Revenue leaders often treat the range as hedging. It is the opposite. A point estimate makes an implicit claim of certainty that no forecasting model can support, and the interval simply publishes the uncertainty the model already carries. Two forecasts of $10.4M mean entirely different things when one comes with a $700K band and the other with a $4M band.
Where the width comes from
The interval is built from the model's historical error. Fit the model, compare what it would have predicted against what actually happened over prior periods, and the spread of those residuals sets the band. A model whose past errors clustered within 4% produces a tight interval. A model that has swung 20% either way produces a wide one, and it should.
Simulation-based forecasts build the interval directly. Resolve every open deal against its close probability thousands of times, and the percentiles of the resulting totals are the band. That approach handles deal concentration properly, which matters because a quarter carried by five large opportunities genuinely has a wider range than one carried by two hundred small ones.
Interval width also grows with horizon. The next month is far more predictable than the quarter after next, so any interval that stays the same width across four quarters out is not being computed, it is being asserted.
A narrow interval is not automatically better
The goal is calibration, not tightness. An interval that is right about how often it is right is doing its job even when it is wide. Teams get this backwards and squeeze the band until it looks decisive, which produces confident forecasts that miss.
ORM's benchmark shows what a properly tight band requires. Forecast accuracy on new and expansion revenue typically lands around 90% and takes heavy manual effort to produce, and it goes stale as conditions change. ORM targets 95% and holds it without manual adjustment from day 1 through day 90 of the quarter, updating as the quarter progresses. Stability like that is what earns a narrow interval. Narrowing the band without earning it just relocates the miss.
Turning an interval into a committed number
Pick a percentile and hold it constant. Many teams commit near the conservative end of the band and carry the distance up to the midpoint as upside, which maps cleanly onto existing commit and upside categories without pretending they are exact. The specific percentile matters less than never moving it to reach a desired answer.
Then audit the band the way you audit the number. Track how often actuals land inside it across many quarters and the interval becomes a real accountability tool. It exposes both directions of failure, overconfidence and vagueness, which a single point estimate of forecast accuracy cannot do on its own. Compare the interval floor against plan before deciding whether pipeline coverage needs a mid-quarter intervention.
Frequently Asked Questions
What is the difference between a confidence interval and a prediction interval?
A confidence interval covers uncertainty in the estimated relationship, such as where the true trend line sits. A prediction interval covers that plus the randomness of a single future outcome, so it is always wider. What revenue teams call a forecast confidence interval is almost always a prediction interval, since the question is where next quarter lands rather than where the average sits.
What confidence level should a revenue forecast use?
80% is the common working level because it produces a band tight enough to plan against while still being right most of the time. 95% intervals on quarterly bookings get so wide they stop informing any decision. Pick one level and keep it fixed, since a level that moves between quarters makes the intervals uncomparable.
Why is my confidence interval so wide?
Three causes dominate. Short history gives the model little to estimate from. Deal concentration means a handful of large opportunities decide the quarter. Volatile conversion rates make each stage transition unpredictable. Width is information about the business, and narrowing it requires changing the pipeline rather than changing the model.
How do you check whether the interval is honest?
Count how often actuals land inside it. Across twenty quarters, an 80% interval should contain the actual about sixteen times. If it contains twenty, the interval is too wide and useless for planning. If it contains ten, the model is understating uncertainty and the band gives false comfort.
Put these metrics to work
ORM builds custom revenue forecast models that turn concepts like forecast confidence interval into prescriptive action for your team.
Schedule a Demo