Churn-save Playbook

Trigger conditions

Agent product early churn signals surface 2-3 weeks before SaaS equivalents. The critical difference: task completion rate decline typically precedes login frequency decline. A customer using the product 5 times a day but failing every time is more at risk than one using it once a week with all successes.

Automated monitoring of the following signals; any hit triggers this flow:

Signal	Threshold	Severity
Task completion rate decline	> 15 pp drop in 14 days	High
HITL intervention rate rise	> 10 pp rise in 14 days	High
7-day return rate	falls below 30%	Medium-high
Monthly active workflow count	drops by ≥ 2	Medium
Primary user inactive for 30 days	trigger on hit	Medium
Monthly token usage	drops > 40%	Medium

High-severity signals: CSM contacts within 48 hours. Medium-severity: contact within 1 week.

Roles and responsibilities

Role	Primary responsibility
CSM	Owner; customer outreach, root cause identification, internal coordination
Engineering	Engaged when root cause is product quality (completion drops, HITL spikes)
Product	Engaged when root cause is workflow mismatch — adjust recommendations and re-activate
Sales	Engaged when customer raises contract/pricing issues; not involved for non-contractual issues

Intervention timeline

Days 0-3: signal confirmation and root cause analysis

CSM:

1. Within 24 hours, verify automated alert and filter false positives (customer PTO, scheduled product downtime, etc.)
2. Analyze past 30 days of task-level logs: completion rate, error types, HITL trigger reasons
3. Categorize root cause into one of three:
   • Product quality issue — completion decline correlates with specific workflow / task type → escalate to engineering
   • Usage issue — user attempted tasks beyond agent capability, or workflow misconfigured → CSM education / re-activation
   • Business change — customer business shift, primary contact departure, contract issue → sales engaged

Days 4-7: targeted intervention

By root cause:

Product quality issue:

1. Engineering analyzes and locates within 24 hours (reference incident-response)
2. CSM proactively notifies the customer’s primary contact (don’t wait for them to ask), states expected fix timeline
3. After fix, CSM follows up to confirm customer perception

Usage issue:

1. CSM holds a 1-on-1 conversation to understand the gap between customer’s original expectation and actual usage
2. Recommends 2-3 fitted workflows / task types
3. Provides re-activation seeds: 3-5 high-completion first-task examples

Business change:

1. Sales contacts the customer’s primary contact / new owner
2. Evaluates contract flexibility: smaller scope, lower tier, pause rather than cancel
3. Do not discount easily — discount doesn’t resolve root cause and sets a low anchor for the next negotiation

Days 8-14: track recovery

CSM:

• Monitor the same 6 signals for rebound
• Task completion rate returning to ≥ 80% of pre-decline level counts as initial recovery
• 7-day return rate > 50% counts as habit re-formation

Day 30: verdict

CSM and sales joint review:

• Saved — signals recovered, customer proactively expresses continued usage intent. Archive case
• Stable but low-activity — customer continues but at half magnitude. Accept current state, mark as low-priority account
• Churn confirmed — customer stops paying / does not renew at contract end. Complete exit interview, archive failure cause

Critical things not to do

• Do not rely on discounting to save — trading one-time cashflow for the customer’s future discount expectation costs more long-term than short-term gain; agent product cost structure makes discount-to-vendor margin damage far more severe than SaaS
• Do not recommend new workflows before root cause is confirmed — the customer has experienced quality decline; pitching a new untested workflow reads as “now you want me to invest more”
• Do not let support handle churn in lieu of CSM — support resolves specific tickets; only CSM can see the pattern. Conflating roles results in churn being logged as “ticket closed”
• Do not defend in the exit interview — the goal is to collect honest reasons for product iteration, not to win the argument

Measurement metrics

Track quarterly:

• Early-signal triggers vs actual churn: ideal ratio 3:1 — not every trigger is real churn, but the system should catch most real churn cases
• Save success rate: percentage of triggers where signals recover within 30 days (healthy baseline 40-60%)
• Median save duration: days from signal trigger to recovery
• Churn reason distribution: by root cause; periodically review top-three causes and reverse-engineer product / sales process changes

Cross-section connections

• Specific early signal metric definitions: metrics/overview
• Product quality escalation path: incident-response
• ROI impact on customer when agent completion rate drops: economics/controls-and-roi
• Internal support ticket correlation with early churn: support-runbook