Preventing Churn

• Source: Report research library
• Frameworks applied: Gainsight four-driver churn diagnosis and save plays, health band alerting with scorecard overrides, Valuecase time-to-value measurement discipline (signature to first value milestone, median), Help Scout six-stage journey map with milestone red flags

Four-driver diagnosis table with distinct data and conversation signatures, and a recovery criterion that re-diagnoses rather than extending a failed play.

Known weakness, published as found: There is no fenced pre-output validation gate like the sibling skills use: add one (driver assigned exactly once, recovery criterion has metric+threshold+date, rejected drivers stated). Cut the health-band and GRR/NRR benchmark material that duplicates scoring-customer-health and modeling-saas-revenue-metrics, and link out instead.

1. 1.Copy the SKILL.md text below, or download the raw file.
2. 2.Create a folder named exactly preventing-churn in your agent's skills directory.
3. 3.Save the file inside that folder as SKILL.md.
4. 4.Ask the agent one of the trigger requests below.
5. 5.Check the output against what you already know before it leaves your desk.

• Usage dropped off a cliff at our biggest account and renewal is in 90 days — what do I do?
• Our first-year churn is way too high, how do we fix onboarding and time to value?
• Build me a save play for an account whose champion just left the company

• What weights and thresholds should our customer health scorecard use?
• Put together the QBR deck for next week's executive business review

---
name: preventing-churn
description: >-
  Diagnoses the actual driver behind an at-risk account, runs the matching save
  play, and instruments time-to-value and journey milestones so early churn is
  removed at the source rather than fought at renewal. Use when the user says
  churn, at-risk account, save play, churn risk, "why are customers leaving",
  "this account is going to churn", customer retention plan, early churn,
  onboarding is too slow, time to value, TTV, "customer went dark", "usage
  dropped off a cliff", or asks how to reduce logo or gross revenue churn. Use
  this skill whenever the task is diagnosing or intervening on a specific
  retention risk. Do NOT use for building or weighting the health score itself
  (see scoring-customer-health), for renewal forecasting and expansion
  sequencing (see planning-renewals-and-expansion), or for preparing a
  customer-facing review (see running-executive-business-reviews).
metadata:
  version: "1.0"
---

# Preventing churn

Diagnose which of four drivers is actually causing an account's risk, run the matching save play, and fix the upstream time-to-value and milestone failures that manufacture the next cohort of at-risk accounts. One job: churn diagnosis, intervention, and prevention. Score construction, renewal forecasting, and review decks are out of scope.

There is no published, source-backed churn-prediction model with named inputs and intervention thresholds; substitute the weighted health score plus journey-milestone red flags described below and label any predictive claim accordingly.

## Workflow

Copy this checklist into your reply and tick items as you complete them:

```
- [ ] 1. Confirm the risk signal and its recency
- [ ] 2. Diagnose the driver (one of four)
- [ ] 3. Run the save play matching that driver
- [ ] 4. Verify recovery against a measurable criterion; loop until it clears
- [ ] 5. Instrument TTV to remove the upstream cause
- [ ] 6. Attach journey red flags so the next account is caught earlier
```

**1. Confirm the signal.** Risk detection runs off the weighted health score with bands **Healthy 71-100 / At Risk 31-70 / Critical 0-30**, with alerting configured per segment and threshold ([Gainsight](https://www.gainsight.com/blog/customer-health-scores/)). Before acting, confirm the score is current and check for a manual override condition the telemetry cannot see, such as stakeholder loss or known product underuse ([Gainsight Communities](https://communities.gainsight.com/predictive-health-scoring-321/build-a-foundational-health-scoring-framework-using-dear-26486)). A stale score sends the CSM into the wrong play, which costs more credibility than no outreach.

**2. Diagnose the driver.** Talk to the customer to determine whether the issue is **unclear value**, a **missing feature**, or an **onboarding gap**, and analyze usage for drop-offs and underused features ([Gainsight](https://www.gainsight.com/blog/customer-health-scores/)). Add **relationship loss** as the fourth driver, because a sponsor departure produces the same score decay with a completely different remedy.

Judgment step — you assign the driver. Decision criteria:

| Driver | Signature in the data | Signature in the conversation |
|---|---|---|
| **Unclear value** | Steady usage, no ROI evidence tied to stated goals, flat depth of feature use | "We can't tell what we're getting for this" |
| **Onboarding gap** | Never reached the first value milestone; activation incomplete; setup done but no outcome | "We're still getting set up" past the target window |
| **Missing feature** | Usage concentrated in a narrow path, workaround tickets, competitor evaluation | A named capability blocking a named workflow |
| **Relationship loss** | Usage steady, engagement collapsed, exec touches at zero | New leadership, restructure, or champion departure |

Assign exactly one primary driver. Running two plays at once produces a scattered outreach that reads as panic to the customer.

