What is an assumptions log for a business plan financial model?

A dated log — typically a spreadsheet tab or simple table — that traces every material number in the financial model back to a named, explicit assumption, such as a conversion rate, an average deal size, or a churn figure, along with the evidence behind it and a record of when and why it changes.

How is an assumptions log different from just updating the financial model?

Updating the model changes the output number going forward. The assumptions log preserves the history of why it changed — what the assumption used to be, what it is now, when it changed, and what evidence triggered the change. The model shows the current forecast; the log shows the reasoning trail behind every version of it.

How many assumptions should I track in the log?

Focus on the assumptions that materially drive your key numbers — revenue growth, unit economics, headcount cost, and cash runway. A handful of well-documented, high-impact assumptions is more useful at review than an exhaustive log of every minor input.

Do endorsing bodies ask to see an assumptions log directly?

Not typically as a named, mandatory document — but the underlying discipline is exactly what a strong answer at a contact-point meeting or annual review is built from. Having it ready means you can answer variance questions with dated specifics instead of reconstructing your reasoning on the spot.

What should I do when an assumption in my log turns out to be wrong?

Add a new dated entry explaining what the actual figure turned out to be, what evidence revealed it, and what you changed in response — in the model, the plan, or the operating strategy. Don't delete or quietly edit the original entry; the record of the original assumption being wrong is itself valuable evidence of a functioning process.

The assumptions log: your defence against forecast variance

Every number in a financial model came from somewhere. A monthly revenue figure is really a chain of smaller claims — this many leads, at this conversion rate, at this price point, minus this much churn. The problem is that once those claims get compressed into a single line on a spreadsheet, the reasoning behind them disappears. Six months later, when the actual number comes in at half the forecast, the founder is often left trying to reconstruct, from memory, what they were even assuming in the first place.

Why the number alone is never enough

Take a common example: a plan forecasts £18,000 in month 6 revenue. Eighteen months later, month 6 actually landed at £9,000. On its own, that's just a shortfall — a red number in a spreadsheet, and precisely the kind of gap founders worry an endorsing body will treat as evidence of failure.

But £18,000 was never really "the forecast." It was the output of a specific, if unstated, chain of reasoning: something like 40 leads a month, at a 12% conversion rate, at a £450 average contract value. Once you make that chain explicit, the shortfall stops being one vague miss and becomes a specific, traceable question: which link in the chain broke? Maybe leads came in as planned but conversion was 6% instead of 12%. Maybe conversion held but average deal size came in lower because customers bought a smaller package than assumed. Those are two completely different businesses to be running, and they call for different responses — but you can only tell them apart if the original assumption was written down.

What the log actually looks like

You don't need special software. A spreadsheet tab or a simple table is enough. Each row traces one material assumption: the number it drives, the assumption itself, the evidence behind it at the time, and a running history of changes.

Date	Driver	Assumption	Evidence	Status
2026-01-15	Month 6 revenue	40 leads/mo × 12% conversion × £450 ACV	Conversion rate based on 3 discovery-call pilots run pre-launch	Superseded
2026-07-10	Month 6 revenue	40 leads/mo × 6% conversion × £450 ACV	Actual conversion across 68 qualified leads, months 1–6; pilots overstated pre-launch enthusiasm	Current
2026-07-10	Response	Repriced entry tier from £450 to £320 ACV to reduce sales friction; targeting 10% conversion at new price	First 12 leads post-repricing convert at 11%	Monitoring
2026-09-02	Headcount cost	1 FTE sales hire at month 9, £42k loaded cost	Deferred to month 11; runway model showed month-9 hire left under 18 months of cash at current burn	Current

Four things matter in every row: the specific number the assumption drives, what the assumption actually was (not a vague "revenue will grow"), the evidence behind it at the time it was made, and what happened when it changed. The "evidence" column is what separates a real assumptions log from a wish list — "based on 3 discovery-call pilots" is falsifiable and specific; "we believe demand will be strong" is neither.

Why this beats reforecasting silently

The instinctive response to a wrong assumption is to just fix the model and move on — update the conversion rate, let the forecast recalculate, keep going. That's necessary, but it's not sufficient. Reforecasting silently changes the output; it doesn't preserve the reasoning trail that explains the change.

A founder who silently reforecasts is hoping nobody asks what happened. A founder with a dated assumptions log is prepared for the question before it's asked.

— Editorial framing

The distinction matters most in exactly the moment it's tested: a contact-point meeting or annual review where the endorsing body notices your current numbers don't match your original plan. See what the annual review actually checks — the review is testing whether the underlying business rationale still holds, not whether the original numbers were exactly right. A founder with a log can answer immediately, with dates and evidence: "conversion came in at half our pilot estimate, we identified that by month 6, repriced by month 7, and early data on the new price point supports the revised assumption." A founder without one is reconstructing the same story on the spot, and it reads very differently even if the underlying facts are identical.

This is also the raw material for writing the variance narrative that some founders prepare for a 12-month update — the log is where that narrative's specifics come from, rather than being invented at the point the narrative is needed.

Know exactly where your application stands.

Get your free AI assessment in 90 seconds.

Get your assessment

What this proves that the model alone can't

An endorsing body reviewing a business at 12 or 24 months is not really trying to verify that a founder is a good fortune-teller. Nobody expects an eighteen-month-old forecast, written before a single customer existed, to have been exactly right. What they are trying to verify — per the financial-model red flags that assessors are trained to spot — is whether the founder understands their own business well enough to know what's driving it and to respond intelligently when reality diverges from the plan.

An assumptions log is direct, dated evidence of exactly that competency. It shows:

The founder made specific, falsifiable claims rather than vague aspirations.
The founder tracked those claims against reality on an ongoing basis, not just at review time.
When a claim turned out to be wrong, the founder identified which one, understood why, and changed something in response.
The response is itself evidenced — a new assumption, a new data point, a new decision — not just a hopeful adjustment.

That's a much stronger position than defending the original £18,000 figure, and a much stronger position than quietly hoping nobody notices it didn't happen.

Getting started without overengineering it

Don't try to log every input in your model — that turns into its own burden and defeats the purpose. Focus on a handful of high-impact assumptions: the core revenue driver chain (leads, conversion, price), your biggest cost assumption (usually headcount), and your cash runway assumption. Update the log at the same cadence you already refresh your business plan — see build a living business plan, not a museum piece for the version-history habit this log slots directly into. Most quarters, updating it is a matter of minutes once the initial version exists.

Sources and further reading

Key takeaways

A single forecast number hides a chain of assumptions — an assumptions log makes that chain explicit, dated, and traceable.
Each log entry should name the assumption, the evidence behind it, and what changed when it turned out to be wrong.
Reforecasting silently changes the output but destroys the reasoning trail; the log preserves both.
At review, a founder with a dated log answers variance questions with specifics instead of reconstructing a story under pressure.
The log is direct evidence of the founder competency endorsing bodies are actually testing: understanding your own business well enough to know what moved and why.