Johnson v. Midland Credit Management, Inc. et al - Document 188

Court Description:

Order For the reasons stated in the Order, Plaintiff's Second Renewed and Amended Class Certification Motion (ECF No. 155 ) is granted, and Defendant's Motion to Exclude the Opinions and Testimony of Ruoming Jin (ECF No. 159 ) is denied. As a result of the hearing held by the court, Plaintiff's Motion for Hearing on Defendants' Motion to Exclude the Opinions and Testimony of Ruoming Jin and Plaintiff's Motion for Conference/Hearing (ECF Nos. 164 , 167 ) are granted. Plaintiff's Motion for Extension of Time and Plaintiff's Motion for Appointment of Special Master are denied (ECF Nos. 172 , 177 ) as moot. Signed by Judge Solomon Oliver, Jr on 11/29/2012.(K,K)

Loading PDF...
UNITED STATES DISTRICT COURT NORTHERN DISTRICT OF OHIO EASTERN DIVISION ZLORO JOHNSON, et al., Plaintiffs v. MIDLAND CREDIT MANAGEMENT INCORPORATED, et al., Defendants ) ) ) ) ) ) ) ) ) ) Case No.: 1:05 CV 1094 JUDGE SOLOMON OLIVER, JR. ORDER On April 29, 2005, Plaintiff Zloro Johnson (“Plaintiff” or “Johnson”)1 filed the abovecaptioned case, on behalf of himself and all those similarly situated, against Defendants Midland Credit Management, Incorporated (“Midland” or “MCM”) and Encore Capital Group, Inc. (“Encore”) (together, “Defendants”). Plaintiff demonstrated, and the court found, that Defendants violated Johnson’s rights under the Fair Debt Collection Practices Act (“FDCPA”), 15 U.S.C. § 1692g, by failing to resend a required debt validation letter (“DVL”) to Johnson after the original letter was returned as undeliverable. (Summ. J. Order, ECF No. 81). Plaintiff, however, now seeks to certify the suit as a class action. This court has twice denied Plaintiff’s Motions for Class Certification (August 24, 2006 Order, ECF No. 81; June 24, 2008 Order, ECF No. 108.) In its most recent Order, the court denied Plaintiff’s Second Renewed and Amended Class Certification Motion because Plaintiff failed to demonstrate a feasible method to identify class members. (June 24, 2008 Order, ECF No. 108.) Following that ruling, the court allowed Plaintiff additional limited discovery 1 The court previously granted summary judgment for Defendants against a second named Plaintiff, Rod L. Feyedelem. (Order, ECF No. 81, 32-33.) in an effort to determine if MCM’s computers contained the information needed to identify class members. (July 22, 2008 Order, ECF No. 113.) Currently pending before the court are the following motions: Plaintiff’s Second Renewed and Amended Class Certification Motion (ECF No. 155), Plaintiff’s Motion for Hearing on Defendants' Motion to Exclude the Opinions and Testimony of Ruoming Jin (“Jin”) (ECF No. 164), Plaintiff’s Motion for Conference/Hearing (ECF No. 167), Defendants’ Motion to Exclude the Opinions and Testimony of Ruoming Jin (ECF No. 159), Plaintiff’s Motion for Extension of Time (ECF No. 172), and Plaintiff’s Motion for Appointment of Special Master (ECF No. 177). To resolve the complicated issues in these motions, the court held a hearing on September 17, 2012, in which both parties presented expert testimony on the feasibility of determining the class. For the reasons stated herein, Plaintiff’s Second Renewed and Amended Class Certification Motion is granted, and Defendant’s Motion to Exclude the Opinions and Testimony of Ruoming Jin is denied. As a result of the court’s hearing on these matters, Plaintiff’s Motion for Hearing on Defendants' Motion to Exclude the Opinions and Testimony of Ruoming Jin and Plaintiff’s Motion for Conference/Hearing are granted. Plaintiff’s Motion for Extension of Time and Plaintiff’s Motion for Appointment of Special Master are denied as moot. I. FACTS AND PROCEDURAL HISTORY2 A. Debt Validation Letter Requirement 2 A more detailed discussion of the facts and procedural history can be found in the court’s Order of August 24, 2006, ECF No. 81. -2- Midland is a consumer debt collection agency. It is undisputed that, with exceptions not relevant here, the FDCPA requires that the initial communication between a debt collection agency and a consumer be a debt validation letter (“DVL”) containing specific information. 15 U.S.C. § 1692g(a). Midland admits that, during the relevant time period, it employed a computer system that identified the following two distinct categories of mail as “undeliverable” and tracked them both as “RTM” (returned mail): (1) notices that were physically returned to Midland and never delivered; and (2) notices that were forwarded by the United States Postal Service (“USPS”). When debt validation letters were forwarded to new addresses, no FDCPA violation occurred. However, when debt validation letters were returned to Midland and it found a new address for the consumer, Midland did not always resend a debt validation letter to the new address. Instead, Midland sent a second type of notice. By communicating with consumers without sending the required debt validation notice, Midland violated the FDCPA. B. Plaintiff Granted Partial Summary Judgment By Order dated August 24, 2006 (ECF No.81), the court granted partial summary judgment for Plaintiff against Midland, finding that Midland violated the FDCPA, 15 U.S.C. § 1692g(a), by failing to send a debt validation letter to Johnson’s correct address. (Id. at 18-25.) In so doing, the court stated: The court [in Mahon v. Credit Bureau of Placer County, 171 F.3d 1197 (9th Cir. 1999),] noted that under the common law Mailbox Rule, “proper and timely mailing of a document raises a rebuttable presumption that it is received by the addressee.” Id. (internal quotations omitted). . . . [T]he [Mahon] court indicated the Mailbox Rule presumption was rebuttable . . . . In the instant case, Midland’s computer records indicate the validation notice it sent to Johnson was returned as undeliverable. However, based on the design of Midland’s computer system, this could mean the mail was actually physically -3- returned, or it could mean the mail was forwarded and the Post Office notified Midland of an address change. In this case, Johnson testified he had not lived at the Harrison Street address to which Midland sent the notice for more than five years. Therefore, according to the Post Office guidelines, the mail could not have been forwarded. While Midland has shown evidence that it did send a validation notice, Johnson has rebutted the presumption of delivery. Thus, the burden shifts to Midland to show that its notice to Johnson was not returned in the mail, since Midland receives returned mail from the Post Office in the regular course of business. The mere fact that Midland’s computer system was not designed to properly record the difference between returned mail and mail forwarding postcards is not sufficient to defeat summary judgment. Midland has put forth no evidence to suggest it actually sent Johnson a validation notice at a valid address. Midland’s violation in this case resulted when it obtained a new address for Johnson, and instead of sending out a new validation notice, it proceeded with its debt collection activities as if its first correspondence was sent to a correct address. Accordingly, since Midland failed to send Johnson a validation notice within five days of the first correspondence it actually sent to a valid