Renewal Forecasting

⚠️ Handbook ahead of software. The CSP currently has four forecast categories and no weighted revenue rollup. The fifth category (Will Not Renew) and the weighted monthly forecast are scoped in BUILD_BACKLOG.md for the Prism v2 push.

What forecasting is for

A forecast is a commitment to a judgment, made early enough to act on. Its value isn't prediction accuracy for its own sake — it's that categorizing an account forces the CSM to articulate why, which surfaces gaps in what they actually know.

An account nobody can categorize is an account nobody understands.

Why forecast is decoupled from health

Forecast is deliberately not derived from the health score. They answer different questions and rely on different inputs.

Health is computed from observable behavior — usage, tickets, engagement, plan progress. Forecast folds in things the data model cannot see:

• Competitive evaluation the customer mentioned on a call
• A budget freeze announced internally
• Procurement or security review that could delay signature past the renewal date
• Sentiment — the champion who is enthusiastic but has just lost an internal argument
• M&A, restructuring, or a client loss on their side that changes what they need

A healthy account can be a renewal risk. An account with mediocre health can be a certain renewal because the buyer has already committed budget. Deriving forecast from health would erase exactly the judgment that makes forecasting worth doing.

The tradeoff, named honestly: manual forecasting is only as good as the CSM's discipline, and it's vulnerable to systematic optimism. That's what the movement rules and the quarterly accuracy review exist to counter.

The five categories

Two rules with teeth

Unknown is banned inside the renewal window. An Unknown at T-60 is not a forecast — it's an admission the renewal motion hasn't started. The weekly checkpoint pass (2.3 Operating Rhythm) exists to prevent this; any account still Unknown inside the window is a prioritization failure, not a data gap.

Will Not Renew requires a customer signal. A strong inclination that an account will churn is At Risk, not Will Not Renew — keep working it. This matters because Will Not Renew moves the account toward offboarding (3.8 Churn & Offboarding) and effectively ends save effort. Entering it on a hunch is self-fulfilling: write the account off, stop investing, watch the prediction come true. The category records what the customer has told us, not what we fear.

Movement rules

Forecast changes only with a stated reason, logged as an activity.

"Moved to At Risk: champion departed, no successor identified, EBR postponed twice."

Not: a silent category change during a monthly pass.

This does three things: it prevents drift, it creates an audit trail of the CSM's own judgment, and it's the raw material the quarterly forecast-accuracy review reads (2.3 Operating Rhythm). A forecast history with no reasoning attached teaches nothing in hindsight.

When forecast is reviewed:

• Weekly — for accounts crossing a renewal checkpoint (3.7 Renewal Motion)
• Monthly — full pass across the renewal horizon, not just imminent renewals
• Immediately — when something material happens: champion departure, escalation, stated dissatisfaction, competitive mention

Direction is not sticky. Moving an account from At Risk back to Likely is legitimate when the threat is genuinely resolved — and requires a stated reason like any other move. Categories that only travel downward aren't forecasts, they're pessimism with extra steps.

Weighted revenue forecast

Each category carries a probability weight. Applied per account and summed, this produces the forecast number.

Calculation: for each renewing account, ARR × category weight. Summed by month, reported against total renewing ARR so both the ceiling and the expectation are visible.

Why these numbers. Likely sits near the book's gross retention base rate — if Likely were 75%, that would imply a 25% churn rate among healthy accounts with no contrary signal, which describes a failing book, not a forecast. Unknown sits modestly below base rate: absence of information is not presence of bad news, but unknown accounts skew toward neglected ones, so a discount is warranted. At Risk sits well below because a named, active threat genuinely changes the odds.

Two things to keep straight:

1. This is a book-level planning number, not a per-account one. No individual account renews at 85% — it renews or it doesn't. The weighting only means something in aggregate.
2. These weights are starting estimates, to be calibrated against actual outcomes at the quarterly accuracy review. A weight never checked against what happened is decoration.

Forecast accuracy review

Quarterly, as part of the system review (2.3 Operating Rhythm):

• Did Commit accounts renew? A Commit that churned means the evidence standard wasn't applied — someone forecast a feeling.
• Did At Risk accounts churn, or were they saved? If most At Risk accounts renew fine, the category is being applied too readily and the weight is too low.
• Was Unknown used inside the window? Each instance is a process failure to name.
• Is the aggregate biased? Systematic optimism is the most common forecasting failure and only shows up in hindsight, across many accounts.

Output: recalibrated weights, or an explicit decision to leave them.

Related pages

• 1.3 Customer Data Model — forecast category as a stored field
• 1.6 Health Scoring Methodology — why health and forecast are separate signals
• 2.3 Operating Rhythm — when forecast is reviewed
• 2.4 Account Prioritization Framework — how forecast interacts with triage
• 3.7 Renewal Motion — the checkpoints where forecast is set and revised
• 3.8 Churn & Offboarding — what Will Not Renew hands into