2.6 Banks per Customer: Why Growth Now Has to Come From the Existing Base

European bank customers now hold relationships with 2.6 banks on average, up from two in 2021. That figure, from a McKinsey analysis published on 3 September 2026, describes the state of retail banking more precisely than any satisfaction survey: customers are not leaving, they are spreading out. The salary account stays with the primary bank, savings move to a direct bank, the travel card comes from a neobank, the brokerage account sits with an online broker. Each relationship on its own is harmless. Taken together, the share the primary bank holds of its customer's financial life shrinks, and revenue per customer shrinks with it.

This article puts the McKinsey numbers in context, adds the Visa study "From AI Promise to AI Performance" of 325 European bank decision-makers presented at the end of September, and draws conclusions for how banks manage their existing customer base.

Growth sits in the existing base, not in acquisition

The McKinsey authors analysed 112 million customer relationships across six European markets. The result: 48 percent of customers hold exactly one product with their bank, and 81 percent hold two or fewer. Put differently, four in five customers are little more than an account from the bank's point of view. At the same time, 69 percent of the European banks surveyed say they want to achieve organic growth primarily by developing existing customers better.

Together, those two numbers describe the gap: the intent is there, the results are not. Banks in the upper quartile of product penetration achieve 13 percent higher balances, 5 percent higher revenues and 22 percent more products per customer than their weakest peers, according to McKinsey. The difference is not in the product range, which is broadly comparable across Europe. It lies in how systematically a bank turns what it knows about its customers into conversations and offers.

The intensifying competition for customers has a clear driver. Digital-first neobanks already attract a quarter to a third of customer relationships in some markets. Revolut has passed 70 million customers, McKinsey notes, and is beginning to compete for primacy with subscription models and a broader product set. Add to this that 99 percent of all service interactions now happen digitally. The branch, where advisers used to notice needs in passing, no longer works as an early-warning system. A bank that wants to understand its customers today has to do so from their data.

Four building blocks of base-oriented customer management

McKinsey describes four building blocks that high-performing banks combine in their customer value management. None of them is new, but the numbers behind them are.

First: customer data and decisioning. Most banks have analytical models that suggest offers. According to McKinsey they are too static and produce too little breadth. Leading institutions complement them with hundreds of behaviour-based triggers in near real time: an unusual change in the account balance, a flight purchase, an abandoned product application. Such triggers improve click-through rates two- to threefold, the authors report. The decisive shift is from the product campaign to the decision at customer level: not "Who do we send the loan offer to?" but "What is the most useful action for this customer right now, and through which channel?"

Second: personalised campaigns and journeys. Banks communicate with their customers three to five times less often than fintechs or online retailers, according to McKinsey benchmarks. Observed implementations suggest banks can send up to one message every one to two days before opt-outs rise noticeably. That number deserves caution, as discussed below. More robust are the results of disciplined sequencing: banks that structure initial and follow-up contacts across push and pull channels have improved lead generation by up to 70 percent. One European bank raised the conversion rate of its mobile account opening from below 2 percent to close to 10 percent by simplifying identification, data explanations and data entry.

Third: measurement and marketing technology. An end-to-end performance dashboard across all channels, A/B testing, app personalisation and a customer data platform with identity resolution form the technical base. Without them, every campaign remains a one-off event from which nothing is learned.

Fourth: operating model and talent. The shift from product-led to customer-led management requires a cross-product, cross-channel mandate with clear accountability. Many banks do not fail for lack of models but because nobody is responsible for the customer as a whole.

One point deserves particular attention in regulated European markets: McKinsey explicitly notes that data collection and the personalisation built on it must comply with applicable privacy rules and customer consent frameworks. In practice this means that purpose limitation, consent status and the right to object have to be part of the decision logic itself, not a downstream filter. A trigger derived from transaction data may lead to a marketing message only where the consent for it exists; as a service notification without sales intent, it is often permissible without separate consent. Mapping that distinction cleanly is less a legal task than an architectural one, and it determines how much of the described potential can actually be realised.

