The forecast confidence gap: why bank CFOs don’t fully trust their own numbers, and what it takes to close the gap

Bedford Consulting | Banking Insight Series

There is a particular kind of discomfort that most bank CFOs know well but rarely discuss openly. It usually happens before a board presentation, or during the quiet moment before an earnings call, when the analyst questions are to be fielded and the slides finalised. The numbers are defensible. The narrative hold together. But somewhere, in a place that doesn’t show up in the footnotes, sits a nagging awareness: I’m not entirely sure these numbers tell the full story.

This isn’t imposter syndrome; it’s pattern recognition. The CFO has been around long enough to know where the assumptions are thin, where the models were last updated six months ago, where internal targets and performance management outcomes are shaping the numbers, not the macros or business reality. the reconciliation between finance and risk was done by hand on a Thursday afternoon. The forecast looks solid, it’s likely materially there. But how much noise is in the signal is the question, where the forecast will actually land is often less than clear.

That gap between the confidence a forecast projects and the confidence the CFO actually feels is what we call the forecast confidence gap. And in banking, where the numbers inform regulatory submissions, capital allocation, dividend policy, and strategic direction, it is one of the most consequential problems nobody talks about.

Where confidence breaks down

The forecast confidence gap doesn’t originate in bad maths. Bank finance teams are full of talented people doing careful work. The gap opens up in the spaces between teams, between models, and between the assumptions that different parts of the organisation are working from.

Consider a revenue forecast for a retail banking division. The finance team builds it from product-level drivers: loan volumes, deposit rates, fee income, NIM assumptions. Each input has been debated and agreed. But the forecast depends on assumptions about customer behaviour that the marketing team owns, cost trajectories that operations controls, and credit loss expectations that sit with risk. If those upstream assumptions change and nobody updates the downstream model, the revenue forecast becomes a snapshot of a world that no longer exists.

This kind of assumption drift is remarkably common. It happens not because people are negligent, but because the planning process wasn’t designed to keep assumptions synchronised across functions. The finance team locked their version of reality in a spreadsheet. The risk team locked a different version in another spreadsheet. By the time the CFO stands up to present, the two versions have quietly diverged.

PwC’s analysis of recent bank earnings calls captures the external consequence of this problem. Analysts are increasingly focused on whether management teams can deliver clear, evidence-backed outlooks. Roughly 30% of analyst questions now probe the quality and resilience of revenue growth. Another 25% test whether the cost base is genuinely flexible. These aren’t questions about whether the bank’s spreadsheets add up. They’re questions about whether the story behind the spreadsheets is coherent and durable.

A CFO who can’t trace a revenue forecast back to its operational assumptions, or who can’t show how a cost plan connects to a strategic initiative, will struggle to answer those questions with the conviction the market now demands.

The five sources of forecast doubt

In our work with banking clients, we’ve found that forecast confidence tends to erode from five recurring sources. None of them is exotic. All of them are fixable. But they rarely get addressed because they sit in the seams between functions rather than within any single team’s remit.

Assumption drift. An assumption agreed in January goes stale by April. The interest rate outlook shifts. Customer acquisition costs change. But the model still reflects the January view because nobody triggered an update. We’ve seen cases where forecasts feeding executive decisions were built on (legacy information) different time horizons, with nobody aware of the mismatch.

Definition mismatch. The word “capacity” means one thing in finance and something subtly different in operations. “Fully loaded cost” includes different components depending on who built the model. These definitional gaps are invisible until a decision goes wrong and someone traces the problem back to a cell in a spreadsheet where two teams were talking past each other.

Orphaned models. A model built for a specific purpose two years ago is still being used to inform decisions, but the person who built it has moved on. Nobody else fully understands its logic. Nobody is updating its inputs. It produces numbers, and those numbers get used, but the governance and understanding of what the outputs represent around them has quietly evaporated.

Manual execution reconciliation. The Empyrean Bank Risk and Performance Survey found that 35% of financial institutions identified budgeting and planning tools as their biggest FP&A challenge. In practice, this often means that the link between two models is a person with a spreadsheet. That person is a single point of failure. When they’re on holiday, the reconciliation doesn’t happen, or someone steps in not fully understanding the process. When they’re under pressure, it happens quickly and without full scrutiny.

Scenario avoidance. Scenario plans exist on paper, but they were never designed to be run in parallel or compared side by side. The base case gets all the attention. The upside and downside cases get built off that structure, completed once and filed. Nobody revisits whether the structure or assumptions in the alternative cases reflect the risks or opportunities that actually matter. The forecast looks confident because it only presents one version of the future.