address, the January 13, 2005 NY50 letter, Midland violated the FDCPA, 15 U.S.C. § 1692g(a). (Id. at 20, 23-24.) The court also found that Midland was not entitled to the bona fide error defense, which would have precluded FDCPA liability. By the same Order, the court denied summary judgment for each party regarding Defendant Encore’s liability as an indirect debt collector. (Id. at 25-32.) C. Plaintiff’s First Expert, Yuri Breitbart Plaintiff’s class allegations assert that Midland failed to send debt validation letters to all persons similarly situated to Johnson until in or about November 2006 when Midland updated its computer system to distinguish between the two types of undeliverable mail. In a previous Order, the court allowed Plaintiff to amend his Class A definition and denied his request to amend his Class B definition (Id. at 18-25), and denied Plaintiff’s Motion for Class Certification without prejudice -4- (Id. at 36). Plaintiff subsequently filed a Renewed and Amended Motion for Class Certification, which was denied. (ECF No. 108.) Defendants filed a Motion to Exclude the Opinions and Testimony of Yuri Breitbart, which was denied in part and granted in part. (Id.) The court found Breitbart’s algorithm to be unworkable because it relied on a non-existent forwarding fees database within the Midland database. Breitbart thought that these forwarding fees identified which addresses incurred forwarding fees and which did not, allowing him to identify the putative class. The court denied class certification at the class definition stage because the class was not properly defined, and therefore did not consider the other arguments advanced by the Defendants as to why Plaintiff’s expert’s testimony and report should be stricken regarding numerosity. D. Additional Discovery Following the denial of class certification, the court held a telephonic status conference, during which Plaintiff’s counsel indicated he wanted another opportunity to conduct discovery by having an expert directly examine the capability of Midland’s computer systems to determine whether the system could generate a list of persons who would be members of the putative class. After both parties submitted letters detailing their positions on this issue, the court entered an Order on July 22, 2008, allowing “Plaintiff a limited time to conduct additional discovery.” (Order, ECF No. 113.) This additional discovery was “limited to examining Midland’s computer databases to determine whether they contain a way to identify potential class members whose undeliverable debt validation letter was returned to Midland rather than forwarded by the United States Postal Service.” (Id.) Midland submitted a Motion for Reconsideration, wanting to prevent further discovery by Plaintiff, which was denied. (ECF No. 117.) Midland proceeded to file a Writ of Mandamus with the Sixth Circuit, requesting its intervention, which was also denied. (ECF No. 133.) Thereafter, -5- Plaintiff filed a Motion for the Appointment of an Independent Expert to Undertake the Inspection of Defendants’ Computer Data. (ECF No. 127.) The court denied this request and in the same Order, set forth a protocol, schedule for inspection, and instructed on the submission of expert reports. (Order, ECF No. 130.) This Order was later modified by the court several times pursuant to conferences conducted by Magistrate Perelman, the Special Master appointed by the court to handle the inspection. (ECF Nos. 138, 141, 143.) Plaintiff’s expert, Dr. Ruoming Jin (“Jin”), conducted his computer inspection over a threeday period, from September 30 to October 2, 2009. Defendants’ expert, R.E. Kurt Stirewalt (“Stirewalt”), also attended the inspection. On September 17, 2012, the court held a hearing to further clarify the experts’ positions on the feasibility of identifying the putative class members. E. Experts’ Findings The parties disagree about whether Defendants’ computer system contains an administratively feasible way to identify potential class members. Jin opines that he can “determine, absolutely that he can generate the information to print a list of Midland class members that meet the class definition.” (Pl.’s Second Renewed and Am. Mot. Class Certification at 9, ECF No. 155.) Plaintiff contends that Midland’s database contains sufficient information to identify class members. Defendants disagree and seek to strike Jin’s opinions and testimony, primarily on the ground that his methodology is unreliable because his theory is based on unreliable data and therefore not valid. 1. Jin’s Findings In his report, Jin concludes: -6- to a high degree of professional certainty that [the] MCM database system contains sufficient information for identifying absent class members, such as Johnson. Specifically, there are database tables and fields maintained in MCM database system which can directly resolve subproblem (a): identifying consumers whose undeliverable debt validation letter was physically returned to MCM (rather than forwarded by USPS). There are also adequate database tables and data fields to demonstrate that for those for whom MCM has obtained a new address, and subsequently sent a second letter demanding the payment of alleged debts without information required by the FDCPA (subproblem (b)). In addition, there is also adequate information to eliminate certain unique cases, such as death or bankruptcy of the class member. To sum, it is my expert opinion that MCM database system will generate the names and addresses of absent class members. (Jin Report at 2, ECF No. 145.) During the inspection, Jin looked at “several key source code files which handle the returned mail (RTM), address change services (ACS) from USPS, and the sending of more than one DVL [debt validation letters].” (Id. at 7.) Jin also inspected “several database tables which contain the key information for class member identification.” (Id. at 8.) He looked at the Return Mail History File, which records whether the mail was forwarded, and the Letter History File, which records every piece of mail which was sent by MCM. (Id. at 8.) He also looked at the Commentary File, which contains a field which states whether the returned mail had a new address (RTM1) or had no address (RTM2).3 (Id. at 9.) Jin’s final step was to develop and refine 3 MCM receives information regarding new addresses several ways. If the forwarding service by the USPS has stopped, but a new address is known, the original letter will be returned to MCM with a yellow sticker containing the new address. (Jin Report at 10-11, ECF No. 145.) If the post office has forwarded the mail, MCM is notified through a postcard from the USPS with the new address, photocopy of the envelope, or through electronic notice via CD or Tape. (Id.; Stirewalt Report at 7, ECF No. 148.) If the mail has been forwarded, the vast number of those notifications appear on the CD or Tape. Jin found the number of postcards was very small (2,525) as compared to the number of returned physical mail (3,521,038). (Jin Report at 11, ECF No. 145.) In addition, Jin states that his searches found that the vast majority (98%) of the returned mail did not come in -7- a list of search queries in order to identify the class and “validated these queries through the inspection of individual