The Renewal You Lost Because the Health Score Was Green

It is the 14th of the quarter, 10:18 AM. Your CSM opens the customer health dashboard in Gainsight. The $146K account renewing in 38 days reads green. Login activity over 30 days reads 412 sessions. Support ticket count reads 4, all resolved. NPS from the last pulse reads 8. She marks the QBR as healthy in the renewal forecast, moves the slot to next Tuesday, and starts prepping the upsell pitch for seats 80 through 120.

On Friday at 3:47 PM the procurement contact sends a 90-day non-renewal notice. The reason field on the offboarding call reads "internal consolidation." Your CRO walks into Monday's forecast meeting and reads the loss off the slide. The CSM pulls the account history and finds the signals. The champion left in week three of last quarter. The new VP of Engineering ran a vendor review in week six and never opened a ticket. Three power users dropped from daily to weekly login in week eight. The 412 sessions were 38 service-account API calls a day from a Zapier integration nobody owned.

The dashboard never moved. The function that should have caught it never existed. The score formula was built by a consultant in 2024 and the integration only reads login count and ticket count. Eighteen months of product changes, two pricing updates, and a new buyer persona later, the formula still ships the same green dot on the morning a customer churns.

Customer health is a function. Most Series B and C teams have not staffed it because Gainsight or Vitally is installed, a score formula was written 18 months ago, and somebody on the CS team checks the colors before the QBR. The function lives in the gap between the CSM who watches 32 accounts and reviews each one once a month, the support lead who closes tickets without tagging sentiment, the product analyst who pulls feature adoption on request, the AE who logs the champion change in Salesforce three weeks late, and the renewal calendar that surfaces an account 60 days out. On the org chart it sits inside Customer Success. In the calendar it eats zero hours because nobody owns the synthesis.

The green dashboard nobody validated

Pull every signal the score formula does not read. The CRM contact-role history for champion turnover. The Outreach and Gong call frequency by stakeholder. The support ticket text for sentiment and unresolved frustration, beyond raw ticket count. The product analytics for feature breadth, beyond raw session count. The billing system for plan downgrades and seat reductions. The community Slack for the customer's posts. The shared Notion or Confluence the customer stopped opening.

Most teams find 8 to 14 leading indicators of churn that are not in the health score formula. Walk the account. The 412 sessions trace to a single service account once you filter by user_id. The four resolved tickets contain the phrase "this is the third time" in two of them. The champion's calendar invite to the last QBR was declined and the new VP of Engineering was never added. The Slack channel went from 14 customer messages a week in February to two in May.

The CSM saw none of this because none of it lives in Gainsight. The team that should own this knows it is broken. The CSM watches the dashboard read green on three of the last four churned logos. The CRO marks customer health unreliable on the board pack and uses a gut-feel forecast instead. The head of CS rebuilds the score formula every six months and the new version still misses the same shape. The function sits unstaffed while gross retention drifts down a point a quarter.

Hiring a CS ops lead is the slow answer

The textbook fix is a customer success operations lead or a retention analyst. Loaded comp in the US runs $120K to $180K a year. Months one through three go to rebuilding the score formula, mapping data sources, and reconciling the CRM to the product analytics. Months four through six are when the first cohort analysis ships and the first three at-risk accounts get flagged before the renewal window opens.

The fractional version is faster to start and stops at the same wall. Five to nine thousand a month buys ten hours a week of senior CS ops time and a refreshed score formula in 60 days. The renewal forecast accuracy moves up 8 to 12 points. The 40 enterprise accounts get a manual quarterly review. The 280 mid-market accounts stay on the old formula because the hours run out at the top 40.

Both versions assume the work is human bottleneck work. The work itself is reading every support ticket for sentiment within a day of close, mapping champion changes from Salesforce contact roles weekly, filtering session counts by user type and segment, and pulling feature adoption breadth per account per week. Cross-reference billing changes against usage trends. Roll the signals up into a per-account narrative the CSM can read in 90 seconds before the renewal call. On 320 accounts that is 280 to 440 hours of senior analyst work a quarter, and no CS ops lead clears that pile while also running the QBR cadence and the renewal motion.

