By Gertrude Van Horn, SVP & CIO, Cien.ai

“Win rates are averages. Propensity curves are decision tools. The difference is what separates a forecast that looks right from a pipeline that actually behaves as expected.”

– Gertrude Van Horn, SVP & CIO, Cien.ai

Why Is Your Forecast So Confident?

Most CRM forecasts are built on a shortcut: every deal in the same stage gets roughly the same probability.

Stage 2? Maybe 25%.
Stage 3? Maybe 50%.
Stage 4? Maybe 75%.

But two deals in the same stage can have very different odds. One may be from a strong-fit account with real buyer activity. Another may be a cold new logo deal in a weak segment with little urgency.

Yet many forecasts treat them almost the same.

That is the blind spot. Static stage percentages ignore what actually drives outcomes: account size, industry, deal type, source, buyer behavior, rep activity, and patterns from similar past deals.

The result is a pipeline that looks stronger than it really is.

What Should Replace Stage-Based Forecasting?

CRM stages still matter. They show where a deal is in the sales process. But they should not be treated as true win probability.

RevOps teams should use AI to estimate deal probability when the opportunity is created, not just when the forecast is reviewed. The question should shift from “What stage is this deal in?” to “How likely is this deal to actually close?”

A good model looks beyond the stage field. It weighs fit, segment, company size, opportunity type, lead source, buyer engagement, and outcomes from similar deals.

That creates a more realistic view. Some Stage 3 deals may deserve a 70% probability. Others may deserve 12%. The difference is evidence, not a default CRM rule.

Why Does This Matter?

Bad probability math creates bad decisions.

If pipeline is overstated, companies may delay demand generation, over-commit to targets, or push reps to chase deals that were never likely to close.

It also hurts margin. Late in the quarter, teams often discount to “save” deals. But if the real win probability was low from the start, price was not the problem. Pipeline quality was.

A cleaner model gives Sales, RevOps, and Finance one shared view of pipeline value.

For example:

• A $1M deal at a default 50% probability looks like $500K in expected value.
• If AI shows the real probability is 12%, it is worth $120K.
• That gap changes forecasting, coaching, prioritization, and discounting.

What Does Success Look Like?

Success is when pipeline value becomes objective, measurable, and trusted.

At Cien.ai, this means using clean GTM data and AI-driven models to measure pipeline quality as deals are created. Leaders can see which opportunities are worth working, which segments produce weak deals, and where reps are spending time on opportunities with less than a 10% chance of closing.

The outcome is not just a better forecast. It is a better revenue engine. Reps focus on winnable deals. Finance sees future revenue more clearly. RevOps spots pipeline problems earlier. Leadership protects margin instead of relying on end-of-quarter heroics.

Win rate tells you what happened. True deal probability helps you decide what to do next.

About the Cien.ai Growth Essentials Series

This article is part of our Growth Essentials Series, inspired by our work with B2B executives, GTM consultants, and PE operating partners. These articles focus on the non-technical aspects of improving GTM performance. If you want to dig into the technical details of how to measure the concepts we use here, check out our Practical RevOps Analysis Series.