consumer accounts.” (Id. at 8.) From examining Midland’s database, Jin concludes that the Commentary File and the Return Mail History File can enable him to determine which returned mail pieces had new addresses, and if the mail was forwarded. (Id. at 8-12.) This information would allow him to identify people whose undeliverable debt validation letter (“DVL”) was physically returned to Midland, rather than forwarded by the USPS. (Id.) Jin opines that he can find those people to whom Midland sent at least one other letter besides the DVL, since the second letter would have been sent to another address, different from the DVL, and this letter is not recorded as returned within Midland’s database. (Id. at 14.) Jin contends that he can use the letter history database, which is used by Midland to record every piece of mail it sends to alleged debtors, to see if a non-DVL was sent to the alleged debtor and if its address is different from the one used in the DVL. (Id.) Additionally, Jin states that the letter history file was annotated with an “R” or “U” (returned or undeliverable) when mail was returned, and if this field does not contain an “R” or a “U,” then the mail was successfully delivered by the USPS. Therefore, he concludes, the potential class members are identified from this second step, following the determination as to whether a non-DVL was sent to the alleged debtor. (Id.) Based on his inspection, Jin contends that he has identified a total of 568,076 potential class members. (Id.) He states that this list can be further refined by eliminating “special cases, like death or bankruptcy, for satisfying a more refined legal criteria of class member identification.” (Id.) During the inspection, Jin contends he sampled a list of consumer accounts the form of a postcard, or contain a forwarding address. (Id. at 12). -8- for validation purposes. He provides four of these accounts for illustration. Jin points out that one of the four cases, is one of the “special” cases that does not belong in the class because a bankruptcy was filed, but states that those cases can be “easily recognize[d]” by browsing the consumer accounts information, which is the “information necessary to identify those special cases.” (Id. at 14-15.) Such “special” cases have their own codes, and thus “the information for determining those special cases is available and...can be utilized in the class member refinement.” (Id. at 18.). 2. Stirewalt’s Findings Defendants’ expert, Stirewalt, firmly disagrees with Jin’s conclusions. Stirewalt was previously unable to conclude whether the MCM database could identify individuals who are similarly situated to Plaintiff. However, having reviewed all of the information currently available, he concludes with his “strongest of professional conviction that the MCM database CANNOT be used to reliably identify those debt validation letters that were returned to Midland rather than forwarded by the USPS and that, consequently, the database cannot be used to identify those individuals similarly situated to the Plaintiff.” (Stirewalt Rep. at 2, ECF No. 148.) Stirewalt takes issue with several of Jin’s conclusions, including his “implicit assumption” that the codes in the MCM database “reliably encode information about the presence or absence of forwarding address on mail pieces.” (Id.) Stirewalt opines that the correct conclusion to be reached from the codes regarding the returned mail and new addresses should include the qualifier that “if the human operator enters a new address,” then the database encodes the particular code relating to whether a new address has been provided. (Id. at 2-3.) Stirewalt contends that Jin’s interpretation of the codes leaves no room for any operator error, “which is known to be a major problem with manual mail handling processes.” (Id. at 3.) -9- Stirewalt also finds there is “no source of reliable data that tracks the delivery status of physical mail sent out by MCM.” (Id. at 6.) If this data was located and was reliable, Stirewalt contends that this would have been akin to a “smoking gun,” demonstrating that the MCM database can distinguish returned from forwarded letters. (Id.) Since there is not a direct way to make this identification, an inference is necessary, which Jin uses. Stirewalt contends that there are three criteria that must be met in order to properly validate an inference: “(1) establishing the reliability of the subject data upon which the inference is based, (2) establishing the internal consistency, or logical soundness of the inference, and (3) testing the external consistency of the inference, i.e. empirically checking the consistency of an inference’s predictions against some objective set of expected results.” (Id.) Stirewalt states there is no way to meet the third criteria, since there is no objective means for judging whether a letter was actually returned to MCM, or forwarded to its intended recipient. (Id.) Any returned letters, photocopies of forwarded letters, or postcards from the USPS with the new address, have long since been discarded. (Id.) Therefore, the inference must be evaluated according to the first two criteria. Stirewalt states that a “[d]eficiency in either of these two criteria completely undermines any scientific basis for accepting the inference as a reliable predictor of the data it purports to infer.” (Id. at 7.) He contends that Jin’s inference fails the first criteria because the codes in the commentary and databases are “unlikely to accurately predict the delivery status of actual, physical letters.” Jin asserts that any new address entered in the system will classify the DVL as forwarded. (Jin Report at 10-11, ECF No. 145.) However, some of the new addresses provided to MCM are from the actual returned letters, which have a yellow sticker with the new address on it, and have not been forwarded. (Id.) Therefore, the codes Jin relies on will predict the letter was forwarded when it was actually returned as undeliverable, and will - 10 - underestimate the number of returned letters. (Stirewalt Report at 7, ECF No. 148.) Additionally, Stirewalt contends this inference is likely to overestimate the number of returned letters, since some letters that were actually forwarded will be classified as undeliverable, due to operator error or bugs in the software process. (Id.) Stirewalt notes that the “effects of bugs in the software on the integrity of the subject data codes are unknown and were not explored during inspection.” (Id. at n.8.) Stirewalt concludes that although the relationship Jin infers may be an identification relationship, it is not the one he is claiming, since he reverses the mandatory role in the relationship. (Id. at 8.) This means that “the relationship defined using his inference does not use information in the Midland database to identify letters returned to Midland as undeliverable. Rather, it uses letters returned to Midland as undeliverable to identify information in the Midland database.” (Id.) Therefore, Jin uses the absence of information to infer a result, and not the presence of particular data to make an inference. Stirewalt also takes issue with how the inspection was conducted, noting that Jin spent a majority of the time “writing and debugging queries and comparably very little