Churn Signal Digest
Weekly digest of accounts whose usage is dropping.
role growthrole csmonitoringdigestreportingintermediatepro planorigin selfupdated 2026-10-24
01Requires
what are these? skills vs memories vs plugins →
churn signals — what 'dropping' means for your product (logins, key actions, seats active) memoriesaccount tiers — which accounts are worth a save call vs a nudge email memoriesgoogle-sheets mcpread access to the usage-export sheet permissionsno write access — the digest is for humans to act on permissions
02Setup — 6 steps
- 01Create the agent. Open Grok Bot → Agents → New Agent, name it
churn-signals. - 02Paste the prompt. Copy the prompt below into the system field.
- 03Set two memories. Memory one: your churn signals with thresholds. Memory two: account tiers and who handles each.
- 04Connect usage export. Attach the weekly product-usage export (account, logins, key actions, seats).
- 05Calibrate week one. First run flags too much or too little — tighten thresholds until the list is act-on-able.
- 06Schedule Monday. Weekly run before the CS/growth standup so saves get assigned same-day.
03Prompt
agent prompt · system field
You are a churn-signal analyst. From this week's usage rows:
- Flag accounts matching a churn signal from memories (e.g., logins down >50% WoW, key-action zero).
- Rank flags by account tier — high-value first, with the specific signal cited per row.
- One line per account: account name, signal, tier, suggested play (call / nudge / watch).
- Close with the week's count and biggest mover. No verdicts on accounts not flagged.
04Config
| name | example | description |
|---|---|---|
signals memory | "logins -50% WoW OR zero key actions for 7d" | Thresholds tuned to YOUR product's usage shape |
tier memory | "tier-A = top 20 accounts by ARR → save call" | Who handles what — turns the list into assignments |
05Sample output
Example output shape: 6 flagged accounts — 2 save-call (tier-A, logins -70%), 3 nudge, 1 watch.
06Pitfalls
Cry-wolf threshold. If every week flags 30 accounts, nobody reads it. Tune until the flag list fits on one screen.
Seasonal dips look like churn. Add seasonal context to the signals memory (holidays, industry slowdowns) or you'll re-save the same quiet accounts quarterly.
07Variants
- seat-shrink-watch — narrower: only tracks seat-count contraction, not usage depth
- winback-scorer — same signals applied to already-churned accounts for winback ranking
- health-score-digest — turns signals into a 3-band score (healthy/watch/at-risk) instead of flags
08FAQ
Where does the usage data come from?
Whatever exports to a sheet — your analytics tool's CSV export, a warehouse query, or a manual pull. The agent just needs rows.
Can it email the account owners directly?
Keep it a digest. Auto-emailing customers 'we noticed you left' is the fastest way to churn them.
Growth or CS — whose job?
Both. Growth owns the metric, CS owns the save calls — that's why the digest ranks by tier, so each side sees its part.
09Run it — or hand it to us
Tell us what you want Grok Bot to do: hello@awayhand.com
10Related templates
Onboarding Nudge Drafter
Drafts activation nudges for stalled signups.
growth · 2026-10-24
Weekly Ticket Themes Digest
Monday digest: what customers complained about, clustered.
support · 2026-10-24
↩Also referenced by
unofficial template from a community-run directory — not affiliated with the product's vendor · disclaimer
New templates as they land. No hype.
Email to subscribe →