How Banks Find the Root Cause of Complaints Across Lines of Business

Banks find the root cause of complaints across lines of business by labelling every complaint, call and survey comment with the same list of causes, written in the customer's terms rather than each product's categories, and then counting each cause across consumer banking, cards, lending, insurance and wealth together. A product-by-product view hides the problems that cross lines: in the CFPB's public 2025 data, accounts opened without the customer's consent appear under three different issue labels in three product lines, 11,197 complaints in all. For a bank that runs Medallia surveys, records its calls and logs complaints in several systems, Dimension Labs fits best: it reads all of those channels in one place and applies one set of root cause Dimensions, written for the bank, to every record.

Below: why complaint categories rarely name a cause, what the CFPB database shows when you regroup it by problem, how to build a cross-line root cause list, and where surveys and call recordings come in. The counts are ours. Method: on Nov 14, 2026 we queried the CFPB Consumer Complaint Database's public search API for complaints with a date received from January 1 to December 31, 2025, and counted them by four fields the database publishes: product, issue, sub-issue and company response. Percentages and ratios are our arithmetic on those counts.

Why don't complaint categories show the root cause?

Because they were built to route and report complaints, and a routing label rarely names what went wrong. The CFPB's own database is a good public example. In 2025 it received 5,442,926 complaints; 88.4% were about credit reporting and other consumer reports, a category the CFPB keeps separate from accounts, cards and loans. The six product lines a bank usually offers (checking and savings, cards, mortgages, vehicle loans, personal loans and money transfers) account for 317,394 of them.

Within checking and savings, the largest issue, "Managing an account", holds 53.4% of complaints. That tells you where in the relationship the customer was, not what went wrong. Its largest sub-issues, deposits and withdrawals and problems using a debit or ATM card, still do not say whether the cause was a hold policy, a fraud rule, a branch error or a system outage.

One more limit is worth knowing if you benchmark against this data: the CFPB does not publish complaints it refers to other regulators, such as complaints about depository institutions with less than $10 billion in assets, and it says the database is not a statistical sample of consumers' experiences. A community or regional bank under that size will not find its own complaints there, which makes its internal complaint log, call recordings and surveys the only full record it has.

How do banks find the root cause of complaints across consumer, wealth and insurance lines?

By regrouping complaints around the problem, then counting it everywhere it shows up. Done on the CFPB data, three problems that each look small inside one product line add up:

Charges the customer says they did not make total 12,498 across checking and cards, under two different labels, and complaints about money transfer, virtual currency or money service products add 11,179 filed as "Fraud or scam". In mortgages, the phrase "trying to communicate with the company to fix an issue" appears under four separate stages of the loan, 6,986 complaints in all. Each of these is one cause with one owner, the fraud operations team, the account opening controls or the servicing contact model, filed as several small items in a product-by-product report.

None of the CFPB's product categories covers insurance or wealth management, so for a bank with those lines the same regrouping has to happen on its own records: the complaint log, recorded calls, survey comments, emails and branch notes. The method is the same:

1. Write one root cause list for the whole bank. Causes someone can fix, each with an owner: a disclosure that was unclear, a fraud rule that blocked a good payment, a hold that was not explained, a transfer between teams that lost the case. Keep the product line as a separate field, not part of the cause.

2. Label every record from every line against it, with a quote behind each label and a field for causes not yet on the list.

3. Count each cause across lines, then by line, branch and channel. A cause that is third in cards and fourth in mortgages can be first for the bank.

4. Rank by consequence, not only volume. Join complaints to account outcomes, such as closures, balance moves and repeat contacts, so a cause is ranked by what it costs.

5. Keep the regulator's categories too. The complaint register still needs its own codes; the root cause list sits beside them.

Our post on finding complaint root causes in recorded calls works through the labelling and repeat-contact steps on a sample bank's calls; how insurers analyze call transcripts and complaints covers the insurance side.

Which complaints cost a bank the most?