time testing and looking at actual data.” (Id.) He contends that over the course of 3 days, they looked at fewer than 10 customer accounts, most of which turned out to be the “exception cases” Jin references. (Id.) Though Jin asserts that an individual verification could easily rule out the exception cases, Stirewalt states that he finds it troubling that “the small number of data points that he considered yielded such a high percentage of exceptions.” (Id.) Stirewalt maintains that these exceptions completely undermine any strength of Jin’s conviction that potentially 568,076 class members exist. (Id.) Stirewalt concludes that this number has to be both under and over inclusive. (Id.) Jin does not know how many people within his set of potential class members should not be a part of the - 11 - proposed class, what the characteristics are of these misidentified accounts, nor whether the MCM database contains the information needed to refine his queries. (Stirewalt Supplemental Report at 3, ECF No. 153.) Based on Jin’s deposition testimony, Stirewalt understands Jin to be claiming that the legal experts would determine what additional criteria could be used to refine the queries after examining each of the accounts, and that the database must contain the details a legal expert would need to determine if a member is a part of a proposed class. (Id. at 4.) However, Stirewalt states that these legal experts would have no way of knowing that a forwarded DVL was mistakenly classified as undeliverable. (Id.) Stirewalt interprets Jin’s conclusion that the database must contain the necessary information as a “statement of faith,” rather than a reasoned conclusion. (Id.) In addition, Stirewalt criticizes Jin’s deposition testimony that indicates that he did a random sampling of the outputs of his queries in order to affirm his results. (Id. at 5; Jin Dep. at 107-08, ECF No. 157-1.) Stirewalt contends that Jin has to know that he did not conduct a proper random sampling. (Stirewalt Supplemental Report at 5, ECF No. 153.) Stirewalt states that “[r]andom sampling does not mean just picking a few points from a very large population and then assuming any property that holds over the small sample will hold over the population.” (Id.) Stirewalt explains that how accurately a statistic represents an attribute of a population depends on the size of the population, the size of the random sample, and the variance of the attribute in the population, with small sample sizes and high variability adversely impacting the degree to which the statistic accurately represents the attribute of interest in the population. (Id.) Lastly, Stirewalt takes issue with the accounts used by Jin for illustrative purposes in his expert report. (Id. at 4.) All of the accounts predate Plaintiff’s class period by several years. Therefore, not even Jin’s examples are a part of the putative class. - 12 - 3. Plaintiff’s Rebuttal to Defendants’ Conclusions Plaintiff argues that Stirewalt’s opinion that the Midland database cannot “reliably identify” the putative class is his only contention with Jin’s report, and that he largely agrees with Jin otherwise. (Reply at 13, ECF No. 158.) Plaintiff contends that Stirewalt’s emphasis on reliability is misplaced because the error affecting the reliability of Jin’s algorithm is human error, which is present in any recording system in which persons are involved. (Id. at 15). Plaintiff then notes that Stirewalt agreed, in his deposition, that without human error, Jin’s inference would be true. (Id.) Plaintiff next contends that Stirewalt’s definition of “identify” is also misplaced, because it assumes that 100% accuracy is needed to properly define the class. (Id. at 16) Plaintiff notes that Stirewalt would, if tasked with finding as many class members as he could, have ended up with something very similar to Jin’s algorithm. (Id.) Jin states that it was not his responsibility to define or to develop a complete or final product of the algorithm. However, from his investigation and development of his report, he states that he has a “rather complete algorithm.” (Jin Dep. at 6, ECF No. 157-1.) Jin states that his algorithm would still be good as it stands, even if there were 10,000, 50,000, or 500,000 “exception cases” as he calls them, since he did not believe that there was a point at which the number of exceptions would make the algorithm unreliable. (Id. at 65-67.) However, he does state that he does not believe there would be that many exceptions. (Id.) Plaintiff contends that these exception cases are people who meet the class definition, but simply for some reason Midland would not have liability for, such as for the reasons indicated by Jin: death or bankruptcy. (Resp. at 16, ECF No. 161.) Plaintiff also asserts that the exceptions are cases where the second letters are not attempts to collect debts, which would mean they would need - 13 - to be excluded from the potential class Jin generated, but that Midland has all the letter history information to exclude the individuals. (Reply at 18, ECF No. 158.) In addition to Defendants’ challenges regarding exception cases, Plaintiff also disputes the characterization by Defendants that Jin thought that all of the accounts identified by his algorithm must be individually verified. (Resp. at 17, ECF No. 161.) However, Defendants argue that it was Jin who brought up these “special cases” that would need to be reconciled before he could fully identify class members, and that this list was non-exhaustive, and would therefore require individualized inquiry. (Jin Report at 23-24, ECF No. 145.) Plaintiff disputes Defendants’ contention that based on Jin’s admissions, and Stirewalt’s conclusions, Jin’s methodology for identifying class members results in both an over and under inclusive result. Plaintiff argues that this “purported dilemma is small in scope,” like any other class action where files are reviewed to determine which individuals meet the class definition. (Pl.’s Reply at 17, ECF No. 158.) Plaintiff also maintains that because “Midland was deficient by not specifically tracking the returned mail, it is possible to find some, but not all of the class members.” (Id.) Plaintiff concedes that Defendants are partially correct in regard to their claim that all of the accounts provided for illustrative purposes predate the class period, noting that “Dr. Jin did not specify that the results generated from the search queries be limited to the specific class period stated.” (Resp. at 8.) Plaintiff contends this could be corrected though, since as Jin indicated, there is “enough data in the Midland database that this list can be ‘filtered’ to limit it to a certain time period, or to list only those accounts that were sent a specific form letter.” (Reply at 20, ECF No. 158.) Plaintiff indicates that some of the information needed for the accounts was not available to - 14 - Jin. (Id. at 22.) Midland routinely purges this information and it can only be recovered from backup tapes. (Id.) Plaintiff states that “[a]dditional testing or refinement of the search queries and the recovery of purged data from backup tapes was contemplated by Dr. Jin.” (Id.) Plaintiff further states that Jin agrees the exact queries run during the inspection are not the final product. (Id. at 23.) However, Plaintiff maintains that Jin has identified the information in the database which would allow him to create the final algorithm. 