What a fractional AI customer health function does

Hand the CRM, the product analytics, the support ticket text, the Gong library, the billing system, the community Slack, and the renewal calendar to a fractional AI agent that runs on a continuous cadence. The agent does the work a CS ops analyst, a support lead, and a product analyst would do together. The cadence is daily on new signals, weekly on per-account narratives, monthly on cohort drift, quarterly on score formula recalibration. The CSM stops trusting the green dot.

Every signal read, beyond the ones in the formula. Support tickets get sentiment-scored on close. Champion changes in Salesforce surface within a day of the contact role update. Session counts get filtered by human user versus service account. Feature adoption breadth gets pulled per account per week. The $146K account would have flagged on the champion change in week three, not on the non-renewal notice in week 38.

Per-account narratives, not a single color. Every account gets a 90-second written brief refreshed weekly. The brief names the three leading indicators that moved, the stakeholders by name, the open and closed tickets with sentiment, the feature adoption trend, and a confidence band on the renewal. The CSM reads the brief before every customer touch instead of glancing at a color.

Early warnings 90 days out, not 30. When two or more leading indicators degrade in the same 14-day window, the account moves to a watch list with a recommended play. The play might be a champion re-engagement, an exec sponsor outreach, a feature adoption push, or a pricing conversation. The same shape the win-loss interview function runs on the sales side, applied to the post-sale motion.

Renewal forecasts the CRO can defend. The agent rolls the per-account confidence bands into a renewal forecast for the next 90 and 180 days. Each accountable line on the forecast carries the leading indicators behind it. The CRO walks into the board meeting with a forecast that is wrong by 3 to 5 points instead of 12 to 18.

Score formula recalibrated quarterly from real outcomes. Every churned account gets a postmortem against the score it carried 90 days out. The signals that moved before the score did get added to the formula. The signals that never predicted anything get dropped. The formula stops being a 2024 artifact a consultant built and becomes a living model tuned to the last four quarters of outcomes.

The unit economics of a green dashboard that lies

A Series B company at $14M ARR running 320 customer accounts on a 2024 health score is renewing against a dashboard that misses the leading indicators on roughly one in four churned logos. Two preventable losses a quarter at $120K average ACV is $960K in gross retention a year. Net revenue retention slips three to six points. The CRO carries the variance on the board pack and the forecast accuracy never improves.

Layer in the people math. The CSM team, the head of CS, and the support lead spend a combined 40 to 80 hours a quarter on manual health reviews, ticket sentiment scans, and renewal prep. That is 160 to 320 hours a year of senior retention work on a playbook a fractional function covers, against a fully loaded hour of $180 to $240. The CS ops hire runs $150K to $230K loaded with a four to six month ramp before the first refreshed score ships. The fractional analyst stops at the top 40 accounts.

A 14-day sprint to stand up the agent runs in the low to mid five figures. Ongoing cost lands closer to one senior contractor than a CS ops org. The per-account narrative lands in week one. The first three at-risk accounts get flagged with a recommended play in week two. The score formula recalibrates against the last four quarters of churn data before the sprint closes.

What changes after the sprint

Picture the same 14th of the quarter, 10:18 AM moment, fourteen days after the sprint ships. The CSM opens the account brief for the $146K renewal. The brief reads yellow, not green. Champion turnover in week three is named. The 412 session count is broken out as 38 service account calls and four human user sessions a week. Two support tickets are flagged for unresolved sentiment. The new VP of Engineering is named as the missing stakeholder. The recommended play reads exec sponsor outreach within 14 days.

By Friday the AE has booked a 45-minute meeting with the new VP. The CSM has run a feature adoption push on the three modules the original champion was using. The exec sponsor on the EOI side has sent a one-paragraph email referencing two roadmap items the new VP cares about. The renewal forecast moves the account from 92 percent to 64 percent confidence with a flagged save play. The 90-day notice never gets sent because the conversation happened in week 35 instead of week 38.

If your customer health currently lives in a dashboard the CSM glances at before the QBR and a score formula a consultant built 18 months ago, the version where every signal gets read the day it lands and every account carries a 90-second narrative is fourteen days away. Customer health is a function. You can hire against it, you can outsource a fractional CS ops lead for it, or you can scope a sprint and have it running this month. The work is the same. The math is not.