Most renewal forecasts are works of fiction with a confident font. They are assembled from CSM judgment calls, a column of dates, and a column of "likely / at risk / committed" that each owner fills in with whatever they felt during the last call. Then leadership commits a number to the board, the quarter plays out differently, and everyone agrees the forecast was "directional." Directional is a polite word for wrong.
A renewal forecast you can actually commit to has one property: it is grounded in observed account behavior rather than self-reported confidence. Here is how to build one.
Why the optimism-based forecast fails
The standard process asks each CSM to categorize their renewals by gut feel. This breaks for predictable reasons.
- Optimism bias is structural, not personal. People who own relationships are wired to believe the relationship is fine. They are close to the account and emotionally invested in its success. This is a feature in a CSM and a bug in a forecaster.
- Categories are inconsistent across owners. One CSM's "committed" is another's "at risk." Without a shared, evidence-based definition, you are aggregating noise.
- It updates too slowly. A status set at the start of the quarter goes stale the moment an account's usage drops or a champion leaves, and nobody goes back to flip it until it is too late.
The ingredients of a grounded forecast
Replace gut feel with three layers of evidence, combined.
1. A health score per account, weighted by behavior
Start from an objective health score that reflects outcome attainment, adoption breadth, sentiment, and commercial fit, not a CSM's mood. We cover what belongs in that score in what a health score should measure. The score gives every renewal a defensible baseline probability before any human judgment enters.
2. The trajectory, not just the snapshot
A 70 that is falling is more dangerous than a 55 that is climbing. Direction carries more predictive weight than the absolute level, because it captures momentum in the relationship. A forecast that ignores trajectory will be repeatedly surprised by accounts that looked fine on the day they were scored.
3. CSM judgment as an override, not the foundation
Human context still matters. A CSM may know that a low-scoring account just signed a verbal commitment, or that a high-scoring one is about to lose its sponsor to a competitor. The right design uses the data as the default and treats CSM input as a deliberate, logged override, so you can later check whether the overrides made the forecast better or worse.
A forecast is only as trustworthy as your ability to explain, after the fact, why each call was right or wrong. Gut feel cannot be audited. Evidence can.
Close the loop
The fastest way to a trustworthy forecast is to grade the old one. After each quarter, compare predicted outcomes to actual renewals and ask where the model and the overrides diverged from reality. Accounts that churned while scoring healthy reveal a blind spot in your inputs. Accounts a CSM saved despite a red score reveal context worth systematizing. A forecast that learns from its own misses gets sharper every quarter; one that is rebuilt from scratch on vibes every quarter never improves.
Where Merrily fits
Merrily supplies the first two layers automatically: a behavior-weighted health score per account and its trajectory over time, both tied directly to the renewal date and the ARR at stake. It leaves the override in human hands and keeps a record of every one, so the forecast is both grounded and auditable. The result is a renewal number you can put in front of a board without crossing your fingers, because every line in it points back to evidence you can see.