4. The Court’s Hearing on Class Identification On June 19, 2012, this court determined that a hearing would be needed to further clarify some of the issues presented by parties’ briefs on class identification. The hearing was held on September 17, 2012, and both parties’ experts presented testimony on the examination of MCM’s databases and Jin’s algorithm. At the hearing, Jin explained the methods detailed in his report, and further clarified that MCM kept a variety of different codes for DVLs that could be used to narrow the class members down to only those who had received a debt collection letter without first receiving a DVL. He confirmed that his method, which uses the RTM1 and RTM2 codes to identify which DVLs were returned to Midland, could be used to narrow which potential class members had subsequently been sent a non-DVL letter in violation of the FDCPA. He confirmed that while his task was only to determine whether such an identification procedure was possible, he was certain his methodology could identify the class members as defined by Plaintiff. Jin testified that his method could be adjusted to account for a narrower time period, different DVL codes created by Midland, letters not violating the FDCPA, and exception cases like death and bankruptcy. He based this opinion on the fact that Midland had codes for all of these potential exceptions to the class membership in their - 15 - databases or on tapes. Jin also confirmed that the vast majority (98%) of undeliverable letters were coded as RTM2, that is, they posed no identification problems because they were not accompanied by a new address. Jin further confirmed that he had based all of his prognosis about his methods on what he had discovered in the databases. Finally, Jin estimated that it would take a few days to perfect and calibrate the algorithm to extract the putative class from Midland’s databases. This would include time to extract information contained on Midland’s tapes, as well as testing the algorithm and debugging it. Stirewalt testified that Jin’s algorithm had not yet been properly tested. He pointed to the fact that the algorithm currently did not account for all the exception cases mentioned above. He also testified that it had not been properly debugged. Stirewalt estimated that to properly test and debug the algorithm would take about a month. Stirewalt then pointed to two main sources of error which could undermine the class definition. First, he reiterated that there was no way to account for mail handler error. However, he also acknowledged that if there was no operator error, Jin’s inference would be true. Second, he pointed to the fact that both the postcards, as well as the returned mail with a forwarding address, could not be differentiated using Jin’s algorithm. Stirewalt acknowledged that the number of returned mail without any forwarding addresses was overwhelmingly larger than any of the returned mail with forwarding addresses and postcards. The court then asked the parties to address the issues related to the inability to differentiate within the database between those who had not received a DVL (i.e., where the DVL was returned to Midland with a forwarding address) and those who had been forwarded the DVL (i.e., where Midland only received a postcard, or the information was contained within the CD/Tape given to them by the USPS). Plaintiff acknowledged that it would be impossible to differentiate between - 16 - the two due to Midland’s record keeping procedures, but argued that this represented such a small number of potential class members that it should have no effect on class certification. Defendant agreed that due to this problem, the entire class could not be identified. The court also asked the parties if there was any value or figure that could be used to calibrate the algorithm to account for the operator errors. Stirewalt stated that he knew of a study saying that where the error rate for enterprise databases is unknown, an error rate of 5% could be expected. Jin stated that he thought a 5% error rate would be too high in this case because of the simplicity of the mail sorting process at Midland. Jin felt an error rate of 1% would be more realistic in this case. Jin also stated that a 5% error rate would not change his opinion about the reliability of his algorithm. II. LAW AND ANALYSIS A. Legal Standards for Class Certification and Class Definition Plaintiff seeks certification of the proposed class under Federal Rules of Civil Procedure 23(a) and 23(b)(3). A court must engage in a “rigorous analysis” of the plaintiff’s ability to meet the requirements of Federal Rule of Civil Procedure 23(a) before certifying a class. Gen. Tel. Co. v. Falcon, 457 U.S. 147, 161 (1982). To obtain class certification, Plaintiff bears the burden of satisfying the requirements of Rule 23(a), commonly known as numerosity, commonality, typicality, and adequacy of representation, and must demonstrate that the class fits under one of the three subdivisions of Rule 23(b). Coleman v. GMAC, 296 F.3d 443, 446 (6th Cir. 2002). A district court has broad discretion in determining whether to certify a class, within the dictates of Rule 23. Gulf - 17 - Oil Co. v. Bernard, 452 U.S. 89, 100 (1981); In re Am. Med. Sys., 75 F.3d 1069, 1079 (6th Cir. 1996). Plaintiff seeks to certify the following class: All persons and entities who were sent a notice from defendants, between March 18, 2004, and November 30, 2006, alleging a debt owed to a third party and seeking to collect that debt which notice advised them of their right to dispute the validity of the debt within 30 days and/or request in writing that Defendants obtain verification of the validity of the debt before continuing collection activities, which same notice was returned in the mail to Defendants, which afterwards the same persons and entities were subsequently mailed another notice seeking to collect that debt and which subsequent notice did not advise them of their right to dispute the validity of the debt within 30 days and/or request that Defendants obtain verification of the validity of the debt, which subsequent notice was not returned in the mail to Defendants. Excluded from this Class are employees, officers, directors, legal representatives, heirs, successors, and assignees of Defendants. (Pl.’s Second Renewed and Am. Mot. Class Certification at 7, ECF No. 155.) Defendants contend that the threshold issue in light of the limited discovery previously granted, is whether the computer databases contain a way to identify potential class members whose undeliverable DVL was returned to Midland, rather than forwarded by the USPS. Defendants maintain that whether Jin met this burden must be decided before consideration of Plaintiff’s Second Renewed and Amended Class Certification Motion. Although the Sixth Circuit has not expressly addressed the extent to which the individual class members must be ascertainable prior to class certification, many other courts have this issue. See, e.g., Crosby v. Soc. Sec. Admin., 796 F.2d 576, 580 (1st Cir. 1986); Bentley v. Honeywell Int’l, Inc., 223 F.R.D. 471, 477 (S.D. Ohio 2004); Garrish v. United Auto., 149 F. Supp. 2d 326 (E.D. - 18 - Mich. 2001); see 7A Wright & Miller, Federal Practice and Procedure § 1760 at 134 (collecting cases).4 As Wright and Miller explained, Although not specifically mentioned in the rule, an essential prerequisite of an action under Rule 23 is that there must be a “class.”