A financial model validation SOP checks that your startup's spreadsheet tells the truth: the formulas work, the assumptions come from evidence and the outputs survive a skeptical investor's questions. It takes one to three weeks, depending on how much evidence you still need to collect. Investors forgive a forecast that turns out wrong; they don't forgive one that was built wrong. You finish with an assumption register, a model that balances and a list of the three numbers your company's survival depends on.

What is the financial model validation SOP and when should you use it?

A financial model is the spreadsheet that turns assumptions (prices, conversion rates, salaries, churn) into a forecast of revenue, costs, cash and, for later-stage companies, a full set of statements. Validation tests two layers: mechanics (do formulas calculate what they claim?) and assumptions (is each input grounded in something real?).

Neighbouring SOPs go deeper on single parts. Financial scenario planning builds best, base and worst cases and decision triggers. Unit economics validation tests margin per customer. This SOP checks the whole model as a tool before you trust or share it.

Use it:

Before you send the model to investors, a bank or a grant body.
When the model drives hiring or spending decisions and nobody has checked it in months.
After a co-founder, advisor or AI tool built most of it, so you understand every line you will defend.
When actuals keep missing the forecast and you can't say why.

What do you need before you start?

The model, with a copy you can break.
Your actuals: bank statements and an export from your accounting software for the last 3 to 12 months.
Evidence for inputs: interview notes, landing page conversion data, quotes from suppliers, signed contracts, public benchmarks.
A reviewer who builds models for a living: an accountant, a finance-minded angel or a founder one round ahead of you.
A modelling standard to check structure against, such as the FAST Standard (Flexible, Appropriate, Structured, Transparent) or the ICAEW Financial Modelling Code.

How do you run it, step by step?

1. Separate inputs, calculations and outputs

Move every assumption to one inputs sheet, coloured so it stands out. Calculations reference inputs and never contain typed numbers. Outputs (charts, summary, investor tables) only reference calculations. A number typed into the middle of a formula is the classic hidden error, and it survives every edit.

2. Build an assumption register

List every input with its value, unit, source and an evidence grade: A for your own actuals or signed contracts, B for tests and quotes, C for public benchmarks, D for gut feel. Anything graded D that sits in the top drivers goes on your to-test list.

3. Add error checks that must read zero

Add check rows: cash in the cash flow equals cash on the balance sheet, assets equal liabilities plus equity, total revenue equals the sum of segments, headcount never goes negative. Sum all checks into one cell on the summary sheet. Raymond Panko's research at the University of Hawaii found errors in most spreadsheets that field audits examined, so assume yours has some.

4. Trace the top outputs by hand

Pick the five numbers investors look at first: month-12 revenue, month-24 revenue, peak burn, runway and gross margin. Use Trace Precedents to follow each back to inputs, and recompute one month by hand or on a calculator.

5. Replace top-down revenue with bottom-up

"1% of a large market" is a top-down guess. Rebuild revenue from the actions that create it: leads per channel, conversion at each stage, price and churn. Check growth against known reference points: Neeraj Agrawal of Battery Ventures described T2D3 (triple, triple, double, double, double) as a fast path for SaaS revenue, so a plan faster than that needs strong evidence.

6. Compare forecast with actuals

Put the last three months of forecast next to real numbers. Gaps over 15% on revenue or costs point at a broken assumption; fix the assumption before you touch the outputs.

7. Stress the top drivers and get an outside review

Change each input by plus and minus 20% and rank them by how much runway moves. The top three are your company's real risks. Then hand the model and the register to your reviewer with one question: "Where would you stop believing this?"

Template you can copy

Input	Value	Unit	Source	Evidence grade (A–D)	Next test to improve evidence
Visitor to trial conversion	3.5%	%	Landing page, 1,900 visits	B	Run paid ads test
Trial to paid conversion	18%	%	Guess	D	Concierge pilot with 10 clinics
Monthly price	€149	€ per clinic	14 interviews, 3 pre-orders	B	Pricing model test
Monthly churn	2%	%	Public SaaS benchmark	C	Track first 30 customers
Salary, developer	€5,200	€ per month, employer cost	Two job offers	A

Error check block (each cell must read 0): Balance sheet balances · Cash matches cash flow · Revenue equals sum of segments · No negative headcount · No hard-coded numbers in calculation sheets (count) · Total of all checks.

