Scenarios & uncertainty
Never hand leadership a single number: ranges tied to assumptions, scenarios tied to decisions, and the communication patterns that make uncertainty useful instead of scary.
A point forecast — 'week-9 cash: $1.84M' — is a fiction wearing precision. Every forecast is a distribution; the professional question is how to communicate that without either false confidence or useless hedging. Two tools, each tied to something concrete:
- Ranges tied to assumptions. 'Week-9 cash $1.6M–$2.1M, driven mainly by DSO (31-38 days) and diesel (±12%)' — the range isn't decoration, it's traceable: widen an assumption, watch the band widen. Practical construction for spreadsheet finance: run the model at pessimistic/base/optimistic values of the 2-3 assumptions that matter (the challenger pass told you which), not all of them — a range built from twenty simultaneous worst cases is a horror story, not a forecast.
- Scenarios tied to decisions. A scenario earns its slide only if someone would do something different in it: 'If Marlowe Retail (11% of revenue) churns: cash stays positive through week 13 but the credit-line draw moves from unlikely to probable in week 9 — recommend we extend the facility review now, while it's cheap.' Scenario → threshold → action. Scenarios without decisions attached are speculative fiction with a budget.
Communicating it upward (where forecasts live or die)
- Lead with the base case and its shape, not the machinery: 'Cash stays above $1.2M throughout; the tight week is week 9 (payroll + quarterly insurance); base case assumes current DSO holds.'
- Name the watch items: the 2-3 assumptions that would move the answer, each with its early-warning signal ('if DSO breaches 36 by week 4, week-9 tightness becomes a draw'). This converts your uncertainty into their monitoring plan — uncertainty made useful.
- Score yourself publicly. Forecast vs. actual vs. naive baseline, every cycle, one small table in the pack. Nothing — nothing — builds forecast credibility like visibly tracking your own misses and explaining which assumption broke. It also quietly retires the colleague whose forecasts are never wrong because they're never checked.
- Let AI draft the narrative, then verify every number in it (the Module 1 rule). AI writes fluent variance commentary; it also fluently drifts figures while polishing. The verify pass is two minutes; skipping it once in a board pack costs more than every minute it ever saved.
In the 13-week grid, color-code each cell's mechanism: known (booked AR, scheduled payroll — weeks 1-2 mostly), pattern (seasonal collections, fuel-indexed costs — the middle), assumption (the far weeks). Leadership instantly sees where the forecast is solid and where it's judgment — and stops arguing week-2 precision while under-scrutinizing week 11. One formatting habit, disproportionate trust dividend. The lab implements it.