The Point Estimate Is a Confession of Ignorance Disguised as Knowledge
When a finance team brings the board a single number — revenue of $42M, payback in 19 months, a 24% IRR — that number is almost never what the team believes. It is the mode of a distribution they have privately collapsed into a scalar because the room rewards apparent conviction. The mechanism is social, not analytical: a point estimate is legible, comparable, and easy to hold someone accountable to, so it travels well through a governance process built on minutes and motions. The cost is that the board never sees the shape of the risk it is actually underwriting. Two projects can share a 24% expected IRR while one is a tight cluster around 24% and the other is a coin flip between a triumph and a write-off. Approving them on the same evidence is not prudence; it is blindness with a paper trail.
What an Interval Buys You, and What It Costs
An interval — say, an 80% confidence range of $34M to $51M — restores the missing dimension. It tells the board the dispersion, and dispersion is the thing capital allocation is supposed to price. The discipline of producing one also forces the analyst to separate two kinds of uncertainty that a point estimate fuses: aleatory risk, the irreducible variance of the world (demand, FX, weather), and epistemic risk, the part that more diligence could actually shrink. The first you hedge or size around; the second you spend to reduce before you commit. A point estimate hides which lever applies. The cost is real and worth naming: intervals invite the lazy dismissal "so you don't actually know," and a wide band can read as a failure of competence rather than an honest map of a genuinely uncertain venture. That reputational asymmetry — punished for honesty, rewarded for false precision — is exactly why the practice is rare, and exactly why it has to be made cheap to adopt.
Designing Intervals a Board Will Act On
The failure mode is not that boards reject uncertainty; it is that they are handed uncertainty in a form they cannot act on. A 95% confidence interval, the analyst's reflex, is the wrong default for decisions: it is so wide it looks like an admission that anything could happen, and it implies a tail precision no forecast actually possesses. The fix is to engineer the presentation around the choice in front of the room, not around statistical convention.
- Lead with the decision-relevant tail. A capital request lives or dies on the downside, so anchor on the 10th or 20th percentile — the question is "how bad, with what likelihood," not "what is the symmetric band."
- State the few assumptions that move the bound most, and show the estimate's sensitivity to each, so the board debates drivers rather than the final digit.
- Attach the cost to narrow the interval — the price of a pilot, a study, a delay — converting epistemic risk into an explicit buy-or-proceed option.
The Decision Implication: Govern the Distribution, Not the Number
Once risk is presented as a range, the board's job changes from ratifying a forecast to pricing a bet, and the right questions follow naturally: Can the firm survive the 10th-percentile outcome? Is the spread between downside and upside one we are paid to bear, or one we are merely exposed to? Is the band narrow because the venture is genuinely safe, or because the analyst rounded away the doubt? A board that internalizes intervals stops treating a missed point forecast as a betrayal — the outcome inside the stated range was never a promise broken — and starts holding management accountable for the honesty of the range and the quality of the reasoning that set it. That is the trade the Financial Lens should insist on: a little less comfort in the boardroom, in exchange for capital decisions that are actually priced for the uncertainty they carry.