The Credibility Discount: Why Relationship Depth Stays Flat Across Financial Services

A credit union calls it products per member. A bank calls it share of wallet, a carrier calls it policies per household, and a wealth firm counts the relationships sitting behind assets under management. All four are measuring the same thing, and at most institutions the number has barely moved in a decade. 

The standard explanation is a targeting problem, so the next planning cycle funds better segmentation and tighter timing against the same offers. Results land within a point of the prior year and the cycle repeats with a narrower segment. 

Something else is going on, and it costs more than anything on the marketing budget. Call it the credibility discount: the markdown a customer applies to financial guidance based on who is delivering it, applied before they have read a word of it.

The Discount is Rational 

Financial institutions have spent a decade building education. Rate explainers, buying guides, coverage checklists, and retirement calculators, most of it modeled on the independent comparison sites customers already trusted, and a fair amount of it more accurate than what a customer would turn up alone. Engagement has been disappointing almost everywhere. 

Quality has very little to do with it. A customer reading about home equity from the institution that would write the loan knows precisely what they are reading. The markdown goes on automatically, and production budget does not move a discount that attaches to the byline. 

The uncomfortable part is that customers are right to apply it. An institution with a lending target does have an interest in which product a customer selects, and a reader who assumes otherwise is being naive. When a customer takes a rate question to Reddit or to a coworker who refinanced last spring, they are making a defensible judgment about whose incentives are aligned with theirs. The behavior looks like distrust. It is closer to accurate pricing of a known bias. 

Any strategy built on persuading customers that the institution is different runs directly into this. The customers who need convincing are exactly the ones who have already discounted the argument.

Talk is Cheap and Everyone in the Market Knows

Economists have a durable explanation for what happens next. Michael Spence's 1973 paper on job market signaling, part of the work that earned a share of the 2001 Nobel for the analysis of markets with asymmetric information, established the condition under which a claim carries information at all. The signal has to be costly, and specifically costly in a way that makes imitation prohibitive for anyone whose quality is lower. A statement that a weak institution could make just as cheaply as a strong one tells a customer nothing. 

Financial marketing has spent twenty years producing statements that cost nothing to make. Every institution claims to put customers first. Every institution says the relationship matters more than the transaction. Neither claim separates the honest institution from the rest, because both can be made at the same price, and customers have adjusted accordingly. 

This explains a pattern that otherwise looks irrational. Institutions with genuinely better products underperform in categories where the claim cannot be verified at the moment of decision. The product is not the constraint. The credibility of the channel carrying it is.

What the Branch Manager Used to Do 

The industry solved this problem once, without knowing it had one. 

For most of the twentieth century, credibility in retail financial services was carried by a person. A branch manager who told a member not to take the loan they came in for was making a costly statement, since it cost the branch a booking and the manager a number. That expense is what made the advice believable. The member could see it, and the relationship compounded from there. 

Scale dismantled that arrangement for defensible reasons. Serving more people at lower cost meant moving the relationship into channels where nobody bears a personal cost for honest advice, and where nobody has the context to give it. The credibility discount is the bill for that trade, arriving forty years later. Institutions kept the language of relationship banking and dismantled the mechanism that made the language credible. 

Credit unions and mutual carriers hold the strongest version of the original claim, since member ownership and mutual structure are real and structural. Almost none of it has ever been demonstrated inside a digital experience where a customer could watch it operate.

Only Expensive Statements Carry Information 

The way out follows from the diagnosis. An institution that wants its guidance to count has to say things that would be costly to say if they were untrue. 

In practice this means showing a customer where the institution's own offer sits against the market, including the cases where a competitor is winning on rate or terms, and then deciding whether to match. The statement carries information because it can lose money. A customer who watches an institution concede a comparison has evidence about its incentives that no volume of content marketing can supply. 

Most institutions will find reasons not to do this, and some of the reasons are good. An institution whose pricing is uncompetitive across the board should not build a mirror. The move works for institutions that win often enough to survive being honest about the times they do not. That describes more of the market than the market believes. 

It also requires knowing what the customer is trying to accomplish before the comparison can be framed, and that is where most institutions actually break.

The Knowing Problem 

Customers describe their next product out loud, continuously, to employees who are paid to listen to them. A member whose hours were cut calls to ask about lowering a car payment. A policyholder mentions a kitchen renovation while sorting out a billing question and never gets asked whether the coverage still fits. 

