Score an account's renewal likelihood on a five-level scale with probabilities, and name the next move for the account owner. Use when asked will this account renew, what's the churn risk on this account, should we run a save play, or when a renewal is 30–120 days out and the owner needs a calibrated read rather than a gut feel. Produces the account state in a fixed schema, the five-level renewal score with a probability per level and confidence, the mapped next move, and the single signal that would change the call.
Start a fresh conversation. Replace the highlighted placeholders.
Use renew-or-churn-call to assess [account name, ARR, renewal date, and term] from [usage and engagement evidence] and [champion, commercial, and history signals]; mark unknowns rather than guessing, and return five-level probabilities, confidence, the mapped next move, and the single signal that could change the call.
Required inputs
[account name, ARR, renewal date, and term]
[usage and engagement evidence]
[champion, commercial, and history signals]
User-approved starter prompt based on a pinned source version; check current skill requirements before running. Task evidence ↗
SKILL.md
name: renew-or-churn-calldescription: "Score an account's renewal likelihood on a five-level scale with probabilities, and name the next move for the account owner. Use when asked will this account renew, what's the churn risk on this account, should we run a save play, or when a renewal is 30–120 days out and the owner needs a calibrated read rather than a gut feel. Produces the account state in a fixed schema, the five-level renewal score with a probability per level and confidence, the mapped next move, and the single signal that would change the call."version: 1.0.0
Renew or Churn Call
Health scores drift into vibes because nobody defines the levels. This skill fixes five levels — from will churn to will renew and expand — each with the evidence that puts an account there, and returns a probability for every level. The next move is mapped to the level, so the output is an action, not a colour.
The account state — six fields, filled from CRM and usage evidence
The renewal score — one of five defined levels, with a probability per level and a confidence score
The next move — mapped from the level (save play, playbook, discovery call, standard motion, expansion)
The flip signal — the one change in state that would move the level
The forecast line — for the renewals sheet: level, probability, move, owner, date
Required Inputs
Ask for these if not provided; mark unknowns rather than guessing:
account — name, ARR, renewal date, term
usage — active seats vs licensed, 90-day trend, key feature adoption
engagement — last executive touch, QBR held, support tickets open/closed, NPS or CSAT
champion — still in role? sponsor aligned?
commercial — price change at renewal, competitor in play, procurement signals
history — prior escalations, credits, missed commitments
Framework: Five Defined Levels
Level
Evidence that puts an account here
Next move
0 · will churn
Champion gone, usage collapsing, competitor selected, or procurement told us
Escalate now: exec-to-exec call, save play with a concession budget
1 · at risk
Usage down, exec contact stale, open escalation or price objection
Run the renewal playbook with a 60-day save plan
2 · uncertain
Mixed signals: healthy usage but no sponsor, or engaged but under-adopted
Discovery call to find the sponsor and the value gap
3 · likely renew
Usage steady, sponsor engaged, no blockers, price accepted
Standard renewal motion; confirm timeline with procurement
4 · renew and expand
Growing usage, seats capped, expansion asked for
Expansion proposal before the renewal paperwork
Thresholds. Act on the mapped move when confidence ≥ 0.55 and the top level's probability ≥ 0.6. Otherwise hold and look at the shape: a 0.40 / 0.40 split between at risk and likely renew is a discovery call, not a save play — the split itself says the sponsor question is unanswered.
Getting probabilities. Any calibrated decision model that scores against ordered, described levels serves this contract; the repo ships an adapter (node integrations/jev/decide.mjs renew-or-churn-call --state state.json, contract in integrations/jev/decisions/). Without one, the assistant estimates and labels it.
Verdict:[act: move] or [hold — shape says: …]Flip signal: [the one state change that moves the level, and who would know]
Forecast line:[account] · L[n] p=0.xx · [move] · [owner] · [date]
Quality Checks
Every state field is evidence or unknown; no invented usage numbers
All five probabilities shown; expected level computed
The move is the one mapped to the level, not a freelance idea
A split distribution is read as a question, not averaged into a colour
The forecast line is one line and fits the renewals sheet
Anti-Patterns
RAG colours with no definition — "amber" that means something different to each CSM
Averaging a bimodal distribution into "medium risk"
Save play by reflex — running concessions on an account whose problem is a missing sponsor
Ignoring the champion field — the single most predictive signal, most often left blank
Letting the score replace the call — the account owner owns the forecast; this is the evidence
Example Trigger Phrases
"Will Acme renew in November? Give me a real read."