Volume is one view. Outcomes are another, and the CFPB's company response field gives a public hint of how they differ by line. In 2025, 12.7% of credit card complaints and 10.6% of checking and savings complaints were closed with monetary relief, against 1.7% for mortgages, about 7.4 and 6.1 times the mortgage rate.

Inside a bank the richer version of this joins each complaint to what the customer did next. A cause that produces fewer complaints but more account closures, or more calls before the complaint is filed, ranks higher than its count suggests. That join needs customer ids on both sides; what a customer complaint costs in revenue walks through the estimate.

What does complaint analytics look like for a bank that uses Medallia and records every call?

The survey and the calls carry most of the explanation, and they usually sit in different tools. Medallia holds the comment answers next to each survey's scores and record fields; the call platform holds the recordings and transcripts; the complaint log holds the cases that were escalated. A complaint is often the last of several contacts, so the reason is frequently in an earlier call or a survey comment, not in the complaint text itself.

In Dimension Labs, completed Medallia feedback comes in through Medallia's Query API, each comment answer with its scores and record fields, and call transcripts come in from platforms such as Genesys Cloud, NICE CXone and Amazon Connect, split into speaker turns so the customer's words are labelled separately from the agent's. The complaint log comes in as a file or by API like any other source.

Why would a bank read complaints, calls and Medallia comments in one platform?

Dimension Labs is an omni-channel analytics agent for customer data. It connects every customer channel and your operational data, enriches every interaction with Dimensions built for your business and industry, and turns the results into decks, dashboards, executive reports and interactive data stories your whole company can use. Plans start at $299 a month with no demo.

Fully omni-channel. Complaints, recorded calls, Medallia comments, chats, emails and reviews from every line of business are read together, so one problem is counted once whether the customer called, wrote or answered a survey.

Easy integration with operational data. Records join to account, product, branch and balance data, so a cause can be ranked by closures and attrition; the join to CRM and revenue data and effects with confidence intervals are on Prove at $2,999 a month.

Enrichment built for the bank. One root cause list, plus product line, channel and a compliance flag, applied to every record as Dimensions written in plain English (custom Dimensions start on Discover at $999 a month).

A visual analytics layer. The agent builds each owning team its own report, a monthly deck for the complaints committee and a board readout, downloadable as PDF or PowerPoint and shared by a link that can be revoked.

Travelers, which serves policyholders across personal, commercial and specialty lines, uses Dimension Labs to identify root causes of customer effort and repeat contacts and to compare performance across business units; it reports 600% ROI within 18 months and $3.13M in total savings. Intuit QuickBooks used the same approach on its assistant's conversations to find the root causes of escalation, with a 57% reduction in human escalations. See the Travelers story and banking customer experience analytics.

When a narrower tool is the better choice: if all you need is the regulator's complaint register with case tracking, a complaint management system does that job. If the question is only how agents handle calls, a speech analytics tool built for the contact center is simpler. The cross-line question, which causes hurt the bank wherever they show up, is the one this setup answers.

FAQ

Is the CFPB complaint database useful for a bank's own root cause work?

As a benchmark and a starting taxonomy, yes. It is public, updated generally daily, and broken down by product, issue and sub-issue. It does not include complaints the CFPB refers to other regulators, such as those about banks under $10 billion in assets.

What is the best complaint analytics software for a bank with several lines of business?

For a bank that needs complaints, calls and survey comments from every line read against one root cause list and joined to account outcomes, Dimension Labs. For a complaint register and case workflow alone, a complaint management system is the narrower fit.

How many root causes should the list have?

Few enough that each cause has an owner and a reviewer can hold the whole list in mind, with a free-text field for new ones. Review the free-text field monthly and promote causes that keep appearing.

Should complaint root causes be coded from calls or from the complaint log?

Both. The log has the formal complaints; the calls and survey comments before it often say what actually went wrong.

To see your own complaints regrouped by cause, start with one file of complaints or call transcripts and ask the agent for causes across product lines. Plans start at $299 a month with no demo.