Revenue forecasting in SaaS is hard because you're trying to predict three different things simultaneously: new logo ARR (dependent on sales execution in a given window), expansion ARR (dependent on product adoption and CS effectiveness), and net churn (dependent on customer success and competitive dynamics). Each has different predictability characteristics and different leading indicators.
The top-down forecast mistake: start with "we're growing 40% year-over-year" and build a straight line from current ARR. This produces a number that the CEO wants, not a forecast that reflects operational reality.
The bottom-up forecast components that produce reliable outputs:
New logo: base the forecast on stage-weighted pipeline at current win rates, then apply a judgment factor for pipeline quality. The judgment factor requires the AE who owns each opportunity to give an explicit assessment of close probability. Aggregate AI-generated probability scores without human input are less accurate than individual rep estimates because reps have context the model doesn't.
Expansion: base the forecast on identified expansion accounts with known expansion triggers. A list of 40 accounts with identified triggers and an expansion conversation in progress, at historical expansion conversion rates, produces a more accurate expansion forecast than an aggregate expansion rate applied to total ARR.
Gross churn: base the forecast on accounts in the at-risk bucket of your health score, weighted by their risk score and ACV. Accounts with health scores below a certain threshold that are also in their renewal window are your at-risk ARR.
Combine these three components with an explicit uncertainty range, not a point estimate. A revenue forecast that says "$4.2M new logo ARR in Q3 ± $600K" is more useful than "$4.2M" because the range reflects the uncertainty inherent in the forecast.
Forecast honestly. The miss that isn't flagged in advance is the one that destroys credibility.