How to Review Someone Else's Financial Model

Reviewing a model you did not build is detective work. You inherit someone else's logic, conventions, and shortcuts, and you have to decide whether to trust the output. A structured pass, from broad structure down to individual formulas, finds problems faster than poking at cells at random.

Start With the Structure, Not the Numbers

Before checking any formula, understand the layout. Identify the inputs sheet, the calculation engine, and the outputs. A model with clear separation is already a good sign.

Walk the tabs left to right and read the sheet names. Note where assumptions live and how the model is meant to flow.

1. List every sheet and what it appears to do
2. Find where the headline outputs come from
3. Locate the inputs that drive those outputs
4. Sketch the flow from inputs to outputs in your head or on paper

Run a Sanity Check on the Outputs

Before going deep, ask whether the top-line numbers are even plausible. Does margin trend sensibly? Does the balance sheet balance? Does cash never go impossibly negative?

Implausible outputs point you straight to the part of the model worth scrutinizing first.

Trace the Key Formulas One Layer at a Time

Pick the most important output cell and walk back through its precedents. Understanding what feeds a number, and what feeds those inputs in turn, is how you actually learn a model.

Excel's Trace Precedents under Formula Auditing draws arrows, but they clutter quickly on a dense sheet and clear the moment you edit. Walking one layer at a time keeps the chain readable.

1. Select a headline output cell
2. Trace its direct precedents and note where they sit
3. Step into each precedent and trace its inputs
4. Continue until you reach the underlying assumptions

Hunt for the Usual Weak Spots

Once you understand the flow, target the places errors hide: hardcodes inside formulas, broken row patterns, and links to other workbooks that may be stale.

Check that subtotals sum the right ranges and that sign conventions are consistent across sections.

• Constants typed into formulas instead of referenced
• A single cell in a row that breaks the copied pattern
• External links that could be reading outdated files
• Error-suppressed formulas that might be masking real failures

Pressure-Test the Assumptions

A model can be flawlessly built and still wrong if its assumptions are unrealistic. Flex the key inputs and watch whether the outputs move sensibly.

If a 10% change in a driver barely moves the result, either the model is insensitive to it or the link is broken. Both are worth understanding.