How a Record Turns Into a Risk Signal
This chapter examines the specific mechanism at the center of the RentGrow allegations: how one criminal or eviction matter can appear in a tenant-screening report as several records. The goal is to distinguish a separate case from a separate event, then follow the effect of grouping, labels, and layout on the eligibility signal a landlord or property manager receives. The distinction matters because a report can make an applicant’s history look more extensive without discovering any new underlying proceeding.
The problem is not unique to a particular database design. A court case unfolds through events. It may begin with a filing, move through hearings or charges, and end with a judgment or another disposition. Different repositories may store those events separately. A reporting company that gathers the entries must decide whether they describe one proceeding or several. That decision is part of the report’s meaning.
One proceeding can become several entries
The complaint alleges that RentGrow did not use reasonable procedures to prevent duplicative case records and multiple entries for the same criminal or eviction action. It describes two common ways this can happen. The same conviction may be present in more than one public-record repository. Or a single eviction proceeding may include several events, such as an initial filing and a judgment of possession. If the report treats every incoming line as a separate record, the reader may infer several convictions or several evictions when the sources describe one matter.
A useful distinction is between an event and a case. An event is something that happened in the course of a proceeding: a charge was filed, a hearing occurred, a judgment was entered, or a case was dismissed. A case is the connected proceeding that gives those events their relationship and context. A report does not necessarily have to show only one line for every case. It may need to show the stages of a proceeding. But it should make clear that the stages belong together and should not count them as independent proceedings merely because they arrived from different sources or have different dates.
This is a data-integration problem, but calling it that does not make it less substantive. The system has to compare identifiers, case details, event types, and source structure. It has to notice when one source already groups the stages of a case and avoid discarding that relationship. It also has to distinguish a genuine second case from another entry describing the first one. The retained documents do not disclose RentGrow’s internal matching rules, so they cannot show exactly how those decisions were implemented. They do show why a procedure for grouping and deduplication matters.
The CFPB’s advisory opinion states the principle in terms of how the report is understood. Consumer reporting agencies must have reasonable procedures to prevent duplicate information from being reported in a way that inaccurately suggests one event occurred more than once. When a report includes multiple stages of the same court proceeding, the stages should be collated and presented so the reader can tell they relate to one case. The concern is not that a company may never report more than one line. The concern is that the format should not transform related stages into an inflated count of adverse events.
There are two different ways for context to disappear. The first is duplication: the same conviction or proceeding enters from two sources and is counted twice. The second is fragmentation: the report contains different stages of one case but does not show their connection. Duplication inflates the number of matters. Fragmentation can make an ordinary sequence look like a series of independent matters. Both can produce the same practical impression, but they require different checks. A procedure that removes identical entries may not recognize that a filing and a judgment share one case. A procedure that groups familiar event types may wrongly merge two genuinely separate cases if it does not preserve enough identifiers.
The transformation can be pictured as a sequence:
The critical step is not only the collection of B, but the transition from C to D. The underlying events may be real, yet the displayed records can still be misleading if the relationship among them is lost. Once the report reaches the screening criteria, the landlord or property manager may see the count and labels rather than the source history that would explain them. A later decision can therefore respond to a representation of the record, not to the record in its original context.
A grouping decision is also a decision about what information remains available to the report user. If related entries are nested under one case or clearly labeled as stages, the customer can interpret them as a connected sequence. If they are separated into numbered records, the customer may reasonably read the number as a count of cases. A compact report can be easier to use while still being less informative if it removes the identifiers or headings that explain the relationship. Conversely, a longer report is not automatically more accurate if it repeats the same matter without clarifying why the lines are present. The relevant question is whether the organization preserves the evidence needed to interpret the count.
This distinction also prevents an opposite mistake: assuming that every repeated-looking entry is a duplicate. Two separate cases may involve the same person, the same type of proceeding, or similar dates. The evidence supports a need for reasonable procedures, not a rule that a screening company should collapse all similar entries. Accurate grouping must preserve genuinely distinct matters while preventing one matter from being counted more than once. Without the underlying records and the company’s procedures, the sources cannot resolve every individual comparison.
When presentation changes the signal
The complaint alleges that the problem extended beyond the existence of duplicate information to the way the report displayed it. It describes entries that were listed separately, displayed across multiple pages, or presented without clear headers and breaks. Small type and confusing layout allegedly made it difficult to see that records were related. These details may sound cosmetic, but presentation determines what a customer can reasonably perceive from a report.
