Boards do not lack numbers. They lack numbers they can interrogate. There is a difference between being shown a forecast and being able to ask it questions: what happens if volume drops eight percent, which assumption is doing the heavy lifting, where exactly did this margin figure come from. A model that cannot answer is not analysis, it is theatre with a spreadsheet.
The models we build for business decisions follow a small set of rules, and none of them are sophisticated. They are just disciplines that most models skip under deadline pressure, which is precisely when they matter.
Three cases, honestly different
Base, upside, downside. Not base plus five percent and base minus five percent, which is one case wearing three costumes, but genuinely different states of the world with the operational story that produces each: the contract that lands or does not, the wage settlement at four percent or six, the customer that renews or walks. If the downside case does not make anyone uncomfortable, it is not a downside case.
If a number matters, it should survive interrogation. If it cannot, you needed to know that before the meeting.
Every figure traceable
Each line in the summary traces to its source: the ledger export, the contract schedule, the assumption register with a date and an owner. This sounds bureaucratic and takes surprisingly little effort at build time. Its value appears the first time a director asks "where does that come from" and the answer takes eleven seconds instead of a follow-up meeting.
Sensitivity before opinion
Before the model expresses a view, it shows which inputs move the answer. A tornado chart of six sensitivities tells a board more than twenty pages of narrative: if the whole case swings on one price assumption, everyone should know that first, and the meeting should spend its time on that assumption rather than admiring the base case.
The final discipline is framing. Analysis supports a decision; it does not make one. The model lays out the options and what each costs, and the decision stays where it belongs, with the directors and their advisers, made with the numbers in full view. That division of labour keeps everyone honest, including us.
What to do about it
- Refuse single-case forecasts. Require a downside that hurts and the story that produces it.
- Ask where any figure comes from. If the answer is not immediate, the model is not finished.
- Ask which three assumptions move the answer most, and stress those in the meeting.
- Keep an assumption register with owners and dates. Stale assumptions are how models rot.
Our financial consulting practice builds models on these rules for boards, lenders and transactions. They are less impressive to look at than the forty-tab kind, and considerably harder to argue with.