What the market is really asking for

The external pressure on forecast quality is intensifying, and it’s worth understanding why.

A decade ago, analysts evaluated bank forecasts primarily on accuracy: did the bank hit its numbers? Today, the conversation has moved upstream. Analysts want to understand the machinery that produces the numbers. They want to see that the CFO can explain how a change in one assumption cascades through the P&L. They want evidence that management has tested multiple scenarios and can articulate the trade-offs between them. They want to know that the forecast isn’t just a number but a system of connected, governed, traceable assumptions.

PwC puts it plainly: the market is not asking CFOs to predict the future. It’s asking them to demonstrate control over the present, and to articulate a credible path forward. The CFOs who will stand out are those who can connect growth, discipline, and investment into a single, coherent story grounded in realism rather than aspiration.

For a bank CFO, this changes what a good forecast looks like. A good forecast is no longer one that proves to be right. It’s one whose assumptions are visible, whose logic is traceable, whose scenarios have been genuinely explored, and whose governance is strong enough that the CFO can stand behind it without that nagging feeling in the minutes before the board meeting.

Closing the gap

Closing the forecast confidence gap is not a technology project, although technology plays an important part. It’s a governance project with technology as its backbone.

The first step is making assumptions visible. Every forecast rests on a set of assumptions, but in most banks those assumptions are buried in individual spreadsheets, known only to the people who built the models. Surfacing them into a shared environment where they can be reviewed, challenged, and updated is the single most impactful thing a CFO can do to improve forecast confidence. When assumptions are visible, they can be governed. When they’re hidden, they can only be hoped for.

The second step is connecting the models. A revenue forecast that lives in isolation from the cost plan, the capital model, and the stress test is a forecast that can’t be fully trusted, because nobody can see whether its assumptions are consistent with the rest of the bank’s view of the world. Connected planning, on a platform like Anaplan, makes those connections structural rather than manual. When the NIM assumption changes, the downstream impact on capital ratios, product profitability, and headcount is visible immediately, in the same environment, to every team that needs to see it.

The third step is investing in scenario discipline. This means more than building a bear case once a year. It means designing scenarios that can be run frequently, compared systematically, and used as genuine decision-making tools. One leading bank improved its overall forecast accuracy from over 12% deviation to 5% and below after moving to a connected planning platform. That improvement didn’t come from better maths. It came from better governance of assumptions and a planning architecture that kept models aligned.

The fourth step, and the one most often neglected, is model ownership. Every model that informs a forecast needs a named owner who is responsible for its assumptions, its inputs, and its connections to other models. Without that ownership, models drift. Definitions diverge. Workarounds become business-as-usual. And the forecast confidence gap widens quietly until something goes visibly wrong.

The Bedford perspective

We’ve spent this three-part series making a connected argument. In the first article [LINK], we explored why connected planning is the real competitive advantage for bank CFOs. In the second [LINK], we showed how Basel III Endgame exposes the dangers of disconnected capital planning. Here, we’ve arrived at the human consequence of those structural problems: a CFO who presents numbers they’re not fully confident in, because the planning infrastructure doesn’t give them reason to be.

At Bedford, this is the problem we exist to solve. We help banks see what others can’t: the hidden risks, missed value, and blind spots buried within planning processes and models. We do this by combining deep governance expertise with Anaplan’s connected planning platform, designing the architecture that turns disconnected models into a single, governed, traceable view of the bank’s future.

The result is a CFO who can stand up in a board meeting, or sit down with an analyst, and know that the numbers they’re presenting are built on foundations they can explain, defend, and trust. That’s not a technology outcome. It’s a confidence outcome. And it’s where better decisions begin.

Continue the conversation

If the themes across this series have resonated, we’d welcome you to join us on 12th May in London for a breakfast roundtable with banking leaders, Bedford’s financial services team, and Anaplan. The roundtable will explore how leading banks are closing the forecast confidence gap by connecting finance, risk, treasury, and operations on a single planning platform, and turning forecasts into governed, cross-functional decisions that leadership can stand behind.

Register for the Banking Roundtable →

Bedford Consulting is Anaplan’s longest-standing EMEA Partner of the Year, with deep expertise in financial services planning. To discuss any of the themes in this series, contact our banking team at info@bedfordconsulting.com.