**3. Run the matching save play.**

- **Unclear value.** Rebuild the ROI case against the criteria the customer stated at kickoff, not against your own metrics. Where criteria were never recorded, run a criteria-setting session first — the absence of an agreed measure of success *is* the risk.
- **Onboarding gap.** Deliver walkthroughs and refresher training, and re-run the onboarding path to the first value milestone rather than adding features ([Gainsight](https://www.gainsight.com/blog/customer-health-scores/)). Analyze the new-user funnel to find where users stall before activation, reduce friction with in-product checklists and walkthroughs, and send tailored communications to users stuck at a specific funnel step ([Gainsight](https://www.gainsight.com/essential-guide/product-management-metrics/dau-mau/)).
- **Missing feature.** Convert the gap into a dated product commitment or an explicit workaround with an owner. Do not offer a discount in place of a roadmap answer, because a discounted account with an unsolved blocker churns one cycle later at a lower ARR.
- **Relationship loss.** Re-establish an executive sponsor before any product conversation. Arm the surviving champion with shareable metrics, ROI calculations, and executive briefing material, and set monthly updates to a VP or C-level sponsor ([Rework](https://resources.rework.com/libraries/post-sale-management/land-and-expand-model)).

Across all four, trigger the automated layer as support, not substitute: automated outreach, follow-up tasks, and nurture enrollment ([Gainsight](https://www.gainsight.com/blog/customer-health-scores/)).

**4. Verify recovery, fix, re-verify.** Define the recovery criterion before starting the play, as a measurable threshold with a date — for example, weekly active seats back above 60% of licensed within 30 days, or the first value milestone completed within 21 days. Then:

- Re-score the account at the stated date.
- If the criterion is not met, do not extend the same play. Re-run step 2, because a failed play usually means the driver was misdiagnosed rather than under-executed.
- Only close the risk case when the criterion is met *and* the score has held for two consecutive refresh periods, because single-period recovery is frequently a campaign artifact.

Track intervention success so the program can be quantified: flag churn indicators, target flagged customers for CSM intervention, and track intervention success to quantify ROI ([Gainsight](http://www.gainsight.com/wpcms/wp-content/uploads/2015/07/6-Crushing-Your-QBR-Agenda-and-Template.pdf)).

**5. Instrument time to value.** Early churn is an onboarding problem, so measure TTV as the single onboarding KPI. Measure from **contract signature** to the **first value milestone** — the concrete outcome both parties agreed indicates success, such as first campaign sent, first report generated, or first workflow running live. Do not measure to "kickoff scheduled", "implementation began", or "setup complete", because a customer can be fully set up and still be waiting for value ([Valuecase](https://www.valuecase.com/articles/what-is-time-to-value)).

Non-negotiable measurement rules, because each one changes the number materially ([Valuecase](https://www.valuecase.com/articles/what-is-time-to-value)):
- Start at contract signature; starting at "kickoff scheduled" hides the handoff delay.
- Define the end milestone deliberately at the kickoff call, per customer.
- Aggregate with the **median, not the mean** — one 12-month enterprise rollout distorts the mean.
- Segment by plan, customer size, or onboarding type.
- Manage the trend: whether median TTV falls quarter over quarter.

There is no universal TTV benchmark; a single industry number is not useful, and your own median trend is the benchmark. Typical ranges are hours to days for low-touch or self-serve, a few weeks for mid-market B2B SaaS depending on data setup and stakeholder count, and weeks to months for enterprise rollouts involving data migration, integrations, and security reviews. Enterprise rollouts taking months can be healthy if TTV is steadily improving ([Valuecase](https://www.valuecase.com/articles/what-is-time-to-value)).

Attack the three drivers of long TTV, all of which are process problems rather than product problems ([Valuecase](https://www.valuecase.com/articles/what-is-time-to-value)):
1. **Handoff gaps** — sales closes, onboarding starts cold, goals and context are re-collected from scratch, and post-signature enthusiasm decays during the delay.
2. **Customer-side waiting** — data uploads, system access, and security sign-offs stall because nobody owns the follow-up.
3. **Invisible progress** — the plan lives in email threads and a spreadsheet, neither side can see what is done or who is blocking, and stalls go unnoticed until momentum is lost.

The shortest-TTV teams define the first value milestone at kickoff, template the process so design work is not repeated per customer, collect data through structured forms instead of email chains, give the customer one shared workspace where progress is visible, and automate follow-up and chasing ([Valuecase](https://www.valuecase.com/articles/what-is-time-to-value)).

**6. Attach journey red flags.** Map the six stages — Purchase, Onboarding, Adoption, Retention, Expansion, Advocacy — and for each milestone define what success looks like, then record the signs a customer is failing it as explicit red flags for the CS team ([Help Scout](https://www.helpscout.com/blog/customer-success-journey/)). Define milestone success as a *functional* outcome, not activity: the milestone "inviting colleagues" succeeds only when the invited colleagues accept the invitation and log in ([Help Scout](https://www.helpscout.com/blog/customer-success-journey/)). Using the product once and never again indicates the customer is not succeeding, regardless of the activation flag ([Help Scout](https://www.helpscout.com/blog/customer-success-journey/)).

