When Models Are Technically Right but Strategically Wrong

We sometimes take comfort in the idea that if the model is correct, the decision will be as well. In practice, it is rarely that simple.

There is a particular type of modelling problem that does not look like a problem at all. It does not appear in an error check. It does not break a formula. It does not disturb a reconciliation. In fact, the model may be technically flawless. And yet, something essential can still be wrong. Not in the mechanics. But in the framing.

Most financial models answer the question they are given. They do this with admirable discipline. They follow the logic we define. They apply assumptions consistently. They optimise whatever objective we encode. In that sense, they are very obedient. And obedience, while useful, has never been the same thing as judgement.

A model can perfectly evaluate the return of a project while quietly ignoring whether the organisation is actually capable of delivering it. It can optimise capital structure without asking whether the business model itself is resilient. It can calculate long-term value without pausing to consider whether the underlying assumptions are strategically aligned with reality.

Technically correct. Strategically incomplete.

This rarely happens dramatically. There is no red warning sign. No alarm bell. No dramatic pause in the meeting. The model is built with care. The assumptions are discussed. The outputs look sensible.

Everyone nods. And gradually, almost imperceptibly, the conversation shifts.

From: “Is this the right direction?” to: “Is the model correct?”

The second question is easier. It feels measurable. It feels professional. It feels reassuring. And reassurance, as we all know, can be surprisingly persuasive. But it is not necessarily the important question.

Strategic misalignment rarely announces itself as a spreadsheet error. It tends to show up in quieter ways.

Optimism that reflects ambition more than capacity. Growth projections that look elegant on paper but struggle in operations. Risk buffers that exist numerically but not culturally. Scenarios that vary percentages but never behaviour.

None of these break the model. But they can slowly bend the decision. And the bending often happens politely.

Experienced modellers develop a certain sensitivity to these moments. They notice when discussions focus heavily on mechanics but only lightly on context. They see when sensitivity tables are explored thoroughly while alternative strategic paths are barely mentioned. They become cautious when the base case feels emotionally anchored before it has been intellectually challenged.

These are not formula issues. They are framing issues. And framing, as it turns out, is the most strategic part of modelling.

Not the numbers. Not the formatting. Not even the elegance. The question.

There is no perfect safeguard against this. Models cannot correct the question they are given. But we can design them in ways that make thinking harder to avoid.

We can make assumptions visible and challengeable. We can separate operational feasibility from financial attractiveness. We can resist the urge to compress uncertainty into a single narrative too early.

These choices do not guarantee strategic wisdom. But they reduce the risk of elegant misdirection.

In the end, models are extraordinarily good at answering defined questions. Their weakness is not calculation. It is obedience.

They will follow the boundaries we draw. They will optimise the targets we set. They will confirm the logic we encode. And they will do so with impressive consistency.

Which makes the initial framing – the quiet conversation before the spreadsheet opens – the most strategic part of the entire exercise.

That part does not happen in Excel. It happens between people. Sometimes between people who would prefer the spreadsheet to decide for them.

A technically right model can still support a strategically wrong decision.

The spreadsheet may be flawless. The direction may not be. Knowing the difference is part of responsible modelling.

And, occasionally, part of responsible leadership.