A complaint dashboard describes the records that reached it. It cannot show what was lost before the file was built. An inquiry kept in a partner’s system, a failed digital submission or a repeat call removed as a duplicate may never enter the analysis. The totals can reconcile perfectly to an incomplete extract.
This is a difficult weakness to detect from the final report. There may be no blank field or error message. The missing record leaves no obvious trace. For the CCO, confidence in the analysis therefore starts with the route from the customer’s first contact to the records available for review.
The question is broader than whether all cases in the complaint system were analyzed. It is whether the institution understands which experiences that system captures, which it excludes and how those choices limit its conclusions.
Follow the concern across the handoff
The Apple Card enforcement matter illustrates the importance of this route. In its October 2024 order against Apple, the CFPB found that Apple failed to send transaction disputes to Goldman Sachs. Apple controlled the customer-facing interface, while Goldman was responsible for investigating disputes. The failure occurred between the customer’s report and the institution’s investigation process. [1]
The CFPB terminated Apple’s order on September 22, 2025 and waived alleged non-compliance with it. The case is a historical example here. It is not being presented as an active obligation on Apple or as a requirement imposed on other firms. [1]
The practical question remains useful elsewhere. When a partner accepts a concern, what proves that the receiving team obtained the full record and assigned it? A count of messages sent is weak evidence if it cannot be matched to receipt. A successful transfer is also incomplete if the narrative, attachments or original notice date were dropped.
Sample the handoff from both ends. Select records at the point of first contact and trace them forward. Select records in the receiving queue and trace them back. Differences can reveal missing cases, duplicate records or lost details that neither system’s total would expose on its own.
The CFPB’s Compliance Management Review procedures address complaints and inquiries across channels and the institution’s oversight of service providers. These are examination expectations within the Bureau’s scope. The specific duties and deadlines for a customer notice still depend on the law that applies to that notice. [2]
Check what each exclusion removes
Exclusions often have a reasonable purpose. Duplicate removal can prevent double counting. Privacy limits can restrict access. Short summaries can reduce reading time. A clear review asks what those choices remove as well as what they improve.
A repeat contact may be another copy of the same message. It may also show that a promised fix did not work. Combining both as one “duplicate” loses the second fact. Where records are linked to one underlying event, the sequence of contacts should remain available.
An inquiry can describe the same conduct as a formal complaint. A customer may ask where a payment went without expressing anger or using the word complaint. If the analysis includes only a narrow complaint label, it may miss facts that merit review or trigger a specific duty.
Summaries deserve their own check. A record may say that the customer wanted a refund but omit that several earlier requests went unanswered. If the omitted sequence matters to the concern, classification quality cannot repair the loss. Where calls are transcribed, sample the quality of the transcription against authorized source records.
Privacy controls should remain in place. A clear data inventory can show what information is needed for a particular review and who may see it. Where a restriction prevents a conclusion, state that limit. The response to a gap should be an approved way to test it, not wider disclosure by default.
Use dates and populations that answer the question
Three dates can tell different stories: when the customer experienced the event, when it was recorded and when Compliance received it. A rise in this month’s file may reflect a backlog released by a partner. It does not establish that the underlying failure began this month.
Where the event date is unavailable, describe the pattern as newly visible to Compliance. That wording preserves what the records show. It also directs attention to the delay in access, which may be a separate weakness.
Compare periods with similar coverage. A newly added channel can increase observed complaints even if the process is unchanged. A fall can follow the loss of a feed. Before attributing either movement to service quality, check the source systems, record definitions and inclusion rules.
Meaningful denominators help but have limits. Complaints per active account may answer a different question from complaints per disputed transaction. Choose the activity tied to the suspected problem. Even then, the rate describes observed complaints; it does not measure every customer who may have been affected. The CFPB makes a related caution about its public database, which is not a statistical sample of all consumer experiences. [3]
A useful coverage record need not be large. It can identify the sources expected, records received, exclusions, known gaps and changes since the prior period. The deeper check is to sample what was excluded and ask whether the stated reason holds.
What evidence shows that a quiet channel has few problems, rather than a weak route for bringing those problems into view?