Worked example: a booking tool for dental clinics

This is an illustrative case. Two founders prepare a seed round for software that dental clinics use to fill cancelled appointments. Their 36-month model shows €2.1 million in annual recurring revenue by month 36.

The register lists 41 inputs. Only 12 (29%) have A or B evidence, and the revenue line starts from "1% of 140,000 clinics in Europe". Error checks show the balance sheet off by €14,200 from month 19, caused by a VAT row that counts sales tax as revenue. A formula search finds 23 hard-coded numbers inside calculations.

They rebuild revenue bottom-up: two salespeople, 18 demos each per month, a 25% close rate from their pilot calls and 2% monthly churn. Month-36 recurring revenue falls to €1.3 million. The stress test ranks churn first: moving it from 2% to 3% cuts runway from 22 to 17 months. Two weeks of evidence work raise A and B coverage to 34 of 41 inputs (83%). The reviewer, an angel investor, asks nine questions, and the founders answer each one from the register.

What should you measure, and when do you stop?

Metric	Pass signal	Warning signal
Inputs backed by evidence	80%+ graded A or B, including all top-3 drivers	Top drivers graded D
Error checks passing	Sum of checks reads 0 in every month	Checks deleted or "fixed" with plugs
Forecast vs actual gap	Revenue and costs within 15% for the last 3 months	Gaps over 30%, or no actuals tracked
Runway swing from top 3 drivers	You know the three and have a test running on each	Nobody can name them
Reviewer questions answered	Every question answered from the register	Answers start with "we assumed"

Stop when every row passes and the reviewer's last question needs no change to the model. Re-run steps 3 and 6 every month once you have revenue.

Which mistakes should you avoid?

Polishing charts before checking formulas. Investors trace numbers; pretty outputs from broken logic hurt trust more than plain ones.
Mixing VAT or sales tax into revenue. It inflates sales and breaks cash timing.
Using one churn and one conversion rate for every channel. Paid ads and referrals convert differently.
Plugging the balance sheet. A "misc" row that forces it to balance hides the real error.
Building 10-year forecasts at pre-seed. Three years by month, with the first year evidence-backed, is plenty.

How can AI speed this up?

Export each sheet as CSV or paste formulas into an AI assistant and ask it to find hard-coded numbers, circular logic, inconsistent units and rows that don't add up. Ask it to draft the assumption register from your inputs sheet and suggest which public sources could back each one.

The human check: AI reads the formulas you give it and nothing else, and it will invent a benchmark if you ask for one without a source. Confirm every benchmark at the original publication. The free F/MS business plan generator turns a validated model into the written plan banks and grant bodies ask for.

Where does this SOP fit in your validation plan?

This SOP opens the Financial Validation SOPs:

Inputs come from earlier work: unit economics validation, the break-even analysis SOP and cash flow validation.
Next, build cases around the validated model with financial scenario planning.
Then take it to investors with the fundraising validation SOP and the valuation methodology SOP.

The F/MS investor playbook module shows how investors read a model and which numbers they check first.

FAQ

How do you check a startup financial model for errors?

Separate inputs from formulas, add check rows that must read zero (balance sheet, cash, totals), trace the top five outputs back to inputs and recompute one month by hand.

How accurate does a startup financial model need to be?

Nobody expects forecasts to be right years out. Investors expect correct mechanics, evidence for the top drivers and recent forecasts within about 15% of actuals.

Should I use a financial model template?

A template saves time on structure, but you still need to validate every assumption. Check that the template separates inputs, calculations and outputs before you fill it in.

Who should review my financial model?

Someone who builds or reads models for a living: an accountant, a finance-minded angel investor or a founder who raised a round recently. Give them the assumption register along with the spreadsheet.

Financial Model Validation SOP