Meeting the Growth Challenge Requires Precision Over Volume

Most growth in retail banking isn’t won at the top of the funnel — it’s already sitting in the book. Roughly 75% of deposit growth in a given year comes from customers a bank already has. That doesn’t make acquisition disposable: this year’s new customers are next year’s existing ones, and a book that isn’t replenished eventually runs dry. But acquisition has gotten harder and more expensive – costs have doubled, and fewer than 1% of prospects bring meaningful balances. The opportunity isn’t choosing between existing customers and new ones. It’s getting more value out of both: deepening the relationships already on the books, and targeting the right prospects to bring in. Doing so doesn’t require greater spend, it requires greater precision.

Growth in retail banking has stalled. And it’s not for lack of investment. Banks have spent more than ever on marketing, they’ve gone all-in on digital, and they’ve flooded the market with cash offers and aggressive pricing. Yet according to Curinos data, only one in 10 financial institutions achieved 3% growth for both customers and deposits in 2025. Half the banks experienced no growth in either or both (Figure 1).

The problem isn’t where FIs have been investing, or by how much. It’s that most continue to optimize in silos: marketing reallocates spend, product adjusts pricing, retail rethinks distribution. Each lever improves in isolation, but enterprise growth remains constrained. The deeper problem is structural: most banks have no closed loop between planning, execution and measurement. Strategy is set in one function, executed in another and measured in a third — so decisions stay fragmented and calibrated to activity rather than outcome, blind to the customer relationships that hold most of the value.

The core message is clear: growth can no longer rely on acquisition strategies based on volume alone. It will depend increasingly on improving visibility into the full customer relationship, using AI-driven decision intelligence. And it will depend on shifting investment from broad acquisition tactics to targeted, 1-to-1 action focused on acquiring, deepening, and retaining high-potential existing customers.

This paper draws on over a decade of proprietary Curinos research as well as data across 65+ institutions to lay out what’s happening, and what it will take to compete differently and lay the groundwork for durable advantage going forward.

Part One: The Headwinds Are Structural

The Math Has Changed. The Playbook Needs to.

The cost to acquire a checking customer doubled from 2018 to 2025, ballooning to $559. And more than half of all people who switch providers hold 4+ banking relationships, up from a mere 7% in only six years. This has led to greater churn, with balance attrition up 25% since 2019 (Figure 2).

These aren’t independent trends. They’re interconnected and they compound. At today’s cost of acquisition, replacing a lost customer can be four times more expensive than retaining them. The economics of the old model — acquire broadly, compete on rate, hope they stick — have frayed considerably. The math has turned against it.

Going Digital Is Lowering Acquisition Quality, Not Just Cost.

The industry’s response to rising acquisition costs has been to lean into digital channels. That’s because digital is cheaper and more scalable. And, without question, it’s where consumer preference is headed. In 2025, 40% of new-relationship acquisitions came through digital channels, up from 29% just two years earlier. Curinos projects that by 2027 the number will exceed 55%.

But the quality of digitally acquired relationships remains elusive. After 12 months, a scant 41% of them remain on the books, compared with three out of four for those sourced at the branch. After two years, digitally acquired accounts display only 27% of the balances that are initiated in branches (Figure 3). And it gets worse: While these customers are materially less valuable, total relationship acquisition has barely budged since 2023. It’s clear that digital acquisition is replacing branch sales, not supplementing them.

Source: Curinos Distribution Analyzer, Curinos Analysis

After two years, digitally acquired accounts display only 27% of the balances that are initiated in branches.

As acquisition has shifted from branch to digital, banks will have to make up for reduced balances. For every $1B in balances acquired in 2023, banks will need to acquire 54,000 more new customers to achieve the same level of balance acquisition in 2027. That’s a daunting 41% increase (Figure 4). Making this up through broad-based marketing campaigns and rate offers is too expensive for the typical regional bank.

Instead, banks will need to target the right customers – those most likely to switch brands and bring material balances. They’ll need to improve their onboarding, activation and relationship deepening once the digital customer commits. And they’ll need to use digital and physical channels in tandem to identify the best way to interact with existing customers to grow relationships over time.

These are solvable problems, not a verdict. The sections that follow show how the banks that are winning are converting these same digital relationships into durable value — and why the answer is better decisions, not bigger budgets.

Part Two: Where Growth Actually Comes From

Portfolio Is Three-Quarters of the Game

The single most consequential finding in this paper is this: On average in a given year, 75% of deposit growth comes from existing customers (Figure 5). Not new-to-bank acquisitions. Not promotional rate campaigns. The customers already on the books. In accounts the bank already owns, making decisions the bank rarely sees and almost never anticipates. This doesn’t diminish the importance of acquisition, which is needed to fuel future growth – acquisition and portfolio growth should be treated as complementary, rather than competing. The imbalance this paper targets is one of proportion: most banks over-index on chasing the next customer and under-invest in growing the ones they already have.

