How Reliably Accurate Can a Sales Forecast Be?
Put SalesTech Forecast Accuracy Claims to the Test
Actual SalesTech forecast-accuracy claims. The definitions are unclear. We take a stab at deciphering and stress-testing these claims.
Prioritizing accuracy can be harmful because it shifts attention toward an elusive accuracy goal and away from an achievable one: maximizing sales.
Vendors promise 95% forecast accuracy
The vendors don't publish clear definitions for these claims. But their tools are focused on predicting sales bookings (new contracts), not GAAP revenue. So we interpret 95% forecast accuracy to mean sales bookings landing within ±5% of the forecast. Let's see how reliably an open-pipeline forecast can meet that standard under favorable conditions.
The method we use here may seem different from the way sales leaders normally forecast. Companies may estimate the probability of individual deals, apply historical close rates by stage, incorporate micro-signals, roll up Forecast Categories, build consensus predictions—or rely heavily on judgment.
​
We are not trying to replicate those forecasting processes. We are modeling the underlying reality they all face: every open deal has some element of chance in its outcome. Statisticians call this a stochastic process. A stochastic process produces a distribution—a range of possible outcomes—rather than one predetermined result.
Sales teams work hard to influence those outcomes, and often do so effectively. But before the outcomes are known, uncertainty remains. Changing the forecasting method can change the forecast. It does not eliminate the underlying uncertainty.
​
The explorer therefore asks a more fundamental question: if you knew the true win probability of each Forecast Category before the quarter played out, how reliably would actual sales land within ±5% of the resulting forecast?
​
To make that question easy to explore, we group open opportunities into three familiar Forecast Categories—Commit, Best Case, and Pipeline. Within each category, all deals are assumed to have the same known win probability. And all deals are the same size. You can adjust both the number of deals and the probability assigned to each category. The model also assumes that deal outcomes are independent and that the forecast itself creates no feedback that changes seller behavior or deal outcomes.
​
Real pipelines routinely violate these assumptions. Deal sizes vary, probabilities differ within categories, outcomes may be correlated, and management action can change results. Those effects generally make the range of possible outcomes wider, not narrower.
​
So think of the meter as showing a favorable boundary: real-world forecast reliability will usually be no better, and often worse.
The Forecast Accuracy Meter shows how frequently the 95% accuracy claim would fail under favorable conditions.
Enter the number of deals in each of your forecast categories and their win probability. Default values are illustrative. Adjust the settings to replicate a representative range of scenarios for your business.
​
For most settings, the 95% accuracy claim is not possible. Nor is it necessarily desirable: chasing accuracy can encourage sales teams to manage toward conservative forecasts rather than maximizing the outcome.
​
The greater value of a forecast is therefore in informing action that improves sales—not in precisely predicting the result.
​Accuracy is not the objective
A sales forecast is not a weather forecast. Sales leaders can change the outcome.
Rather than trying to predict the outcome against difficult odds—or worse, managing the team toward a conservative prediction—use the forecast to determine where attention and resources can produce the greatest return.
When sales teams focused on the deals prioritized by Funnelcast, they generated 60% more sales on average than when they followed their existing priorities.