Field notes from model audits

10 financial model mistakes investors catch first

None of these are exotic. They are small formula slips that move net income by six figures, and an analyst on the other side of the table will find them in the first hour. Here is each one, how to spot it, and the fix.

1. Tax charged on losses

The tax line is written as EBITDA × tax rate. In a loss-making month that produces negative tax — a refund the government never sends. Across two years of losses this quietly “earns” the company tens of thousands.

Fix: MAX(0, taxable profit) × rate, and carry losses forward so the first profitable months are not taxed either.

2. Cash moves with EBITDA, not net income

The cash row adds EBITDA each month. Tax, interest and anything below EBITDA never leave the bank, so runway looks longer than it is.

Fix: cash(t) = cash(t−1) + net income + equity raised − debt repaid, plus working-capital changes if you model them.

3. The annual total skips a month

Year 1 is SUM(B:M), year 2 is SUM(N:X) — eleven months. It happens when someone inserts a column or copies a formula one cell short. Revenue looks fine; December’s costs simply vanish.

Fix: build annual totals with SUMIF on a “year” row, so they cannot drift when columns move.

4. LTV that ignores churn and margin

LTV = price × 12 assumes every customer stays exactly one year and costs nothing to serve. Depending on churn this understates or overstates LTV several times over — and LTV/CAC is the number investors look at first.

Fix: LTV = ARPA × gross margin ÷ monthly churn. Show the payback period next to it.

5. Churn applied to this month’s new customers

Churn is calculated on (last month’s customers + this month’s sign-ups). Customers who signed up today cannot have cancelled already, so growth is understated every single month.

Fix: churned(t) = customers(t−1) × churn rate.

6. Growth that starts one month early

new × (1 + g)^month grows month 1 too. The whole curve shifts by a month, and by month 36 the error compounds into hundreds of customers.

Fix: exponent (month − 1), or anchor growth to a start-month input.

7. Price increase off by one

An annual price rise meant to start in month 13 is coded as month ≥ 12. One extra month of higher revenue each year — small, but exactly the kind of thing a reviewer reads as carelessness.

Fix: drive price from a “year” row: price × (1 + increase)^(year − 1).

8. Rates typed inside formulas

=revenue × 0.02 while the assumptions sheet says the payment fee is 2.9%. Change the assumption and nothing happens. Every typed number is a place where the model and the story disagree.

Fix: every rate links to one assumptions cell. Search formulas for digits other than 0, 1 and 12.

9. Year 2 calculated differently from year 1

Marketing is new customers × CAC for twelve months, then switches to all customers × CAC ÷ 12. Someone “simplified” year 2 and nobody noticed because the numbers looked plausible.

Fix: one formula per row, copied across all months. If a row needs a change mid-way, use an explicit switch input.

10. Formulas that are right by accident

=office × 12 / 12 + office × 0 returns the correct number today and breaks the moment anyone edits it. These are the leftovers of old fixes.

Fix: rewrite to the plain link. If you cannot explain a formula in one sentence, it is hiding something.

Check your own model in 30 seconds

Our free checker flags mistakes 3, 8 and 9 automatically — inconsistent formulas, typed constants and short sums — right in your browser. The file is never uploaded.

Run the free checker See a full audit

Every example above comes from a model like this sample audit, where ten such mistakes hid $174k of losses and halved the reported LTV/CAC.