Skip to main content

Updated Sep 30, 2026Verified against the product Sep 30, 2026

Open in ClaudeOpen in ChatGPT

Prevent Churn with a Retention Outcome Agent

A Retention Outcome Agent turns "reduce churn" into a program you can run and measure. It learns from your past renewals which product adoption separated the accounts that renewed from the ones that didn't, finds the accounts renewing soon that are on the wrong side of that line, recommends what to do about it, and re-measures those accounts every day until the run ends. Use it when you want to spend the next quarter's save effort where the data says it will move renewals, and to know whether it worked.

This guide walks through one full run. For every field and screen, see the Outcome Agents reference.

How the loop maps to FunnelStory​

StepWhat it meansWhere it happens
GoalProtect renewals for a product, over a set windowA draft with the Renewal outcome target
Identify cohortsSplit renewing accounts into an at-risk group and the adoption level that historically renews betterCohort Analysis
StrategizeDecide what moves an account from one group to the otherNext Best Actions on each cohort pair
ActWork the at-risk accountsYour team, an AI Agent, or a Play
MeasureTrack who crossed over, who churned, and what revenue that protectedThe Activated report, re-measured daily

Before you start​

  • The Outcome Agents page is on. FunnelStory enables it per workspace; when it's on, Outcome Agents appears in the left sidebar.
  • You can edit workspace configuration. Creating drafts, generating cohorts, and activating them require it.
  • There's enough renewal history. The analysis needs roughly 100 past renewal decisions for the product and at least two usage signals with enough history. If there isn't enough, the form says "Not enough data is available to generate cohorts yet."
  • Products are set up. Cohorts are built per product from its activity models and account metrics. See Products.

Step 1: Set the goal​

  1. Open Outcome Agents and click New Outcome Agent.
  2. Choose the Feature Adoption template and the Product you want to protect.
  3. Set the windows:
    • Target Renewal Window — the accounts the agent acts on. The default, the next 90 days, fits a quarterly save program.
    • Training Window — the past renewals it learns from. Keep the default two years unless your product or pricing changed recently; learning from a different era gives you advice for a different product.
  4. Leave Outcome Target on Renewal.
  5. Under Compare Accounts Within, pick the traits that make accounts true peers — for example Tier, so enterprise accounts are judged against enterprise accounts. Each extra trait makes groups smaller, so add only the ones that genuinely change how accounts behave.
  6. Click Generate Cohorts.

A saved draft can't be edited. To try different settings, create another agent and compare the results.

Step 2: Choose the cohorts worth running​

Click View Cohort Analysis. Each cohort pair compares an At-Risk Cohort (current adoption) with a High Retention Cohort (the next adoption level up) among peers, and pairs are ordered by the revenue they put in play.

Pick pairs with three questions:

  1. Is there enough at stake? Revenue at stake is the ARR of the at-risk accounts renewing in your window.
  2. Is the gap real? Prefer High confidence, and a large rate gap between the two cohorts' historical renewal rates.
  3. Can your team move it? Read Why this cohort pair was chosen. A gap your team can close — a feature they can enable, a workflow they can teach — beats a bigger gap you have no lever on.

Some agents also show one-sided COHORT cards grouped by a recurring Needle Mover theme. Use them to see which accounts share a concern; they can't be activated yet.

Step 3: Agree on the strategy​

Expand Next Best Actions on each pair you plan to run. It's one strategy of two to four ordered steps, not a menu: each step names the state the whole cohort should reach and the signal that proves an account got there, and it combines levers such as adoption, relationship, commercial terms, and enablement.

Before you activate, agree with your team who owns each step. The agent measures movement; people and automations create it.

Step 4: Activate​

  1. On the pair, set Run for to Month, Quarter, or Six Months. Match it to your Target Renewal Window, so the run lasts until those renewals are decided.
  2. Click Activate.

Activation freezes the definition and locks today's at-risk accounts as the run's roster. From now on, the goalposts don't move: every day, FunnelStory checks the same accounts against the same target.

Step 5: Put the actions to work​

Work the roster through whichever channels fit each step of the strategy.

Give your CSMs the list. Open the run from the Activated tab: Audience List is the locked roster. For each account, the account's Renewal → What's Next read and its Account Dossier give the CSM account-specific talking points for the gap.

Automate the first touch with an agent. An AI Agent with a Query trigger can run once for each account on the roster. Cohort membership is in the cohort_accounts table, and the run's label is the name shown on the Activated tab:

SELECT a.account_id, a.name
FROM cohort_accounts ca
JOIN cohorts c ON c.id = ca.cohort_id
JOIN accounts a ON a.account_id = ca.account_id
WHERE ca.role = 'at_risk'
  AND c.status = 'active'
  AND json_extract(c.data, '$.label') = '<cohort label>'
LIMIT 200

Each returned row starts one run, and identical rows aren't re-run, so every roster account is handled once. A good first agent investigates the account and drafts an Action Card — a recommended step and outreach email that a CSM reviews and sends. See Design an Agent: Detect, Investigate, Act for how to build it.

Run a campaign when the fix is the same for everyone. If the gap is education — for example, most of the roster never set up one feature — a Play with a drafted email sequence may reach the whole cohort faster than one-to-one outreach.

Attach the flow. In the run's report, click Attach Flow under Attached Agentic Flow and pick the agent working this roster. The cohort doesn't run the flow; it records which agent's work it's measuring.

Step 6: Measure progress​

Open the run from the Activated tab. FunnelStory re-measures the locked roster once a day.

TileRead it as
ResultRoster accounts that reached the target adoption, out of the roster
ARR ProtectedThe revenue attributed to the run, estimated from the roster's ARR and the pair's historical rate gap
Tracking SummaryAvg. Time to Transition, plus how many are Still at risk and how many Churned

Review it weekly with the owners of each strategy step:

  • Accounts crossing steadily — keep going, and look for what the fastest movers had in common.
  • Nothing moving after a few weeks — the actions aren't reaching the gap. Check that the agent is running and that CSMs are acting on its cards before changing the strategy.
  • Accounts churning from the roster — look at their Needle Movers; a concern that adoption can't fix, such as pricing or a lost champion, needs a different play.

Between runs: catch the risk the cohort can't see​

A cohort run covers one adoption gap. Keep watching for everything else that puts renewals at risk:

  • Predicted churn — in Focus Areas → Revenue Review, filter by the Churn outcome to see accounts the prediction model flags, whatever their adoption.
  • Leading indicators — Needle Movers surface concerns such as pricing pressure, competitor mentions, or a sponsor leaving, often before they reach a prediction. Turn on autosuggested action cards so FunnelStory investigates open needle movers and drafts the next step before anyone asks.
  • Thin relationships — Relationship Maps show accounts that depend on a single contact or where engagement has dropped.

When you understand why an account is at risk, the account page has the detail: Renewal → Driving Factors for the health score, the Needle-Movers tab for open concerns, and Adoption → Event Timeline for when engagement changed.

Last updated on