How I approach customer health

Most health scores are a single number with no explanation. A CSM cannot act on “68”. By the time a score turns red, the causes are usually months old: a champion left, the rollout stalled at one team, or nobody ever agreed how value would be measured.

The second problem is data quality. Teams silently trust dashboards where one signal has not refreshed in two weeks, then make a renewal decision on it.

Score the account from five weighted signals, then show the arithmetic, the evidence and the freshness behind every component. Prioritise a daily queue rather than asking a CSM to scan a table of 12 to 80 accounts. Connect every risk to a concrete next action, and record what was done so the engagement signal stays honest.

Adoption tells you whether the product is used. Outcomes tell you whether it mattered. Engagement tells you whether the relationship can survive a problem. Support tells you whether friction is accumulating. Sentiment usually moves first and moves quietly.

• Product adoption30%
• Business outcomes25%
• Customer engagement20%
• Support experience15%
• Customer sentiment10%

Health Score = adoption × 0.30 + outcomes × 0.25 + engagement × 0.20 + support × 0.15 + sentiment × 0.10. Green 75–100, Amber 50–74, Red below 50. Status is always derived, never hardcoded.

A number that decides renewal risk has to be auditable. Rules can be explained to a customer, challenged by a colleague and corrected when the business changes. A model that cannot show its arithmetic invites the team to either over-trust or ignore it.

Stale data therefore lowers confidence, not the score itself. Penalising the number would create false risk and train the team to distrust the dashboard.

AI is used for explanation, drafting and prioritisation support — never for the decision. Every generated analysis states what changed, the likely root cause, an alternative explanation, what information is missing, and what must be validated with the customer before acting.

Human review required. This analysis supports, but does not replace, CSM judgement.

Earlier risk detection protects renewal revenue. Evidence-based conversations shorten renewal cycles because value is already documented. A prioritised daily queue means a CSM with 30 to 80 accounts spends attention where the ARR and the risk actually sit, instead of spreading effort evenly.

Customer Success and Onboarding professional with 7+ years of customer-facing experience across enterprise technology, Finance EdTech, cybersecurity, and support operations.

At an early-stage Finance EdTech business, I supported a portfolio of 55+ customers across onboarding, engagement, progress tracking, re-engagement, and customer advocacy. I reduced time-to-value from three weeks to one through structured onboarding and first-value milestones.

At TCS, I supported the onboarding and access enablement of 2,000+ enterprise users per month while maintaining 92%+ CSAT and 100% SLA adherence.