Setup and customization checklist

Placeholders

• {{USAGE_SOURCE}} and {{BILLING_SOURCE}}: read-only access to usage events and subscriptions, joined on the customer account.
• {{USAGE_EVENT}}: the one event that means the product did work for the customer. Pick one, so the counts mean something.
• {{HUMAN_ACTION}}: the event that means a person actively used it, as distinct from automated runs.
• {{KEY_SETUP_STEP}}: check it against your churned customers. If a different setup step predicts churn better, use that one.
• {{GENERATED_OUTPUT}}: if your product produces nothing for people to review, drop the review signal.
• {{USAGE_UNIT}} and {{PLAN_PRICING}}: per-unit rates and base fees for the plan-upgrade check.
• {{TOP_TIER_TAGS}}, {{EXPANSION_OWNER}}, {{CS_OWNER}}, {{EXPANSION_CHANNEL}}, {{DIGEST_RECIPIENT}}.
• {{EXPANSION_PIPELINE}}: the pipeline and first stage. Resolve stage IDs at run time instead of hard-coding them.
• CRM properties: lifecycle phase, risk tier, expansion tier, subscription status, plan, renewal date, allowance total and used, 30-day activity, last activity date, MRR, ARR, last synced. Create them before the first run; missing ones are skipped and logged.

Defaults to tune

Default	Value	Notes
Activating window	60 days from first payment
Never activated	No activity and 14+ days since first payment
Went dark	21+ days since last activity (high), 14 to 21 (medium)
Human silence	14 to 30 days since last human action (medium)
Month-over-month floor	50 events in the prior 30 days	Below it, momentum signals are skipped
Volume drop	More than 50% (high), 30% to 50% (medium)	Established customers only
User drop	Half or more, from at least 2 users	Established only
Unreviewed output	50+ items in 14 days, none reviewed
Failure rate	Over 40% of at least 10 events
Top-ups	2+ in 60 days (high expansion)
Running out	14+ days before renewal, two months running (high); 7+ days once (watch)
Plan-upgrade floor	$40 a month in savings	Recompute the volume threshold on any price change
Seat utilization	75%+ (watch)
Autopilot drift	50+ events in 30 days, no human action in 45+ days	Digest only
Running low	85%+ of allowance used	Digest only

Behaviour that should not change

• Three separate tiers; never blend them into one score.
• Read-only against product and billing data.
• A failed pull is logged and skipped; the run continues.
• One alert and one deal per account per run.
• Medium expansion never triggers outreach, and nothing here contacts a customer.
• Subscription status, not the renewal date, decides churn.