. . . . In keeping with the liberal construction to be given the rule, it has been held that the class does not have to be so ascertainable that every potential member can be identified at the commencement of the action. . . . If the general outlines of the membership of the class are determinable at the outset of the litigation, a class will be deemed to exist. Nor is the fact that specific members may be added or dropped during the course of the action important. However, the requirement that there be a class will not be deemed satisfied unless the class description is sufficiently definite so that it is administratively feasible for the court to determine whether a particular individual is a member. . . . Further, the class must not be defined so broadly that it encompasses individuals who have little connection with the claim being litigated; rather, it must be restricted to individuals who are raising the same claims or defenses as the representative. The class definition also cannot be too amorphous. Id. at 134-47 (footnotes omitted) (emphasis added). Thus, the court finds that proper definition of an ascertainable class is a prerequisite to class certification. B. Ability to Identify the Class Members At the outset, the court notes that, on class certification, only the ability to identify class members is necessary; the actual names and addresses of class members are not necessary at this time. “A class is properly identified so long as it is defined by objective criteria.” Saltzman v. Pella Corp., 257 F.R.D. 471, 475 (N.D. Ill. 2009). “This criteria must make it administratively feasible for the court to determine whether a particular individual is a class member.” Id. Defendants argue 4 The parties do not dispute that Plaintiff is a member of the proposed class. - 19 - that the proposed class is not sufficiently ascertainable because, due to the method by which Midland maintained addresses and mail attempts, Midland’s computer system does not contain a way to identify class members. Consequently, determining the eligibility of each potential class member would require the court to address the central issue of liability and to conduct individual factual inquiries at the class certification stage. Defendants further argue that Jin’s method for identifying potential class members cannot exclude those persons whose DVLs were forwarded to them by the USPS as they would be classified under the code RTM1. Defendants also argue that the class is overinclusive because of those “exceptions” to the class that may appear in Jin’s current algorithm, such as debtors who have died or become bankrupt. Plaintiff argues that the class definition does not require the court to “undertake a factintensive inquiry to determine whether each potential class member had rebutted the presumption of delivery of these validation letters by Midland. The evidence of non-delivery is already documented in the Midland computer.” (Pl.’s Second Renewed and Am. Mot. Class Certification at 24, ECF No. 155.) Plaintiff contends that the computer inspection demonstrates that the information needed to rebut this presumption is available in Midland’s databases. (Pl.’s Reply at 27, ECF No. 158.) In addition, Plaintiff maintains that he does not need to prove that the second letter, the non-DVL was received by the class member. The Mailbox Rule states that “proper and timely mailing of a document raises a rebuttable presumption that it is received by the addressee.” (Id.) (citing Mahon v. Credit Bureau of Placer Cnty., 171 F.3d 1197 (9th Cir. 1999)). Plaintiff contends that if Midland has not marked a letter as being returned or having a new address, each class member presumably received the letter. (Id.) Therefore, Plaintiff concludes an individualized inquiry is unnecessary. - 20 - The court finds Plaintiff’s arguments to be well-taken. First, the court notes that the Mailbox Rule would not pose a concern here because the algorithm could account for those DVLs that were sent on or about the same time as other communications. Though Jin’s algorithm is not complete at this point, he has persuasively testified that Midland’s record keeping of letters sent would allow him to exclude those persons who had their first DVL returned but subsequently received a DVL in the statutorily prescribed time frame. Jin explains that Midland’s letter-coding system would be available to narrow the category of undeliverable mail returned to Midland without a forwarding address. Those persons who had received a DVL by the statutorily prescribed time could be identified using Midland’s codes for its various DVLs (e.g. LT1A). Stirewalt does not dispute this. Therefore, the court finds that Defendant’s concern would not require an individual inquiry as Jin’s algorithm could exclude from the class those persons who had still received a DVL within the statutorily prescribed period. Second, the class as proposed by Plaintiff is limited to those persons whose DVLs were returned to Midland, and who subsequently received non-DVL letters in violation of the FDCPA. Defendants argue that Jin’s methodology is questionable, citing as examples his lack of consideration of exceptions and his small sampling size. Jin maintains that with further development, his algorithm could take account of all the exceptions named by defendants. Debtors who were dead or bankrupt at the time the letters were mailed have their own code in the Midland databases, and could be excluded using those codes. Similarly, those potential class members who were in fact sent DVL to the correct address could be excluded by using Midland’s own letter codes, which distinguish DVL from non-DVL communications. The court finds Jin’s explanations persuasive, and that he could further develop his algorithm to ensure these “exception” cases could - 21 - be eliminated from the putative class. While Jin did state that his algorithm could yield around 500,000 exceptions and still be correct, this was a theoretical argument. There is no reason to believe that the actual class would contain anything close to this number of exceptions. Finally, the court finds that the potential errors pointed to by Defendants are not enough to disqualify Jin’s algorithm as a valid method of determining the members of the class. Defendants’ expert first argues that the algorithm cannot properly identify all members of the putative class because of potential operator error during the scanning of the returned mail. However, the court does not find this argument to be well-taken. The class does not need to be perfectly identified, otherwise no class which involved potential human error could ever be created. Further, as Defendant’s expert stated at the hearing on these motions that an industry standard might is typically 5% where no rate of error has been measured. The court finds that this rate of error is not enough to disqualify the entire class. Further, as Jin noted during the same