Your Forecast Is a Guess Wearing a Spreadsheet

Why does the forecast miss?

Not "why did we miss the number" — teams answer that easily enough, usually by naming two deals that slipped. The harder question is why the forecast itself was wrong, and the answer is almost always the same: the probabilities attached to each deal were assigned by a person, based on how the last conversation felt.

A deal marked 80% likely because the champion was enthusiastic is not a forecast input. It's an emotion converted into a number, and once it's in a spreadsheet cell it becomes indistinguishable from real data.

Accurate sales revenue forecasting requires each deal's probability to come from historical conversion patterns rather than seller confidence — and it requires knowing how wrong your model has been.

The four inputs that actually predict

A deal's likelihood of closing is predictable from observable properties, not from sentiment:

Deal stage — the base rate. What percentage of deals historically closed from this stage?

Engagement level — how many stakeholders are active, how recently, across how many channels. A deal with three engaged stakeholders behaves differently from a single-threaded one, and the historical data will show you exactly how differently.

Velocity — how long this deal has been in its current stage relative to your average. A deal moving faster than typical and a deal that's been stuck for 30 days have very different outcomes, even at the same stage.

Qualification score — whether budget, authority, problem, and timeline were actually confirmed. This is where forecasting and qualification connect: a forecast built on unqualified deals is inaccurate for reasons no forecasting model can fix.

Three scenarios, not one number

A single forecast number is a false precision that helps nobody. What leadership actually needs to make decisions is a range:

• —Conservative — what closes if nothing goes right
• —Base — the realistic expectation
• —Optimistic — what closes if the currently-uncertain deals land

The value isn't in the three numbers themselves. It's that the gap between conservative and optimistic tells you how much of your quarter is genuinely uncertain — which is the actual question behind "are we going to hit the number?"

A narrow gap means the quarter is largely determined. A wide gap means you have time to influence the outcome, and it tells you where.

The exclusion floor

Here's a discipline most forecasts lack: deals below roughly 30% confidence should be excluded from the committed forecast entirely.

Not deleted — they stay in the pipeline, they still get worked. But they don't count toward the number leadership commits to, because a long tail of low-probability deals is exactly how a forecast becomes systematically optimistic. Ten deals at 20% confidence contribute two deals' worth of forecast, and in practice they frequently contribute zero, because low-confidence deals fail in correlated ways rather than independently.

Excluding them makes the committed number smaller and considerably more likely to be right. That trade is almost always worth making.

Track how wrong you were

The step that separates forecasting from forecast theater: track forecast accuracy over time and recalibrate when it's consistently off.

Most teams produce a forecast, miss it, and produce another forecast using the same method. Nobody asks whether the method has ever worked. If your forecast has run 20% optimistic for three consecutive quarters, that's not bad luck — that's a measurable, correctable bias in how probabilities are being assigned.

Accuracy within 15% of actual is a reasonable target to design against. More importantly, the direction and size of your error should be known, because a consistent bias is far easier to fix than random noise.

Cadence matters more than sophistication

A forecast refreshed weekly with mediocre inputs will beat a sophisticated model refreshed quarterly, because pipeline changes weekly. A useful rhythm:

• —Weekly refresh against current pipeline data — every Monday, before the week's decisions get made
• —Leadership review every two weeks — enough to catch drift, infrequent enough not to become theater
• —A flag when quarter-end target is at risk, with enough runway to act — a forecast that tells you you'll miss on the last day of the quarter is a report, not a forecast

That last point is the real purpose of forecasting. Not accuracy for its own sake — accuracy early enough that the number is still changeable.

The honest caveat

No forecasting model handles a genuinely unprecedented quarter — a market shift, a major customer event, a competitor doing something structural. Confidence scores derived from historical patterns assume the future resembles the past, and sometimes it doesn't.

What good forecasting buys you isn't certainty. It's the ability to distinguish between "the model was wrong" and "the world changed," which are different problems requiring different responses. Teams forecasting on gut feel can't tell those apart, so they treat every miss as a market condition and never fix the model.