Then the call ends and the signal stays in the system that captured it, shaped for the transaction that system was built to complete. The branch never learns what the contact center heard, and marketing builds the next campaign from demographics because demographics are the only material within reach. Twelve years into a relationship, whoever picks up the phone next meets a stranger. 

We have spent twenty five years building the systems where this happens. Across service platforms, member and policyholder portals, CRM implementations, and marketing operations in financial services and other regulated industries, the pattern repeats with unusual consistency. The core holds the balances. The digital banking or policy administration platform holds the servicing history. The contact center platform holds what was said. The CRM holds whatever someone remembered to type. Each system is correct on its own terms and none of them holds the relationship, so the institution ends up with six accurate partial records of a person and no view of what that person is trying to do. 

Governance is the Price of Admission 

Anything that reads customer intent and recommends a financial product lands in front of risk and compliance early. Plan for that in the first meeting. 

The line that carries weight runs between advice and decision. A system that surfaces a recommendation for a person to act on sits in a different regulatory posture from one that prices, approves, or denies on its own, and the record across the sector is consistent. CFPB Circulars 2022-03 and 2023-03 establish that using a complex algorithm does not relieve a creditor of the obligation to give specific, accurate reasons for an adverse action. NCUA's artificial intelligence resources point credit unions toward model risk, fair lending, member data privacy, and third party vendor governance. The NAIC adopted its Model Bulletin on the Use of Artificial Intelligence Systems by Insurers in December 2023, requiring a written program covering governance, risk management, and internal audit wherever AI systems make or support decisions in regulated insurance practices, and more than half of states have since adopted it or issued substantially similar guidance. State automated decision law is forming above all of it, with Colorado SB26-189 indicating the likely shape. 

There is a useful alignment hiding in the compliance requirement. A recommendation that traces back to something the customer said, reviewed by a person who bears responsibility for it, satisfies the regulator and reconstructs the costly signal at the same time. The examiner and the member want the same thing. Both want a decision someone can be held to.

The Same Mechanism in Three Vocabularies 

The argument holds across financial services. What changes is the metric, the moment, the regulator, and what an expensive statement looks like when an institution makes one. 

Credit unions.

Depth of membership is the executive language and products per member is the number that reaches the board. The moments arrive unsolicited, since a member calling about a car payment after losing hours is describing a retention risk and a lending opportunity in one sentence, and a member mentioning a coming baby at the branch is describing a savings and protection conversation that almost never happens. NCUA expectations, fair lending, and adverse action explainability set the questions the risk team will ask. The costly statement here is telling a member that their existing auto loan is beating what you can offer, and then either matching it or explaining why you will not. Credit unions have the strongest structural claim to that behavior and the least digital evidence of it. No gap in the sector is sharper. 

Insurance.

The metric is policies per household alongside the health of the book. Intent surfaces around renewal and around coverage gaps a client would never raise unprompted, so most of it lands in service calls and claims conversations where marketing has the least visibility. The NAIC bulletin and the departments enforcing it are explicit that the degree of human involvement factors into how risk gets assessed. The costly statement is telling a policyholder they are over-covered on one line before they discover it themselves. That reduces premium in the current period and changes the renewal conversation in every period after. Carriers and brokerages that can afford this in one line of business should test it there first.

Banks and wealth firms. 

The language is share of wallet and assets under management, and the moments are life events that should start a conversation and instead reach a channel with nowhere to pass them. The costly statement is recommending the lower fee product when it fits better. The revenue concession is real, and that is why it registers with a client trained to expect the opposite.

An Agent Cannot be Persuaded 

Customers are beginning to hand comparison shopping to AI agents. A person describes what they want, the agent gathers offers, and a ranked shortlist comes back. That agent carries no loyalty and no memory of a fee waived in 2019. 

The strategic consequence is larger than a lost application. An agent does not apply the credibility discount, because it does not weigh trust at all. It reads terms. For an institution whose stated advantage is the relationship, an agent-mediated market removes the mechanism through which that advantage was supposed to operate. 

That produces an uncomfortable clarification. Either the relationship produces a materially better offer that arrives before the comparison begins, or it produces nothing the agent can see. Institutions do not get to answer that question in the abstract for much longer, since the answer will be demonstrated in renewal and payoff data whether or not anyone has decided what it should be.

What This Asks of an Institution 

None of it requires replacing a core, a policy administration platform, or a CRM. Those systems capture the signal correctly. The gap is a layer above them that holds the relationship while the transactional systems handle transactions. That is a composability question before it is an AI question. An institution that can add capability without disturbing what already works can move within a quarter. An institution locked into a monolith will spend the first year negotiating with its own platform. 