hearing, the operator’s task was so simple that the rate of error would likely have been closer to 1%. The only error that could make the class overinclusive would be where the operator incorrectly marked a forwarded letter as returned. As Jin details in his report, not only is this mistake very unlikely to occur due to the mail sorting process at Midland, but the number of forwarded letters (represented by the postcards) that could be misclassified is extremely small in comparison with the total number of returned DVLs. Most forwarded mail would be present on the CDs/Tapes provided by the USPS, allowing Jin’s algorithm to exclude such forwarded mail. Therefore, the court finds that the class could still be definite enough, even accounting for human error, and that plaintiff has shown an administratively feasible way to identify putative class members. - 22 - As for the postcards, which remain the most difficult point of contention between the parties, the court agrees with the Plaintiffs that they represent such a small proportion of Jin’s identification method that this should not disqualify the entire class of persons. Plaintiffs concede that there is currently no way to separate the postcards from the returned mail with yellow stickers, as those were both coded RTM1. Plaintiffs point out that the number of DVLs that yielded postcards comprises only a tiny fraction of all of the RTM1 codes. Defendants argue that this would still make the class overinclusive, and therefore should make Plaintiff’s class definition invalid. The court does not find Defendant’s argument to be well-taken. A class does not need to be defined to such a precise degree that any error in definition would make the class invalid. Further, the court notes that Plaintiffs may choose to use Jin’s algorithm to find class members based solely on the RTM2 codes (the DVLs returned without any forwarding address), while finding the rest of the putative class members coded as RTM1 via notice. See Saltzman at 476 (citing In re Warfarin Sodium Antitrust Litig., 391 F.3d 516, 536-37 (3d Cir. 2004) (Where class members were consumers of a prescription drug, and names and addresses of these consumers were confidential and not available to parties, notice by publication combined with call center and website was sufficient notice to identify class members; Macarz v. Transworld Sys., Inc., 201 F.R.D. 54, 59 (D.Conn. 2001) (notice by publication used where circumstances “make it impracticable to gain the names and addresses of class members and notify them individually of the action’s pendency”); Mirfaishi v. Fleet Mortg. Corp., 356 F.3d 781, 786 (7th Cir. 2004) (Internet notice of settlement was acceptable substitute for individual notice where Defendant had no record of part of a class of customers)). In either case, the court finds that Plaintiff has shown - 23 - his ability to define the class, and that the additional work required to perfect Jin’s algorithm should not preclude class definition. C. Rule 23 Requirements 1. Rule 23(a) Requirements Plaintiff argues that “once identification of the class is shown to be feasible, the Rule 23 requirements for class certification are easily met.” (Pl.’s Second Renewed and Am. Mot. Class Certification at 22, ECF No. 155.) Plaintiffs state that the numerosity requirement is met because Jin’s report has identified over 500,000 potential class members. (Id. at 24). Further, because the class definition is based on initial DVLs returned to Midland, there is “no need...for this Court to undertake a fact-intensive inquiry to determine whether each potential class member has rebutted the presumption of delivery” of the DVLs. (Id.) Plaintiffs argue that this also applies to the commonality requirement, because “the claims of all potential class members are based upon the identical conduct of Defendants.” (Id.) Plaintiffs further states that the typicality requirement is met because all of the putative class members’ claims arise from Defendants’ failure to send a DVL prior to attempting to collect a debt. (Id. at 27). Finally, Plaintiffs argue that Johnson is a more than adequate representative for the class as he has no adverse interests to the class and is willing to appear at depositions and assist counsel in the prosecution of the action. Defendants contend that the Mailbox Rule would require several mini-trials to allow Midland to rebut the presumption of the receipt of a non-DVL. (Def. Mot. in Opp. at 30, ECF No. 157.) Defendants argue that as a result, Plaintiff cannot meet its burden as to the commonality requirement of Rule 23(a). Plaintiffs contend that Midland cannot rebut the presumption of - 24 - delivery of non-DVLs, and has not offered any method by which they might be able to do so. (Resp. at 28, ECF No. 158) The court finds Plaintiff’s arguments to be well-taken. The court notes that the legal question in this case, whether class members received DVLs before receiving attempts to collect debt in violation of the FDCPA, is a simple one. Defendants’ argument that the common law Mailbox Rule would require individual inquiry is not well-taken because Plaintiffs do not dispute presumption of delivery of non-DVL letters. In addition, Defendants have not shown how they could rebut the presumption of delivery of their own letters. This would in effect amount to having to prove a negative. Therefore, the court finds that Plaintiffs have shown that the commonality, numerosity, and typicality requirements have been met as they all involve the relatively simple question of whether class members received an attempt at debt collection before receiving a DVL. The court also finds that Plaintiff Johnson is an adequate representative for the class, as the facts in his case mirror the class definition. 2. Rule 23(b)(3) Requirements Plaintiffs argue that the requirements of Rule 23(b)(3) are met because the common questions of law and fact will predominate over any questions affecting only individual class members, and because the class action is a superior vehicle in this case. (Pl.’s Second Renewed and Am. Mot. Class Certification at 29, ECF No. 155.) Plaintiffs state that the issue of liability in this case is narrow, and limited only to whether Midland violated its obligation to issue a DVL before attempting to collect a debt. (Id. At 30). Plaintiffs then argue that the class action is the superior vehicle in this case, as the amounts in controversy for each plaintiff would be small, and many plaintiffs would be unaware of their rights under the FDCPA. (Id. at 32). Defendants again argue - 25 - that there are individual factual or legal issues that pertain to each putative class member’s receipt of a second validation letter. (Def. Mot. in Opp. at 32, ECF No. 157.) Defendants argue that Jin’s algorithm will have errors that will force the court to have to make individual determinations as to the putative class members’ claims, making the class action an inferior vehicle. (Id. at 34). The court finds Plaintiff’s arguments to be well-taken. The court has already determined that Plaintiff has met his burden of showing that Jin’s algorithm can properly identify the class. The issue of liability therefore remains quite simple, and does not demand individual inquiry. Rather, the court find that the class