of the letters unacceptably increases the level of confusion; many unsophisticated consumers would be confused even if the letters they received contained nothing more than a statement of the debt and the statutory notice.” Johnson v. Revenue Management Corp., supra, 169 F.3d at 1060 (emphasis in original).
*582The plaintiffs contend that Plaza Associates should have deleted the “valid for” paragraph. To prove this they conducted a survey. The survey, if it proves anything (we don’t think it proves anything), proves the opposite of their claim.
Kubert’s challenge to the “satisfactory proof’ paragraph also produced confusion in the control group. Of the respondents in the survey group, 66 percent opined that the recipient of the letter could not dispute the debt without proof and 26 percent that the recipient could; only 8 percent opted for don’t know/not sure. But in the control group, 20 percent said the recipient of the letter could not dispute the debt without proof, 24 percent that he could, and 56 percent didn’t know or weren’t sure. The confusion probably stemmed from the omission of any clue in the control letter to how one “disputes” a debt.
If we ignore don’t know/not sure, the survey results favor Kubert: 72 percent of the survey group (66 percent divided by 92 [66 plus 26] percent) thought the debtor could not dispute the debt without satisfactory proofs, compared to only 45 percent of the control group (20 percent divided by 44 percent). But the results are vitiated by the absence from the letter read by the control group of any reference to “proof.” The members of the control group were just asked whether the debtor could dispute the debt “without proof.” Since the letter they read did not contain the word “proof,” it was natural for them to assume that the thing the word denotes was not required.
The question asked the respondents in the survey group was also misleading. They were asked whether they could dispute the debt if they didn’t have “satisfactory proof.” To answer in the affirmative would imply that one can dispute a debt with unsatisfactory proof. What is unsatisfactory proof of not owing a debt? A respondent would naturally and we think correctly believe that if he does not have satisfactory proof of not owing the debt, the debt collector will not accept his denial of owing it.
Kubert argues that a consumer can dispute a debt for “no reason at all,” and that is true, DeSantis v. Computer Credit, Inc., 269 F.3d 159, 162 (2d Cir.2001) — provided one understands that “dispute” is a term of art in the Fair Debt Collection Practices Act. It means that the consumer can, without giving a reason, require that the debt collector verify the existence of the debt before making further efforts to collect it. 15 U.S.C. §§ 1692g(a)(4), (b); see Bartlett v. Heibl, 128 F.3d 497, 498-99 (7th Cir. 1997). This was not explained to the respondents. If after receiving verification of the debt, the consumer has no grounds for contesting it, the fact that he “disputed” it will not cancel his liability.
If Plaza Associates’ letter is misleading, it is misleading in containing an incomplete list of satisfactory proofs. But Kubert does not list additional proofs that he thinks a debt collector would be required to accept. The “satisfactory proof’ challenge fails also because of the inadequacy of the control letter and, as before, the omission of the “don’t know/not sure” option from the oral questioning of the respondents.
Suits under the Fair Debt Collection Practices Act have repeatedly come to grief because of flaws in the surveys conducted by the plaintiffs’ experts (often Mr. Gordon). Muha v. Encore Receivable Management, Inc., supra, 558 F.3d at 625-26; Jackson v. Midland Credit Management, Inc., 445 F.Supp.2d 1015, 1020-21 (N.D.Ill.2006), affirmed under the name Evory v. RJM Acquisitions Funding L.L.C., supra; Jackson v. National Action Financial Services, Inc., 441 F.Supp.2d *583877, 882 (N.D.Ill.2006); Hernandez v. Attention, LLC, 429 F.Supp.2d 912, 916-18 (N.D.Ill.2005). District judges may want to consider exercising the clearly authorized but rarely exercised option of appointing their own expert to conduct a survey in FDCPA cases. Fed.R.Evid. 706(a); General Electric Co. v. Joiner, 522 U.S. 136, 149-50, 118 S.Ct. 512, 139 L.Ed.2d 508 (1997) (Breyer, J., concurring); In re High Fructose Corn Syrup Antitrust Litigation, 295 F.3d 651, 665 (7th Cir.2002); Indianapolis Colts, Inc. v. Metropolitan Baltimore Football Club Ltd. Partnership, 34 F.3d 410, 414-15 (7th Cir.1994); Walker v. American Home Shield Long Term Disability Plan, 180 F.3d 1065, 1070-71 (9th Cir.1999); United States v. Microsoft Corp., 147 F.3d 935, 955 n. 22 (D.C.Cir. 1998). Judges can assure themselves of the expert’s neutrality by (as in arbitration) asking the parties’ own experts to nominate a third expert to be the court-appointed expert. In re High Fructose Com Syrup Antitrust Litigation, supra, 295 F.3d at 665; Indianapolis Colts, Inc. v. Metropolitan Baltimore Football Club Ltd. Partnership, supra, 34 F.3d at 414-15; Manual for Complex Litigation § 11.51, pp. 112-13 (4th ed.2004); Daniel L. Rubinfeld, “Econometrics in the Courtroom,” 85 Colum. L.Rev. 1048, 1095-97 (1985). A genuine neutral should be easy to find in the field of survey research because few survey researchers have settled views about debt collection.
The decision to appoint an expert is within the discretion of the trial judge, of course, and we merely invite consideration of the possibility of using this procedural device to improve judicial understanding of survey methodology. Although the judge is authorized to allocate the cost of the court-appointed expert between the parties, Fed.R.Evid. 706(b), we do not suggest that the defendant should be made to contribute to the cost of a survey conducted by the neutral expert, for in cases under the Fair Debt Collection Practices Act defendants rarely conduct their own surveys but are content to point out the deficiencies in plaintiffs’ surveys. A survey conducted by a neutral is a possible alternative to the often unedifying spectacle of a battle of party-appointed experts.
The dismissal of the suits is
Affirmed.