The goal of sales forecasting isn't to predict the future perfectly — it's to reduce the variance between your best estimate and reality enough to make better decisions. Hiring decisions, marketing spend, inventory, cash planning — all of these are downstream of your revenue forecast. When your forecast is consistently wrong, your entire operating plan is built on sand.
Most companies are using forecasting methods that were never designed to be accurate. They feel official because they're built on CRM data, but the data quality and the method itself both undermine the output. This post breaks down the methods that work and the one you should stop relying on.
The Method That Doesn't Work: Rep-Submitted Commit Forecasts
Let's start here because it's the most common approach and the most predictably inaccurate one. A commit forecast is where reps tell their manager what they think will close this quarter. Managers roll it up. Leadership makes decisions based on it.
The problem isn't that reps are dishonest — it's that the incentive structure systematically distorts their estimates. Reps who consistently over-commit face performance management pressure. Reps who under-commit protect themselves but get their quota raised. The result is a forecast that reflects what reps think they need to say, not what they actually believe will happen.
Method 1: Stage-Weighted Pipeline
Stage-weighted forecasting multiplies each deal's value by the historical close rate from its current stage. A $100K deal in "Proposal" with a 35% historical close rate from that stage contributes $35K to the forecast. Sum all deals across all stages and you have a weighted pipeline number.
This method is dramatically more reliable than commit forecasts if your stage definitions are clean and your historical close rate data is accurate. The two conditions are related — if your stage definitions are inconsistently applied, your historical close rates by stage will be wrong, and the weighted number will be wrong.
The fix: audit your stage definitions (see the stage definition guide), calculate your actual historical close rate from each stage, and plug those numbers into your weighted pipeline formula. Revisit the rates every two quarters as your sales motion evolves.
Method 2: Historical Win Rate by Cohort
This method looks at your historical close rate by deal type, segment, or lead source and applies those rates to current pipeline. For example: enterprise deals sourced from outbound have a 22% close rate over 120 days. Apply that rate to your current enterprise outbound pipeline to get a forecast contribution from that cohort.
The advantage over stage-weighted forecasting is that it's more nuanced — it accounts for the fact that different deal types close at very different rates regardless of stage. A deal in "Proposal" with an enterprise prospect isn't the same as a deal in "Proposal" with an SMB. Treating them the same inflates or deflates the forecast depending on your mix.
This method requires more data history to be reliable — you need at least 12-18 months of cohort data by deal type before the rates are statistically meaningful.
Method 3: Pipeline Coverage Analysis
Coverage analysis doesn't produce a single forecast number — it tells you whether you have enough pipeline to hit your target given your historical close rates. If your close rate from qualified pipeline is 25% and your quota is $1M, you need $4M of qualified pipeline to have a realistic shot.
Coverage analysis is most useful as a leading indicator rather than a point forecast. If you run it at the beginning of the quarter and you're at 2.5x coverage instead of 4x, you have a pipeline problem — not a close problem — and you need to address it now rather than in the last two weeks of the quarter when it's too late.
Track coverage by segment and by rep. An aggregate coverage number that looks fine can hide major gaps for specific reps or deal types.
Method 4: Machine Learning / AI Forecasting
AI-assisted forecasting tools (Gong, Clari, Salesforce Einstein) use deal engagement signals — email activity, meeting frequency, stakeholder involvement, time-in-stage — to adjust forecasts based on behavioral patterns rather than stage alone. They can be significantly more accurate than stage-weighted forecasting for companies with enough historical data.
The caveat: these tools require clean underlying data and 12-18 months of deal history to be reliable. Implementing AI forecasting on top of messy CRM data is one of the most common expensive mistakes in RevOps. Fix your data first. Then add the AI layer.
See the AI lead scoring guide for more on when AI GTM tools are worth the investment.
Building a Forecasting System
The best forecasting systems use multiple methods in combination. A weekly forecast review might include: the stage-weighted number as the baseline, a coverage analysis to check whether you have enough pipeline, a cohort-based view to understand the mix, and a manager-adjusted commit number as a sanity check against the model.
When the model and the commit diverge by more than 20%, that's a signal — either the model is missing something the manager knows, or the manager is applying a bias the model is correcting for. Either way, the divergence is worth a conversation.
Build the infrastructure to run those four views. Use them consistently. Measure forecast accuracy every quarter — the gap between your forecast and actual — and use that data to tune your method. That's how forecasting gets better over time.
Build a forecast your leadership actually trusts.
Take the free RevOps Scorecard to see how your forecasting infrastructure compares to what best-in-class looks like.
Take the Free Scorecard →