action vehicle was exactly designed for this type of case, where there is a large number of plaintiffs who present almost identical legal and factual issues. Therefore, the court finds that Plaintiff has satisfied the requirements of Rules 23(a) and 23(b)(3). Accordingly, Plaintiff’s Second Renewed and Amended Class Certification Motion is granted. D. Jin’s Report The parties disagree about whether Defendants’ computer system contains an administratively feasible way to identify potential class members. The party offering expert testimony bears the burden of proving its admissibility by a preponderance of the evidence. Daubert v. Merrell Dow Pharms. Inc., 509 U.S. 579, 592 & n.10 (1993). Federal Rule of Evidence 702 outlines the basic standard for the admissibility of expert testimony: [a] witness who is qualified as an expert by knowledge, skill, experience, training, or education may testify in the form of an opinion or otherwise if: (a) the expert's scientific, technical, or other specialized knowledge will help the trier of fact to understand the evidence or to determine a fact in issue; - 26 - (b) the testimony is based on sufficient facts or data; (c) the testimony is the product of reliable principles and methods; and (d) the expert has reliably applied the principles and methods to the facts of the case. The Sixth Circuit has elaborated on this requirement by stating that, “[t]he relevance requirement ensures that there is a ‘fit’ between the testimony and the issue to be resolved by the trial. The reliability requirement is designed to focus on the methodology and principles underlying the testimony.” Greenwell v. Boatwright, 184 F.3d 492, 496-97 (6th Cir. 1999) (internal citations omitted). In analyzing any expert’s proposed testimony under this rule, the district judge performs a gatekeeping function by considering the relevance and reliability of the expert testimony. See Kumho Tire v. Carmichael, 526 U.S. 137, 147-48 (1999). A district court has “considerable leeway” in how to determine the reliability of expert evidence in a particular case. Id. at 152, 158. Under Daubert and Federal Rule of Evidence 702, “[a]n expert must offer good reason to think that his approach produces an accurate estimate using professional methods, and that estimate must be testable.” Durkin v. Equifax Check Services, Inc., 406 F.3d 410, 421 (7th Cir. 2005) (internal citation omitted). Plaintiff asserts that the differences in Stirewalt’s and Jin’s opinions go only to the weight of the evidence and should be resolved by the fact finder. Plaintiff relies on Ruiz-Troche v. Pepsi Cola of P.R. Bottling Co.,161 F.3d 77, 85 (1st Cir. P.R. 1998), to support this contention. The Ruiz-Troche court stated that “Daubert neither requires nor empowers trial courts to determine which of several competing scientific theories has the best provenance. It demands only that the proponent of the evidence show that the expert’s conclusion has been arrived at in a scientifically sound and methodologically reliable fashion.” Id. - 27 - Plaintiff has demonstrated that Jin’s opinions and methodology are reliable. As Plaintiff points out in his briefs, Jin was not asked to develop a complete algorithm. Instead, he was tasked with examining Midland’s databases to see if a method for selecting the putative class existed based on the databases. Defendants make much of the alleged deficiencies in Jin’s methodology. However, aside from human error, defendants point to no other factors that would make Jin’s algorithm unsuitable to identify to putative class. Plaintiffs have shown that Jin’s method can be adapted to exclude “exceptions,” and that his algorithm can be further refined to properly identify the class as explained in part E above and as discussed herein. While the parties disagree on how long it would take to refine it, they agree that the algorithm could be refined to exclude the nonDVL recipients and other exceptions. Jin has also shown that the complete algorithm could be developed in a reasonable period of time. Even if Defendant’s expert is correct in saying that it would take a month to develop the algorithm, the court finds that this is a reasonable period of time. Plaintiff also contends that Daubert does not require an expert to come in and actually perform any test, and therefore any argument that Jin did not do enough testing should not be proof that he is not a proper expert in this case. (Pl. Resp. at 21, ECF No. 161) (citing Kamp v. FMC Corp., 241 F. Supp.2d 760, 768 (E.D. Mich 2002)). On these points, Plaintiff is correct. Plaintiff has shown that Jin’s conclusion was arrived at in a scientifically sound and methodologically reliable fashion based on the testing he conducted. As the proponent of expert testimony, it is Plaintiff’s burden to prove that he has met the requirements of Federal Rule of Evidence 702 and Daubert. Plaintiff has met his burden. Jin has provided the court with good reason to think his approach is accurate. He has clarified that the potential exceptions to the class could easily be removed from the defined class using Midland’s own codes. Though Jin’s algorithm is not - 28 - complete at this point, he has explained all of the steps required to properly test the algorithm, including testing for bugs in the code, excluding potential exceptions by using Midland’s other database codes, and narrowing the category to only those members how received a debt collection letter without first receiving a DVL. Therefore, Jin has shown his method to be both testable, and to produce an accurate estimate of the putative class members. Durkin at 421. While Jin acknowledges that there may be some operator error affecting the final class determination, this is not enough to exclude Jin’s testimony as unreliable, as some operator error would be inevitable in any such recording system. The court finds that, because Jin’s methodology is reliable, his testimony should not be excluded. Accordingly, Defendants’ Motion to Exclude the Opinions and Testimony of Ruoming Jin (ECF No. 159) is denied. D. Plaintiff’s Other Motions In addition to the motions decided above, Plaintiff also filed Motion for Hearing on Defendants' Motion to Exclude the Opinions and Testimony of Ruoming Jin (ECF No. 164), Motion for Conference/Hearing (ECF No. 167). In light of the court’s hearing on these matters, Plaintiff’s motions are granted. Plaintiff also filed an Motion for Extension of Time to Reply to Defendant’s Response (ECF No. 172), and a Motion for Appointment of Special Master (ECF No. 177) which are denied as moot. III. CONCLUSION For the foregoing reasons, Plaintiff’s Second Renewed and Amended Class Certification Motion (ECF No. 155) is granted, and Defendant’s Motion to Exclude the Opinions and Testimony of Ruoming Jin (ECF No. 159) is denied. As a result of the hearing held by the court, Plaintiff’s Motion for Hearing on Defendants' Motion to Exclude the Opinions and Testimony of Ruoming - 29 - Jin and Plaintiff’s Motion for Conference/Hearing (ECF Nos. 164, 167) are granted. Plaintiff’s Motion for Extension of Time and Plaintiff’s Motion for Appointment of Special Master are denied (ECF Nos. 172, 177) as moot. IT IS SO ORDERED. /s/ SOLOMON OLIVER, JR. CHIEF JUDGE UNITED STATES DISTRICT COURT November 29, 2012 - 30 -