On average in a given year, 75% of deposit growth comes from existing customers.

The obsession with acquisition is hardly irrational. New customers are visible. Campaigns are measurable. CPA is an easy number to track and evaluate. But what gets measured gets managed, and what doesn’t get measured — the quiet compounding of customer relationships over years — systematically gets the short shrift.

Hiding in Plain Sight: Customers Who Look Least Valuable May Be Your Future

A five-year cohort analysis reveals an eye-opening case in point: many of the customers who appear least attractive at acquisition ultimately become the institution’s most profitable.

When a new-to-bank customer opens an account with less than $10,000 in deposits, they look, by almost every standard metric, like a low-value relationship, contributing only 24% of deposit acquisition value in year one (despite representing 87% of the new customer population). A conventional model would typically reduce them to low priority.

But about 20% of these customers evolve into relationships that are 3x more valuable within five years — representing as much as 60% of a bank’s deposit growth — as they consolidate balances, deepen product usage, and increase engagement. The low-balance customers of 2021 have become among the highest-value customers of 2025 (Figure 6).

20% of <$10k-deposit customers become 3x more valuable within five years — and account for as much as 60% of deposit growth.

The problem is that too many banks have already given up on them, having focused instead on balances at account opening rather than on the long-term relationship. The gradual, silent evolution to high value has fallen off their radar screen. The bank that sees this graduation clearly and nurtures these relationships intentionally will capture that value. The bank that ignores them until their value is evident will not.

One thing is clear: relationship depth matters. When a customer holds multiple products and has meaningful engagement with the institution, CD renewal rates reach 88–94%. For CD-only customers, renewal rates fall to 70–88%. Meanwhile, rate-led acquisition strategies generate 64% higher churn when measured for retention over a five-year period (Figure 7). In both cases, the difference isn’t rate, it’s relationship. And a bank can’t act on a relationship it can’t see.

Traditional segmentation models and static scoring systems often miss the dynamic behaviors that signal future value. What’s needed are systems that can continuously observe customer behavior, identify emerging potential, and adapt engagement strategies before those opportunities are lost.

Part Three: Protecting the Portfolio Means Getting Acquisition Right

How You Reach Them Today Predicts Their Value Tomorrow

Precision targeting at acquisition isn’t merely a cost-efficiency play; it’s portfolio protection – keeping price-sensitive, low-value relationships from diluting the book before they ever get there. Why does this matter? Because only 6% of the population switches checking providers per year and only 15% of them bring more than $10k in balances when they do, so a small sliver of prospects – about 1% – represent meaningful deposit value (Figure 9).

Source: Curinos Deposit Analyzer, Curinos One Capture, Curinos Analysis

6% of people switch x 15% that bring more than $10k = <1% of prospects represent meaningful deposit value.

Traditional funnel metrics such as response rates and acquisition volume are insufficient. What’s needed is customer-level decisioning — identifying individuals most likely to deepen balances, consolidate relationships and respond positively to specific engagement strategies. In other words, the ones that will grow profitably once on the books.

And the solution isn’t to flood them with offers once they’ve been identified. Curinos data have shown that only 25% of deposit promotion balances bring in new money. The remaining 75% is cannibalized from lower-rate accounts in the back book. The institution has paid a hefty premium to redistribute the balances it already has. Just as significantly, rate-led acquisition selects for the most price-sensitive customers — the segment most likely to leave at the next competitive offer. The bank that wins on rate will lose on rate.

The lesson here is that traditional metrics such as response rates and acquisition volume tend to fill the funnel with empty calories. The institutions outperforming today are moving beyond broad demographic targeting toward customer-level decisioning. That requires more than analytics dashboards. It requires decision systems capable of turning insight into action, continuously and at scale.

Judging acquisition success solely on lower cost can be a fool’s errand. Consider digital display advertising versus direct mail. On the surface digital display appears significantly cheaper — about $350 per acquisition versus $680 for direct mail, according to Curinos Marketing Analyzer (Figure 10). But value down the road tells a much different story.

Digital campaigns primarily drive customers into digital onboarding journeys, where customers tend to fund accounts at lower rates and maintain smaller balances over time. Direct mail, by contrast, generally pushes customers toward branch-assisted onboarding. Although more expensive to attract upfront, these customers typically arrive with higher balances, and they stick around. At the same level of marketing spend, direct mail generates 34% more retained balances after 12 months because it produces more advantageous funding.

