What it means
Every business runs on models, even when nobody calls them that. A budget spreadsheet, a pricing calculator, a credit scorecard and a three-year cash flow forecast are all models: structured sets of assumptions that turn inputs into numbers people act on.
Model validation is the discipline of making sure those numbers deserve the trust placed in them. Validation matters because model errors are quiet and expensive.
A single mistyped formula in a valuation model can misprice an acquisition by millions, and nobody notices until the deal closes and the promised returns fail to appear. Regulated lenders and insurers are required to validate their models formally, but the same logic applies to a founder deciding whether to hire ten people next quarter.
In practice validation works on several layers. Mechanical checks confirm that totals tie out, that formulas are consistent across a row, and that no hard-coded number is buried inside a calculation.
Logic checks ask whether the structure makes sense, for example whether a revenue line correctly depends on both price and volume rather than growing by an unexplained percentage. The most revealing layer is behavioural testing.
Stress testing pushes inputs to extreme values to see whether the model breaks or produces nonsense, and back-testing compares past model predictions against what actually happened. A forecasting model that consistently overshoots by 15% is not necessarily useless, but the bias needs to be known and corrected.
A common variant is independent validation, where someone who did not build the model reviews it. This matters because builders develop blind spots: they know what the model is supposed to do and unconsciously read that intention into what it actually does.
Larger organisations formalise this through a model risk policy that grades models by importance and sets how often each must be revalidated.
In practice
Real-world examples.
Example
A logistics company builds a fuel-cost model to set customer surcharges. Validation reveals the model uses last year's average fuel price as a fixed input rather than a live variable, so every surcharge has been understated since prices rose. The fix takes an afternoon; the twelve months of underrecovered cost totalled $340,000.
Example
A software business validates the churn model behind its revenue forecast by back-testing it against the previous eight quarters. The model tracks well for small accounts but badly underestimates churn among enterprise customers, prompting the finance team to split the forecast into two segments with separate assumptions.
Example
Before a private equity investment committee meets, an analyst who did not build the leveraged buyout model reviews it independently. She finds the exit multiple is applied to the wrong year's earnings, which had inflated the projected return by roughly a fifth.
Think of it
“Model validation is checking that your model actually works correctly-independent review.
Formula
Calculation
A standard back-testing measure is Mean Absolute Percentage Error (MAPE): MAPE = average of |Actual - Forecast| / Actual, expressed as a percentage.
A retailer validates its monthly revenue model over three months. In January the model forecast $900,000 against actual revenue of $1,000,000, an absolute error of $100,000, which is $100,000 / $1,000,000 = 10%. In February it forecast $1,260,000 against actual revenue of $1,200,000, an error of $60,000, which is $60,000 / $1,200,000 = 5%. In March it forecast $960,000 against actual revenue of $800,000, an error of $160,000, which is $160,000 / $800,000 = 20%.
MAPE = (10% + 5% + 20%) / 3 = 35% / 3 = 11.7%. The validator concludes the model is accurate enough for headcount planning, where a 12% swing is tolerable, but not accurate enough to set covenant-tight cash targets on its own.Case study
Seen in the real world.
In this illustrative example, Harborline Manufacturing, a fictional mid-sized components maker, priced a five-year supply contract using a margin model built by a departing analyst. The model produced a healthy 22% gross margin, and the sales team signed the deal on that basis.
Nine months in, the contract was losing money. A newly hired financial controller ran a proper validation and found two problems: the raw material input was linked to a summary cell that had been sorted, breaking the link, and the overhead allocation excluded warehousing entirely. Corrected, the true margin was 9%, well below the 15% floor the board had set for long-term contracts.
Harborline could not reopen the contract, but it introduced a rule that any model supporting a commitment over $1,000,000 must be reviewed by a second person before signature. Over the following two years that rule caught four material errors, none of which reached a customer.
Watch out
Common mistakes.
- Treating validation as a one-off event at the point the model is built, when assumptions drift and the model needs revalidating as conditions change.
- Letting the model's author validate their own work, which reliably misses the errors that come from misunderstanding the problem rather than mistyping a formula.
- Confusing a clean-looking output with a correct one, since a well-formatted spreadsheet full of wrong numbers looks exactly like a well-formatted spreadsheet full of right ones.
Questions
People also ask.
How long should model validation take?
For a simple operating model a focused half-day of checks is usually enough, while a valuation or credit model driving major decisions warrants several days and a written record of what was tested.
Does validation prove the forecast will be right?
No, it proves the model calculates its assumptions correctly and behaves sensibly under stress; if the assumptions are wrong the output will still be wrong.
Who should own model validation in a small company?
Usually the finance lead sets the standard, but the practical answer is any competent colleague who did not build the model and is willing to question its logic line by line.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%Related