From Scores to Forecast

The forecast is the sum of scores multiplied by deal values. If the model is calibrated — if a score of 80 means 80% — then the weighted forecast is the most accurate number available. No haircuts. No manager gut checks. No "let me add my discount to be safe." The math is direct: each deal value multiplied by its probability score, summed across the pipeline. A pipeline with three deals — $100K at 80%, $200K at 50%, and $150K at 30% — produces a weighted forecast of $225K. That number carries a confidence interval. The interval tells you the range. The range is the honest answer to "are we going to make the number."

The operational discipline is trusting the model when it disagrees with the rep. Every sales leader faces this moment: the rep says the deal will close, the model says 38%. Who is right? Over a sufficient sample, the model is right more often. Not because the model knows the deal better than the rep — it does not. Because the model does not have optimism bias. It does not have a quota to hit. It does not have a champion who assured them it would be fine. The model has data and history. Trust the model for forecasting. Trust the rep for deal strategy. They answer different questions.