In the arrangement described by the complaint, the landlord or property manager receives a screening product intended to help apply rental criteria rather than the raw history of every source database. If the product labels six lines as “record one” through “record six,” the visible count can become a practical signal of risk. A customer may not know that several lines came from the same proceeding or that the same record was repeated across pages. The report’s layout has then become part of the decision mechanism.
The complaint gives three examples that illustrate this path. In one disputed report, RentGrow allegedly displayed three eviction-history events as separate entries even though two belonged to the same proceeding, leaving two records rather than three. In another, six separately enumerated records allegedly included duplicates, so that only two records remained when the duplicates were recognized. In a third, the complaint says RentGrow listed fifteen criminal charges across seven records even though multiple records and charges duplicated one another. These are allegations in a pleading, not a statistical estimate of how often the problem occurred. Their importance lies in the causal sequence: incoming material, a grouping or presentation choice, a displayed count, and a screening result.
The third example also shows why duplication is not merely a recordkeeping concern. The complaint alleges that RentGrow determined the applicant did not meet the prospective landlord’s criteria on the basis of the criminal history, and that the decision was reversed from denial to acceptance after the consumer disputed the report and all fifteen records were removed. That example supports a carefully bounded conclusion. In at least one pleaded instance, the report’s contents were connected to an eligibility outcome. It does not prove that the duplicated entries were the only reason for the initial result, that every similar dispute would change a result, or that the alleged practice was representative of all reports.
The difference between those conclusions is important. A report can contain an error that does not change a landlord’s decision because another criterion leads to the same result. Conversely, a report can influence a decision even when the landlord never sees the underlying error directly. The causal question is not simply whether an entry was duplicated. It is whether the duplication changed what the report communicated and how that communication interacted with the customer’s screening criteria.
This interaction can be understood without assuming that the screening rule itself is unlawful or irrational. Suppose a property manager’s rule responds to the apparent number of eviction proceedings, while the report does not distinguish proceedings from stages. The rule may be applied exactly as written, yet the input no longer means what the rule assumes it means. The failure would lie in the relationship between the report’s representation and the criterion’s interpretation. If the criterion instead responds to another part of the application, removing duplicates might not alter the outcome. The same reporting defect can therefore have different consequences in different decisions, which is why the complaint’s example supports a mechanism, not a universal outcome.
That is why the CFPB frames grouping as part of maximum possible accuracy. A report should not merely contain the right source fragments; it should present them so a user does not draw an inaccurate conclusion about how many proceedings or convictions occurred. The user’s interpretation is part of the practical output. A technically faithful copy of disconnected source lines can still be misleading if the report’s organization implies a history that the sources do not support.
The boundary between a real record and a misleading account
The RentGrow allegations also show why a clear evidence boundary matters. The complaint says that records received from other consumer reporting agencies sometimes arrived with headers and grouped events. It alleges that RentGrow disregarded that formatting in favor of its own presentation. If that account is proved, the relevant failure would not be that the public records never existed. It would be that the report discarded context that was necessary to present them accurately.
The CFPB’s guidance strengthens the general principle but does not prove the facts of the complaint. It says that agencies reporting public records need procedures to identify duplicates and to collate stages of one case. It does not establish that every report from every provider contains the same type of duplication, or that every format difference is misleading. The complaint supplies the case-specific examples; the advisory opinion explains why the mechanism matters under the FCRA.
Nor can a reasonable procedure be judged solely by the final number of lines. A report with two entries may still be misleading if each entry combines several unrelated matters. A report with six entries may be accurate if it clearly identifies six distinct proceedings. The procedure has to preserve enough context to support the conclusion, and the presentation has to make that context usable by the landlord or property manager. The retained evidence does not give a universal threshold for how much detail every report must display. It supports a more basic requirement: the visible account should not imply a number of cases that the underlying information does not support.
This distinction gives the reader a disciplined way to assess a screening report. First ask what distinct proceedings the sources actually establish. Then ask which events belong to each proceeding. Finally ask whether the report’s count, labels, and layout preserve those relationships for the landlord or property manager who uses the result. A line-by-line match to source material is not enough if the relationship among the lines has disappeared.
The same test protects against overcorrection. Combining two genuinely separate eviction cases into one would understate the record just as surely as counting one case twice would overstate it. The objective is not the smallest number of entries. It is a representation that keeps the number, sequence, and status of proceedings understandable. For RentGrow, the complaint makes that representation the point at which an automated report can become a materially different signal from the underlying history.
Once that signal is visible, the broader causal chain is clear: sources provide events, the reporting system groups and displays them, a customer applies criteria to the display, and an applicant may bear the result. The question is no longer whether the system retrieved information. It is whether the system preserved the difference between one case and many, and whether the landlord received a signal that the evidence could support.