Even though more expensive upfront, direct mail generates 34% more retained balances after 12 months than digital campaigns.

This is hardly an argument against digital. Its evolution is structural, and the lost opportunity from resisting it is enormous. The answer isn’t to pull back from it – it’s to treat activation and onboarding as seriously as acquisition itself, so digitally acquired customers convert into the same durable relationships branch customers do. That means channel mix optimized for relationship quality at 12 to 24 months, not for cost per acquisition on day one. It means onboarding experiences built to anchor the relationship early and continuously. And it means having the in-flight intelligence to flag an underperforming direct-mail creative in a single market and pull it before the spend is wasted — not at the next quarterly review.

Part Four: Why Banks Are Stuck

Three Obstacles Between Data and Outcome

Most banks have more data than they can use. They have CRMs, marketing automation platforms, pricing engines, and analytics teams. Yet there’s still a yawning gap between the insight that exists somewhere in the institution and the decision that gets made in front of the customer. Three structural obstacles stand between the data a bank holds and the decision it makes (Figure 11). They aren’t three independent problems but interlocking ones — each makes the others harder to solve.

A bank without a unified customer view can’t orchestrate effectively, even if it wanted to. And even with unified data and orchestration capability, the wrong KPIs will yield the wrong decisions. Neither unit growth nor cost per acquisition will produce relationship quality. Nor will balance acquisition that doesn’t account for back-book cannibalization. The emphasis needs to move to customer-centered decision-making at the bank level. This is the crux of the shift in an omnichannel world. Decisions that may seem entirely rational but are disconnected need to coalesce around the customer instead.

The “old world” approach is no longer sufficient in an environment where customers expect highly personalized, contextual experiences similar to those delivered by leading digital platforms. The “new world” begins with the customer and dynamically determines the right product, offer, pricing action, or engagement strategy delivered at the right time against a unified enterprise objective (Figure 12).

Easy to say, but how to solve this critical business challenge more effectively – especially in a regulated industry where relationships are built on trust, transparency, and fairness? The answer is by putting in place a governed, compliant decision layer to convert data and insights that banks already have into impactful executions.

Individuals are Individuals, Not Segments

Another impediment to profitable growth is relying too heavily on segmentation. To be sure, product design and value propositions will always operate at the segment level. A mass market proposition and a mass affluent proposition reflect genuinely different needs, different price sensitivities, and different feature priorities. Defining them by segment is efficient and correct.

But bringing a proposition to life requires something segments can’t provide: individual context. Two customers who fall into the same sub-segment may need the same financial product, be looking for the same core benefit, and even be seeking the same reason to believe. But the right channel, message, and timing for each will differ materially based on individual preferences, demonstrated behavior, and relationship history (Figure 13).

Executing at the individual level, the customer who is digitally savvy, for example, receives a push notification. The customer who distrusts technology and finds apps frustrating receives outreach from a banker. The customer who is anxious about their finances and watches their accounts carefully receives in-app messages and email that emphasize clarity and control. Same proposition, same proof point, but three radically different executions, each differentiated by what the bank actually knows about each person.

This isn’t a vision for the future; it’s a capability that exists today within the institutions that have built the data and decision infrastructure to support it. They’re making better decisions with what they already have, building the feedback loops that make each subsequent decision more accurate, and compounding their advantage every day.

Part Five: Decision Intelligence: A Different Way to Compete

The Decisions That Banks Depend On Are Disconnected

The systems that power the decisions around growth — who to acquire, how to price, what to deepen, who to keep — aren’t talking to each other. And that means lost opportunity. Yes, banks have signals, channels, CRMs, marketing automation tools, and pricing engines. But what many lack is the intelligence layer to connect them — the capability to take everything the institution knows about a customer and convert it into the action most likely to move that relationship toward higher value.

Decision intelligence isn’t Curinos coinage — it’s an emerging category that industry analysts, Gartner among them, now track in its own right. In plain terms, decision intelligence is the technology that recommends a clear next action — with a predicted, explainable result that fits the goals and constraints the institution has set — and then improves over time by tracking the loop from goal to decision to result. Its organizing unit isn’t the dashboard or the report; it’s the decision itself, justified by the data models behind it.

That, at least, is the standard. But not every system wearing the label meets it. Before trusting any decision engine, a bank should ask three questions: Does it record the decision the bank actually made, including when a committee overrides the recommendation? Does it measure its actions against control groups, so results are demonstrated rather than claimed? And can a reviewer or regulator trace why any recommendation was made? A system that can’t answer yes to all three is only analytics with a new name.

