Startup financial projections without the fiction

How to build startup financial projections from real drivers: revenue, costs, cash flow and runway, plus the assumption mistakes that undermine credibility.

Quick answer

Build startup projections from drivers rather than from a target: model how customers arrive, what they pay, how long they stay, and what it costs to serve and acquire them. Show the assumptions next to the numbers, keep a monthly cash view for at least eighteen months, and state your runway and the point at which it forces a decision.

Drivers first, totals second

A credible model starts with a small number of drivers and derives everything else. If your revenue line is typed in directly, nobody can interrogate it and neither can you. If it is derived from traffic, conversion, price and retention, a reviewer can challenge one input rather than dismissing the whole sheet.

• Acquisition: where new customers come from and at what volume
• Conversion: the rate from interest to paying customer
• Price: the actual amount, net of discounting
• Retention or repeat rate: how long the revenue lasts
• Cost to serve: the variable cost per customer
• Fixed costs: people, tooling, premises, compliance

Three statements, one story

Even a simple model benefits from separating profit and loss, cash flow, and the balance of what you hold. Profitable months with negative cash are common in early businesses — invoices paid late, stock bought early, annual tooling paid up front. The cash view is the one that determines whether you survive the quarter.

• Profit and loss: revenue and cost recognised in the period
• Cash flow: money actually moving, including timing lags
• Position: cash held, money owed to you, money you owe

Model the assumptions explicitly

Put every assumption in one visible place with a stated basis: observed, benchmarked or guessed. When an investor asks 'why 3% conversion?', the answer 'that is what our landing page has run at over eight weeks' is a different conversation from 'it seemed reasonable'. Both are acceptable at pre-seed; only one of them can be defended.

• Observed — measured in your own data, with the period stated
• Benchmarked — taken from a named external source you can link
• Assumed — a placeholder, flagged as such, with a plan to test it

Scenarios instead of a single line

One projection invites a binary judgement of right or wrong. Three scenarios show how you think. Keep the structure identical and vary only the small set of drivers that genuinely move the outcome — usually acquisition volume, conversion, price and churn.

• Conservative: what happens if the main channel underperforms
• Moderate: the plan you are actually running
• Optimistic: what a channel working well would look like
• For each: runway, hiring implications, and the decision point

Runway is a decision date, not a number

Runway expressed as 'fourteen months' is less useful than 'we must either hit £X monthly recurring revenue by April or reduce burn'. Convert cash into the date by which a decision has to be made, and state what the decision would be. That is what a board or an investor actually wants to see.

Common mistakes that cost credibility

Most early forecasts fail in the same handful of ways. None of them are hard to avoid once you know to look.

• Hockey-stick growth with no mechanism behind it
• Percentage-of-market revenue: '1% of a £5bn market'
• Costs that stay flat while revenue multiplies
• No hiring lag — headcount productive from day one
• Ignoring payment timing, VAT, and annual prepayments
• Prices in the model that do not match the pricing page
• A model nobody can rebuild because the logic is undocumented

Keep it in sync with everything else

The forecast is the document most likely to drift, because it changes most often. Every time pricing, headcount or the acquisition plan moves, the plan and deck built on the old numbers become wrong. Decide where the numbers live authoritatively, and re-check the dependent documents whenever they change.

Home · Features · How it works · Pricing · Compare · For founders · Try free