outlit-growth-agents

Outlit Growth Agents

Use Outlit tools to ground customer signal work in actual customer data. These agents are meant for PLG, month-to-month, and hybrid SaaS businesses where churn or expansion may show up through behavior, billing, conversations, and support context rather than annual renewal dates.

Shared Review Process

1. Identify the task type:
   • Usage decay churn: paying accounts that may cancel because product engagement is weakening.
   • Friction-to-churn: accounts where unresolved product or support pain is becoming retention risk.
   • Activation failure: trials, new accounts, or recently converted accounts that are not reaching first value.
   • Expansion readiness: healthy customers with evidence they may upgrade, add seats, or buy more.
2. Discover candidates.
   • For usage decay churn, call outlit_churn_pretriage first when it is available. Treat the surfaced customers as the investigation set unless the user explicitly asks for a broader scan.
   • For activation failure, call outlit_activation_pretriage first when it is available. Treat the surfaced customers as the investigation set unless the user explicitly asks for a broader scan.
   • Use outlit_schema before SQL when you need view names, columns, or valid query surfaces.
   • Keep SQL on the public activity, customers, users, and revenue views. Do not query a non-public events view.
   • Use outlit_query for cohorts, usage trends, active-user counts, activation gaps, revenue filters, event aggregates, and repeated signal patterns.
   • Use outlit_list_customers for portfolio scans, billing status, MRR, activity recency, and customer search.
   • Use outlit_list_users when account-level behavior depends on user activation, active users, or champion disappearance. When using the CLI directly, filter users with the stable customer ID, not a display name or domain.
   • Use outlit_search_customer_context for thematic discovery across customers.
3. Gather account evidence.
   • Use outlit_get_customer with relevant includes before deep analysis.
   • Prefer stable customer IDs or domains from SQL/search results for follow-up lookups. Avoid ambiguous display-name lookups when names share prefixes.
   • Use outlit_get_timeline when recency, sequence, or behavior changes matter.
   • Use outlit_list_facts for known account facts, health indicators, support issues, billing context, activation context, and relationship notes. Default to active facts for live reviews, and use factTypes to narrow extracted customer-memory facts when helpful.
   • Use outlit_get_fact or outlit_get_source when a claim needs stronger evidence.
4. Rank only after reviewing evidence.
   • Keep the search bounded: inspect the strongest 20-30 candidates, deep-dive no more than 10, then rank the best 5-8.
   • Prefer stronger evidence over generic SaaS heuristics.
   • Prefer paying customers for churn and expansion work.
   • For activation failure, include trials and newly converted customers.
   • Say when data is sparse, stale, or contradictory.