The architecture has four components:

• A unified data layer – the customer “spine” that consolidates the bank’s own first-party signals and third-party signals into a continuously updated customer view.
• A decision engine that evaluates each customer against the bank’s objective function.
• An activation layer that determines which channel, message, and timing will be most effective.
• A learning loop that feeds outcome data back into the model to improve every subsequent decision.

On the acquisition side, the questions the engine answers are: Who should be targeted in-market to achieve stated objectives? What is an economically viable customer acquisition cost based on the value that prospect will actually bring? What is their full wallet opportunity, and over what time horizon?

On the portfolio side, it answers: What attributes make a customer valuable? What untapped opportunity exists in the current book? What is the potential future value of each existing customer, and which behavioral signals indicate a graduation is approaching?

Taken together — the architecture and the questions it has to answer — that’s what decision intelligence needs to do.

Don’t Banks Already Have This?

It’s fair to ask why this requires a banking-specific platform at all. Most institutions already own general-purpose orchestration tooling. Many already have models that recommend the next-best-action. And some will hesitate to place decisioning – the thinking part – outside its own walls. Each concern deserves a straight answer.

Orchestration tools do exactly what they’re programmed to do – and that’s their limit. Someone has to design every path in advance – this segment, this trigger, this message, this wait. With a handful of segments and a few branches, that’s manageable. But decide at the level of the individual customer – the right product, offer, channel, and moment for each of hundreds of thousands of people, adjusted as their behavior changes – and the number of paths grows beyond what any team can write, test, or maintain.

This is not decisioning. It’s the wall described in Part Four of this paper: propositions are built for segments, but execution happens one customer at a time. To scale, this executional layer needs direction from something that decides. Agent-based tools now promise to generate those paths automatically. That solves the authoring problem and sharpens the deeper one: an agent that can act at unlimited scale still has to know which actions create funded, durable value. And in a regulated industry, every one of those actions needs to be explainable, recorded, and measured. Automation raises the stakes of each decision. It doesn’t make the decision.

Next-best-action models come closer. Most of them predict who will click, open, or accept. That’s engagement. But none ever decides to do nothing. A journey runs because a trigger has fired; a next-best-action model, by design, always has a next best. Yet for many customers on many days, the most profitable action is no action at all: no promotion for a customer whose money would simply move over from their own lower-rate account, no message for a relationship that’s growing on its own. And no message fatigue for the customer.

So, unless your next-best-action model predicts the outcome that counts – whether the money arrives, whether it stays, whether the relationship grows and treats no action as a choice – the model is falling short of deciding which message is worth delivering, to whom, and whether to send one at all.

As for handing over the decision, the thing worth protecting has never been the software. It’s the bank’s data, its decisions, and what it learns from the results — those are worth protecting. What an in-house build actually costs is time: years of decisions that have gone unmeasured, learning the bank can never get back. The bank that starts learning first keeps the advantage.

Introducing Curinos One

Curinos One is how Curinos does decision intelligence. It uses AI to close the gap between data and outcomes – turning the first- and third-party signals into the decisions that are most likely to grow each relationship. Learning feeds directly back into the engine as it observes customer signals, decides optimal actions, acts through existing channels, and continuously improves from what it learns from outcomes (Figure 14).

The unified data layer is Customer360, the platform’s customer spine, built from first- and third-party signals. Layered on top, Curinos’ proprietary benchmark data sharpen the strategy and expand the system’s library of actions for the human in the loop.

Curinos One is a single native platform, not individual products stitched together. Offers, channels and timing are selected by the engine, and models update automatically from every interaction. It learns continuously from hundreds of optimizations per day, every cycle compounding on the last. That’s in marked contrast to traditional approaches, which can take 120-180 days from insight to optimization – only two to three “optimizations” per year!

Critically, governance and compliance are built into how Curinos One operates, not bolted on afterward. The operating principle is decentralized optimization with central governance: teams act on recommendations within a shared framework of data, KPIs and oversight. Human judgment stays in the loop by design — the platform captures the decision that a bank actually makes, including where it overrides the recommendation. If the pricing engine suggests renewing 12-month CDs at 315 bps and the committee settles on 325, the platform records 325 as the decision and measures the outcome against it. Automated actions are logged and evaluated against control groups, and every recommendation carries a confidence level and a rationale a user can drill into — so a reviewer, or a regulator, can see not just what was decided, but why.

What does this ask of a bank? Less than the scale of the ambition suggests. Curinos One isn’t a rip-and-replace proposition. It acts through the channels and systems the bank already operates, and it doesn’t require a completed data transformation to initiate. A deployment starts with a single program with a defined outcome and an owner accountable for it – whether balance growth, onboarding, CD renewal – along with the corresponding customer data and matched control group, so the value is measured on an incremental basis. First programs are typically live and reading results in less than a quarter.

