What signals belong in a customer health score?
The reliable inputs cluster into six buckets: product usage trend (weekly active users, feature breadth, depth of use), support load (ticket volume, severity, time to resolve), relationship (NPS, CSAT, executive sponsor engagement, champion tenure), commercial (invoice status, days past due, open credits), lifecycle (days to renewal, contract value trend), and outcome (goals from the kickoff, business value delivered). Avoid stuffing in every data point you can wire up. Four to seven weighted signals tuned to your segment outperform a 20-input model that nobody trusts or can explain to a CSM in a stand-up.
Read the full answer →
How do you weight the signals without guessing?
Start with CSM intuition for a v1, then validate against six to twelve months of actual renewal outcomes. Pull every account that churned or downgraded, score them with your proposed weights, and check whether the model would have flagged them at least 90 days before the exit. If reds caught 70 percent or more of the churn and the false positive rate on greens stayed under 15 percent, the weights are directionally right. Re-tune quarterly. Weighting is a product decision, not a math problem, so write the logic in plain English in the admin and keep it reviewable.
Should the score be red/yellow/green or a 0 to 100 number?
RYG wins for daily CSM work because it maps cleanly to a worklist: reds get an executive save plan, yellows get a playbook, greens get expansion and advocacy motions. A 0 to 100 number is better for leadership reporting because it shows distribution shifts, cohort trends, and movement inside a band. The pragmatic answer is both: compute an underlying 0 to 100 score and bucket it into RYG with named thresholds, so operators see traffic lights and analysts see the continuous number behind them.
How often should the score refresh?
Transactional inputs like product usage, support tickets, and invoice status should feed the score daily or in near real time. Survey inputs like NPS and CSAT update when a new response lands. Lagging inputs like exec engagement and QBR cadence update weekly. The composite score itself should recompute at least nightly, with a visible last-updated timestamp on the account record. CSMs lose trust fast when a red lingers for two weeks after the ticket cleared, or when a renewal risk flips green the day after a champion leaves.
Who owns the health score in a CS org?
The CSM owns the score for their book: validating it against reality, logging save plans on reds, and keeping the account record current. CS Ops owns the model: signal definitions, weights, thresholds, and refresh logic. The CS leader owns the output: distribution across the portfolio, trend quarter over quarter, and the retention forecast the board sees. Product and support feed signals in. If no single person owns the model, the weights drift, nobody trusts the number, and the whole system quietly stops driving behavior inside six months.
How does executive engagement get measured?
Track three concrete signals, not vibes. First, cadence: has the exec sponsor met with the CSM or AE in the last 60 or 90 days, logged on the account timeline. Second, response: are emails to the sponsor acknowledged within a defined SLA. Third, participation: has the sponsor shown up to the last scheduled business review. A champion-change event (sponsor leaves the company or role) should drop the exec engagement signal immediately and raise an alert. All three are activity-based inputs the CRM already captures if calendar and email sync are turned on.
What are the most common ways health scores fail?
Four failure modes repeat. First, false positives: the score flags greens that churn because the model missed a sponsor change or a quiet usage decline inside one key team. Second, lagging indicators: the score moves 30 days after the signal, so the save window is already closed. Third, no action: reds pile up on a dashboard and nobody runs a playbook, so the score becomes decorative. Fourth, score inflation: CSMs manually override reds to greens to clean up the view, and the model loses signal. The fix is governance, not more signals.
How do you handle lagging versus leading indicators?
Leading indicators move before the renewal is at risk: weekly active user trend, feature adoption velocity, champion response time, support severity mix. Lagging indicators confirm risk after it has formed: NPS drop, invoice past due, missed QBR. A healthy score weights leading indicators at roughly 60 to 70 percent of the composite so the model fires early. Keep the lagging inputs in the mix as confirmation and for reporting, but do not let them drive the color on the account. If the score only moves on lagging data, it is a scoreboard, not a steering wheel.
How does the health score integrate with QBRs?
The score should seed the QBR agenda, not be the whole conversation. Open with the trend over the last two quarters, the signals that moved, and the outcomes delivered against the goals set at kickoff. Reds and yellows anchor the risk conversation and the save plan. Greens anchor expansion and advocacy asks. The CSM pre-reads the score with the AE and the exec sponsor one week before the meeting so surprises get handled offline. The QBR deck should pull the score and the signals directly from the CRM, not a separate slide the CSM retypes.
Can Strkr AI generate or tune a health score automatically?
Strkr AI assists with two parts of the job. It drafts a v1 signal set and weights from your segment, book size, and historical renewal data, which gets a new team to a working score in days rather than months. It also surfaces anomalies: accounts where the composite score and the raw signals disagree, or where a quiet signal like champion response time is drifting ahead of the headline number. The weights, thresholds, and model logic stay editable in the admin so CS Ops always owns the final rules and can explain the score to any CSM.
How many health score tiers should you use?
Three tiers (red, yellow, green) is the sweet spot for operators. Five tiers (critical, at risk, watch, healthy, advocate) adds precision for reporting but creates ambiguity in daily work: the difference between watch and healthy is rarely actionable. If leadership wants more granularity, keep three operator tiers and layer a separate advocate flag for green accounts worth a case study or reference ask. Fewer tiers also makes the save plan playbook library easier to build, since each tier maps to a defined motion rather than a sliding scale.
What does a good health score dashboard look like?
The dashboard answers four questions in one view. How is the portfolio distributed across tiers this week versus last. Which accounts moved tiers in the last 30 days and why. Which reds and yellows have a save plan with a named owner and a target date. Which greens are candidates for expansion or advocacy. Each row drills into the account record with the full signal breakdown. Avoid separate reporting tools: the dashboard belongs inside the CRM so CSMs act from the same surface where they log notes and update next steps.