Mine two additional early-warning sources: analyze support tickets for friction, and investigate knowledge-base articles receiving outsized traffic, which signals customers struggling with that specific feature ([Help Scout](https://www.helpscout.com/blog/customer-success-journey/)).

## Output format

Produce a save plan in this exact structure. The driver, recovery criterion, and check date are parsed into the risk register, so keep those fields; the narrative is yours.

```
# Save plan — <Account> — <ARR> — renewal <date>
Health: <score> (<band>), <delta> vs prior refresh   Override applied: <yes/no, condition>

## Diagnosis
Primary driver: <unclear value | onboarding gap | missing feature | relationship loss>
Evidence (data): <signals with dates>
Evidence (conversation): <who said what, when>
Rejected drivers and why: <one line each>

## Play
| Step | Action | Owner | Date |

## Recovery criterion
Metric: <metric>   Threshold: <value>   Check date: <date>
Hold requirement: score sustained for 2 consecutive refresh periods
If not met: return to Diagnosis, do not extend the same play

## Upstream fix
TTV for this cohort: median <n> days (signature -> <named first value milestone>)
Journey stage where the milestone failed: <stage>
Red flag added to the journey map: <flag>
```

## Gotchas

- Activation is not value. A customer can be fully set up and still be waiting for value, so an onboarding dashboard showing 100% setup completion is compatible with a churn in month five ([Valuecase](https://www.valuecase.com/articles/what-is-time-to-value)).
- Mean TTV is the wrong statistic and will hide a worsening onboarding process. Use the median, because one 12-month enterprise rollout distorts the mean enough to mask a rising median ([Valuecase](https://www.valuecase.com/articles/what-is-time-to-value)).
- Generic "we haven't seen you in a while" reminders are a false save. They bring casual users back for a day and increase MAU more than DAU, which worsens the stickiness ratio while the health score appears to improve ([Gainsight](https://www.gainsight.com/essential-guide/product-management-metrics/dau-mau/)).
- Milestone completion counted as activity rather than outcome produces false green. Invitations sent is not a milestone; invitations accepted and logged in is ([Help Scout](https://www.helpscout.com/blog/customer-success-journey/)).
- Do not push upsell or cross-sell into a save conversation. Customers should not be pushed with expansion opportunities indiscriminately ([Help Scout](https://www.helpscout.com/blog/customer-success-journey/)), and an expansion ask inside a save play confirms the customer's suspicion that the vendor is not listening.
- The score alone will miss stakeholder loss, since usage lags a sponsor departure by a full quarter. Configure a manual override for high-risk situations such as stakeholder loss or product underuse rather than waiting for telemetry to catch up ([Gainsight Communities](https://communities.gainsight.com/predictive-health-scoring-321/build-a-foundational-health-scoring-framework-using-dear-26486)).
- Save-play investment has a ceiling. GRR is table stakes with little direct correlation to growth, and the stated parity floor is **GRR at least 90%** against a median of **91%** ([SaaS Capital](https://www.saas-capital.com/wp-content/uploads/2025/09/RB32WS1-2025-B2B-SaaS-Retention-Benchmarks.pdf)). Once GRR clears 90%, additional retention headcount returns less than the same headcount pointed at expansion, since median NRR is **101%** and the growth separation happens above 110% ([SaaS Capital](https://www.saas-capital.com/wp-content/uploads/2025/09/RB32WS1-2025-B2B-SaaS-Retention-Benchmarks.pdf)).
- Month-to-month contracts run **89% GRR** versus **94%** for multi-year ([SaaS Capital](https://www.saas-capital.com/wp-content/uploads/2025/09/RB32WS1-2025-B2B-SaaS-Retention-Benchmarks.pdf)), so a book weighted toward monthly terms will show elevated churn that no save play fixes; that is a contracting problem, not a CS execution problem.

What does the Preventing Churn skill do?

Diagnoses the churn driver, runs the matching save play, and fixes time-to-value at the source

Where does the Preventing Churn skill come from?

Report research library. It was written by The Revenue AI Report against a 12 criterion quality rubric and graded in an independent scoring pass.

Why was the Preventing Churn skill chosen for this library?

Four-driver diagnosis table with distinct data and conversation signatures, and a recovery criterion that re-diagnoses rather than extending a failed play.

When should the Preventing Churn skill not be used?

Do not use it for: What weights and thresholds should our customer health scorecard use? Or: Put together the QBR deck for next week's executive business review

How do I install the Preventing Churn SKILL.md file?

Download the file, create a folder named exactly preventing-churn inside your agent's skills directory, and save the file inside it as SKILL.md. The agent loads it when a request matches the description.