It is also an operating question, and the harder half. The four motions that determine enterprise performance are converting demand, growing relationships, servicing efficiently, and retaining the people who create long term value. Each one needs the same customer signal, and in most institutions those four functions do not share a definition of the customer, let alone a record. Aligning them is most of the work, and the technology matters only to the degree it makes the alignment hold. 

We have a stake in this argument and it is worth saying plainly. Loominate is the layer we built for it. It captures what customers express in their own words, interprets what each signal means, and delivers the next move into the tools each team already uses. One customer's history lives in one Thread that gets stronger with every interaction, and the guidance it surfaces is benchmarked against the market instead of drawn from the product catalog. That is the specific capability the costly signal requires. Every recommendation traces back to something the customer said, and a person decides what happens next. 

An institution could reach the same place by other means, and some will. The requirement is structural rather than vendor specific. Whatever holds the relationship has to be continuous, has to interpret language instead of clicks, has to reach every team that touches the customer, and has to put a person on the end of every consequential decision.

The Decade This Sets Up 

The institutions that close the credibility discount will do it by making statements that cost them something, in front of customers who have been trained by twenty years of free claims to expect nothing of the sort. 

That is a harder program than another content calendar, and a more durable one, since a competitor can copy a campaign in a quarter and cannot copy a decade of visibly honest behavior at all. Customers have been describing what they need for as long as institutions have been serving them. Most of that description is still sitting in the channel where they said it.

FAQs

What is share of wallet in financial services?

Share of wallet in financial services is the percentage of a customer's total financial business held by one institution. A customer with a checking account, a mortgage, and a credit card at the same bank represents a high share of wallet. A customer with only an auto loan at that bank represents a low one, regardless of the balance.

Each segment measures the same idea under a different name. Credit unions use products per member and depth of membership. Insurance carriers use policies per household. Wealth firms use assets under management alongside the number of relationships behind it.

Share of wallet matters because customers holding several products are more profitable and less likely to leave than customers holding one. It matters more now because acquisition is weakening. National Credit Union Administration data for the year ending Q3 2025 shows membership declined 0.5% at the median credit union, and about 55% of federally insured credit unions had fewer members than a year earlier, even though aggregate membership grew.

Why do financial services cross-sell programs underperform?

Financial services cross-sell programs underperform for two reasons that better targeting cannot fix. Customers discount guidance that comes from the institution selling the product, and institutions rarely know what a customer is trying to accomplish at the moment the offer goes out.

Gallup's April 2025 survey of 2,036 U.S. adults found Americans seek financial advice from friends and family (43%), financial advisers (41%), and financial websites (36%) before banks and credit unions (32%). Among adults aged 18 to 29, social media ties financial websites at 42%, and both rank above banks and credit unions at 34%. An offer arriving from a party with an obvious interest in the outcome gets marked down before a customer evaluates it.

The second reason is a knowledge gap inside the institution. Customers describe what they need in service calls, branch conversations, and chat sessions, and that signal stays in whichever system captured it. Marketing then builds the next campaign from demographic attributes, since demographics are the only material available to it. Programs built on

Can AI be used in lending or underwriting decisions?

Yes. AI can be used in lending and underwriting decisions in the United States, and the regulatory expectations governing it are already established. The distinction that determines exposure is whether a system recommends an action for a person to take or prices, approves, and denies on its own.

The practical position for most institutions is advisory intelligence with a person making every consequential call, and recommendations that trace back to something the customer said so the audit trail exists during normal operation rather than being reconstructed during an examination.

How do financial institutions compete with AI comparison shopping?

Financial institutions compete with AI comparison shopping by making a better offer before the comparison starts, since an AI agent cannot be persuaded. An agent does not weigh brand, tenure, or trust. It gathers offers, ranks terms, and returns a shortlist, and it carries no memory of a fee an institution waived years earlier.

The loss is difficult to detect. An institution absent from the shortlist finds out when a payoff request arrives or a renewal does not, and the customer made no decision to leave. They asked a question somewhere the institution was not present.

Two capabilities determine whether an institution can act in time. The first is visibility into customer intent while it is still forming, in a search, a service call, or a question asked in chat, rather than learning about it from the outcome. The second is a willingness to benchmark honestly against the market, including when a competitor is winning on rate or terms, since a customer who watches an institution concede a comparison gains evidence about its incentives that marketing cannot supply.