Curinos One is built around five outcomes that span the full customer relationship. Capture finds high-quality new customers, and Compound expands balances and relationships with existing ones. Both are available today, with the results (below) to show for them.

Three further outcomes – currently in development – extend the same decision layer across the rest of the relationship: Conserve, retaining balances and relationships; Commit, deepening primacy so the bank becomes the customer’s main financial home; and Calibrate, managing the portfolio holistically against rate moves, competitor actions and macro conditions.

Together, these outcomes describe where decision intelligence is headed — one engine, applied across the entire customer lifecycle.

What the Evidence Shows

Curinos One deployments quantify what decision intelligence produces at scale across two program types.

Acquisition: Curinos One Capture

This acquisition case study is from a regional bank with ~$60B in assets that sought to capture new-to-bank households through new checking relationships. It activated programs in the email and banker channels that focused on core value propositions rather than promotional rates. Results below are measured against a matched control group over four months (Figure 15).

• Prospect targeting models identified more receptive, higher-value prospects: The targeted group response rates ran 200% higher than organic acquisition – evidence the model found people likely to act, not just a larger list to mail.
• Those relationships carried real weight: Mailed prospects held an average total deposit balance of $13,800 at 120 days, 44% higher than the $9,600 average for the control group – evidence the model wasn’t just filling accounts, it was filling them with customers who brought and retained meaningful deposits.
• The relationships were sticky: Retention at 120 days was 82%, right in line with organic acquisition, without requiring attractive promotional rates to maintain the balances.

Portfolio: Curinos One Compound

This portfolio case study is from a digital-first bank with over $100B in assets that deployed the platform across three programs: deposit augmentation, new account onboarding, and CD renewal (Figure 16).

At each decision point, the engine evaluated individual customer behavior, value potential, and responsiveness, selecting the action most likely to drive the desired deposit outcome. With no broad campaigns and no promotional rates. Results:

• 10x ROI on platform investment in year one
• $1.6 billion in new-to-bank balances generated since launch
• Total cost of funds: 20 basis points — 75% less than the bank’s standard cost of funds through acquisition
• All results achieved without promotional rates or incentives

Conclusion: The Moment Requires a Different Answer

The evidence is clear and consistent. Portfolio management is just as valuable as acquisition, if not more so, but it receives less investment. Relationship depth predicts retention better than rate, but most banks optimize for rate. The customers who look least valuable in year one generate the majority of growth in year five, but most institutions stop investing in them after the account is opened. The digital shift is permanent, but digital acquisition is producing lower-value customers at scale, and no institution can outspend its way out of that math.

None of these challenges require the industry to stop acquiring customers, stop pricing competitively, or stop investing in digital. It requires that it shifts from functional optimization to customer-centered decision-making. It means moving from asking what each product team needs to asking what each customer relationship requires — and having the data, the decision layer, and the organizational alignment to act on it, at scale. And that in turn means:

• Identifying relationship value earlier
• Coordinating decisions across silos
• Moving from static campaigns to adaptive decision systems
• Aligning marketing, product, pricing and servicing around shared customer outcomes
• Continuously learning from those outcomes

The banks that will define retail growth over the next five years are making this shift now. Not because they have better products or bigger marketing budgets, but because they have the visibility to see their customers clearly, the intelligence to act on what they see and the discipline to measure what actually matters: not accounts opened, but relationships built.

The true growth opportunity isn’t at the top of the funnel; most of it’s already in the book. The question is whether an institution can see it.

Go Deeper

This paper covers the structural picture. These Curinos Perspectives go further into the detail.

Decision Intelligence is Redefining Retail Banking — 5 Takeaways

Five takeaways from a live institutional discussion on moving from insight to action across the customer lifecycle.

Achieving Profitable Deposit Growth in the Age of AI

Why decision intelligence is becoming banking’s next competitive battleground — and what separates the leaders from the rest.

Decision Intelligence is Not Agentic AI – and Any Confusion Will Cost Banks Billions

Most banks are investing heavily in agentic AI while underinvesting in the decision layer that governs what those agents execute. The fix isn’t about choosing one over the other, but in building both layers and understanding which one governs the other. Decision Intelligence is Not Agentic AI – Curinos

Two CEOs Opine on What’s Troubling Bank CEOs

The strategic pressures at the top of the house — and why the answers point back to decision capability, not spend.

When you’re ready to see how this applies to your institution: sales@curinos.com