Hiring guide
Financial model assumptions: build a register founders can challenge
By CFO Index · Published
A forecast can be perfectly calculated and still rest on a guess nobody owns. An assumptions register makes the important estimates visible before management relies on them. Start with the inputs that could change a hiring, purchasing or funding decision, not a catalogue of every cell in the workbook.
Separate evidence, estimates and targets
A signed price agreement, last quarter's observed conversion rate and next quarter's sales target are different kinds of input. Label them accordingly. Historical evidence can inform a forecast without guaranteeing the future; an approved target does not become the most likely outcome simply because management signed it. Record the population and period behind an observation. A conversion rate from repeat customers is not automatically suitable for a new acquisition channel.
Give each important assumption one identifiable home
Use a stable identifier and link it to the model's input location. ICAEW's spreadsheet principles support clear documentation of assumptions, a logical input-to-output structure and keeping fixed values outside formulas. The register below is our suggested management template, not an ICAEW-prescribed form. Keep the value in one authoritative input location and reference it elsewhere; a separate register must not become a second, contradictory source of numbers.
| Field | What to record | Question for the reviewer |
|---|---|---|
| Definition and model reference | Identifier, unit, population, period and input location | Does 20% mean leads converted, customers retained or something else? |
| Value and evidence | Current case, source date and link or document reference | Can another person reproduce or challenge the evidence? |
| Status and alternative | Observed input, estimate or target; justified alternative case | Is the alternative plausible, rather than an arbitrary percentage change? |
| Owner and approval | Person supplying evidence and person approving its use | Who resolves disagreement before a commitment? |
| Review date and trigger | Next check, quote expiry or event requiring reassessment | Will evidence expire before the decision is made? |
| Decision effect and dependencies | Affected output, related inputs and relevant commitment | Would changing this input alter what management should do? |
Test the assumption through the actual model logic
Hypothetical example: 1,000 eligible leads, a 20% conversion assumption and €300 revenue per completed order imply 200 orders and €60,000 revenue before a capacity constraint. If only 180 orders can be fulfilled in the period, a model that recognises only completed orders gives €54,000 instead. Reducing conversion to 15% produces 150 orders and €45,000. With that capacity limit, the revenue sensitivity is €9,000, not the €15,000 obtained from an unconstrained calculation. All figures and model rules here are illustrative, not industry benchmarks or accounting advice.
Record which dependencies are being held fixed
That example assumes the same eligible leads, completed-order value and capacity in both cases. If extra marketing changes lead quality, or discounts change conversion and price together, changing just one cell may misrepresent the decision. Keep a one-input test for understanding the formula, then build a separately labelled combined scenario where warranted. Explain how unfulfilled demand is handled: lost, deferred or backlogged. Do not quietly assume both immediate revenue and a later backlog sale from the same order.
Review weak evidence before polishing low-impact estimates
Prioritise assumptions where uncertainty could change a near-term commitment. A poorly supported input with almost no decision impact can wait; a supplier quote expiring before a large order cannot. Avoid assigning confidence percentages without a defensible method. Use plain evidence labels such as confirmed for this period, provisional or awaiting confirmation, and state the limitation beside the affected output. Approval means the organisation has accepted use of an assumption for a purpose, not that the assumption has become true.
Keep replacement evidence separate from rewriting history
When an assumption changes, retain the previous value, evidence date, reason and forecast version. Distinguish new information from a correction to an earlier mistake. If a review deadline passes without evidence, flag the input as overdue and show the consequence; do not silently turn it into zero or roll the date forward. Let management decide whether to defer the commitment, use a clearly qualified scenario or seek more evidence.
Use one decision as the acceptance test
Ask someone other than the model author to select a material assumption, find its evidence, change it in a controlled copy and explain the resulting output. They should also identify who must refresh it and when. The useful deliverable is a short, maintained register connected to the model and the next decision. It is not a guarantee of forecast accuracy, and it should never hide uncertainty behind a longer spreadsheet.