What the Visa study adds: efficiency or trust

The McKinsey perspective describes what is possible. The Visa study of 30 September shows what banks actually do. Of 325 decision-makers surveyed across 17 European markets, 48 percent invest in AI primarily for operational efficiency or employee productivity. 30 percent name customer experience or fraud prevention as their main driver. Visa calls the first group "Efficiency Seekers" and the second "Trust Builders"; 15 percent mainly follow competitors and 7 percent regulation.

The finding for Germany is sharper still: according to the July release of the same study, 77 percent of German bank managers name cost reduction as the most important goal of their AI use, compared with 63 percent across Europe. 54 percent of respondents in Germany and the UK see legacy systems as the central brake on data-intensive applications.

Where the value emerges is telling. Banks that use AI in high-volume, real-time situations, meaning decisions in the moment, fraud detection and personalised services, are 40 percent more likely to report transformational results than banks that use AI mainly in product development or lending processes, Visa finds. Among the Trust Builders, 42 percent of employees save at least two hours a week; among the Efficiency Seekers, 28 percent do. Banks that aim for customer outcomes appear to get the efficiency as a by-product. The reverse does not hold.

The frequency question: more messages are not the answer

The most striking McKinsey figure, one message every one to two days, deserves pushback, at least in a European reading. First, it is an upper bound from observed implementations, not a target. Second, it comes from a context in which every message had a concrete reason. An alert about a suspicious login, a reminder about an instalment due or a warning about a new payee is service, not sales. That customers who receive many such service alerts generate 30 percent more sales, according to McKinsey, argues for relevance, not volume.

Third, GDPR, national marketing law and customers' own expectations set limits. The Celent study for Temenos we discussed at the end of September shows that customers expect their bank to anticipate and explain their needs, not to advertise more often. The right steering metric is therefore not the number of contacts but the share of contacts the customer finds helpful, measured through response, opt-out and complaint. Contact rules per customer that take these signals into account belong in every decision logic before frequency is increased.

What follows for customer management

Taken together, the two studies give a clear picture. The customer is not lost, but distributed. The bank knows more about them than any competitor, yet rarely uses that knowledge in the moment it counts. And AI budgets flow mostly into processes the customer never sees.

In practice, the starting point is not a comprehensive data platform but a concrete use case with the data it needs. McKinsey explicitly advises against building all data pipelines first. A use case such as re-engaging single-product customers to whom the bank can explain a specific second product based on their transactions can be set up and measured within weeks. Each further stage then extends the data base, not the other way round.

Progress becomes measurable through a handful of metrics: the share of customers with two or more products, the response rate per trigger, the opt-out and complaint rate per contact sequence, and revenue per customer against the prior year. Those four are enough to tell whether the approach deepens relationships or merely generates messages.

Accountability matters just as much. As long as product divisions run their own campaigns, they compete for the same customer and produce exactly the arbitrariness that drives opt-outs. A cross-channel mandate for the customer base, with its own budget and its own metrics, is according to both studies the organisational prerequisite for turning models into results.

Five takeaways

1. European customers hold 2.6 banking relationships on average, up from two in 2021; the competition happens inside the existing base, not at the point of switching.

2. 48 percent of customers have only one product and 81 percent two or fewer; banks in the upper quartile of penetration achieve 13 percent higher balances and 22 percent more products per customer.

3. Behaviour-based triggers and sequenced contacts work measurably (click-through 2–3x, leads up to +70 percent) but require decision logic per customer and contact rules.

4. 48 percent of European banks invest in AI primarily for efficiency, and 77 percent of German banks name cost reduction as their main goal; the largest effects, however, are reported by banks that put AI into real-time customer decisions.

5. The steering metric is not contact frequency but the share of helpful contacts, measured through response, opt-out and complaint.

We will continue to follow how banks manage their existing customer base and put the next benchmarks in context as they appear.

Illustration: AI-generated. AI-assisted content: We use AI technologies and automated agents in the creation of our articles, including from Microsoft, Google, OpenAI, Anthropic and other providers. Topics, editorial direction and final approval remain with our team.