A customer health score is the single number that tells Customer Success which accounts to save, which to grow, and which to leave alone this week. Done well, it drives the renewal playbook, the QBR agenda, and the exec briefing without a side spreadsheet. Done poorly, it is a vanity gauge that lights up green while churn builds underneath. This guide walks the full launch cycle: pick the right signal mix, weight and threshold the score, wire a daily refresh, train CS on the playbooks it triggers, and tune the model every quarter so the color on the dashboard matches the reality in the account.
Time: 1 week to design, 2-3 weeks to pilot and launch
Start by writing down every signal CS believes drives renewal, then cut the list to the handful that correlate to historical outcomes. The six that earn their place in almost every B2B SaaS book are product usage depth, support ticket pressure, NPS or CSAT trend, invoice and payment status, executive engagement, and time to renewal. Usage depth is weekly active seats and depth of feature adoption, not just logins. Support pressure is ticket volume, severity mix, and time open. NPS covers sentiment. Invoices catch late payers, which are a leading indicator of churn no one wants to admit. Exec engagement counts QBR attendance and sponsor meetings in the last ninety days. Renewal proximity surfaces accounts that need attention before the window closes. If a signal does not change CS behavior, it does not belong in the score.
The two scales worth using are a 0 to 100 numeric score or a red, yellow, green traffic light. Pick based on how CS already works. A 0 to 100 score reads well on dashboards, sorts cleanly in lists, and supports fine-grained thresholds for playbooks. A traffic light reads faster at a glance and forces decisive action because there is no ambiguous middle. Many teams run both: compute the 0 to 100 under the hood and surface the band color on account cards and queues. Avoid five-point and ten-point scales; they look precise and get read like a traffic light anyway. Whatever scale wins, publish it in one place and ban every side spreadsheet that invents a parallel score within sixty days of launch.
Weights decide which signal dominates when accounts give off mixed messages. Start with a defensible first pass: usage depth and exec engagement carry the heaviest weight because they predict renewal earliest. Support pressure and NPS come next because they catch active dissatisfaction. Invoice status and renewal proximity act as modifiers rather than core drivers because they are binary and time-bound. A sensible starting split is roughly thirty percent usage, twenty percent engagement, twenty percent support, fifteen percent NPS, and the remaining fifteen percent across invoice status and renewal timing. Then set band thresholds that match CS capacity. Red should capture the accounts a CSM must touch this week, not every account that looks a little off. If more than fifteen or twenty percent of the book is red, the threshold is wrong, not the book.
The score is only useful if it refreshes on a cadence that catches drift before the renewal window. Daily is the right default for mid-market and enterprise books. Weekly is enough for very small customer counts or when usage data lands slowly. Build the pipeline to pull each signal from its system of record, normalize it to a common 0 to 100 subscale, apply the weights, and write the result back onto the account record with a timestamp and a version tag. Store the component subscores alongside the composite so CS can see why the color is what it is. Log every computation so a disputed score can be audited later. Do not let the score live only in a BI tool; it must land on the account record where CSMs already work or no one will trust it.
The score has to trigger action or it is decoration. Build the renewal playbook so each band maps to a defined response. Red accounts inside the renewal window get an executive escalation, a save plan with a named owner, and a weekly check-in until the color moves or the renewal closes. Yellow accounts get a targeted intervention from the CSM: a usage review, a QBR schedule, or a champion check depending on which component dropped. Green accounts inside the renewal window get a confirm-and-expand motion rather than a save motion. Attach the playbook steps directly to the account card so a CSM landing on a red account sees the next action, not just the score. The dashboard is the pull, the playbook is the push.
The score should drive upstream conversations too. Every QBR agenda for the next quarter should pull from the score: the component subscores show the customer where the relationship is strong and where it is slipping, and the trend over the last two quarters shows whether CS has been moving the needle. The monthly exec briefing should roll the book up by segment and band so leadership sees aggregate health, not just the save-list exceptions. Use the score to prioritize which accounts the CCO visits in person. One score, three audiences, no duplicate reporting. If a QBR deck or exec briefing calculates a different health view, consolidate it onto the shared score within the quarter.
A new scoring system without training is a loyalty program for the people who already liked it. Run a kickoff session with every CSM that walks through the signal set, the weighting, the thresholds, the daily refresh, and the playbook per band. Walk through five real accounts live so the room sees how the number behaves on familiar names. Open a dispute path so a CSM who disagrees with a score can flag it with evidence, and route disputes to the owner of the score, not the queue. Treat the first thirty days as a calibration period: the goal is not to defend the model, it is to find the places where the model is wrong and fix them. Office hours twice a week in month one closes the loop before the score loses the room.
The score is a living model, not a stone tablet. Every quarter, run a short retro: which signals predicted churn and expansion, which were noise, and which were gamed. Measure precision and recall on the last cohort. Did the accounts that churned actually trend red in the ninety days before they left. Did the accounts that expanded trend green in the ninety days before the deal closed. If precision is low, tighten the thresholds. If recall is low, add or re-weight a signal. Hold the core formula stable so quarter-over-quarter trends remain comparable, but tune weights and thresholds aggressively. Publish the changelog with the date, the change, and the reason, so six months in the team knows why the model behaves the way it does.
A customer health score is a composite number, typically rendered as 0 to 100 or as a red, yellow, green band, that blends product usage, support ticket pressure, NPS, invoice status, executive engagement, and renewal proximity into a single signal of how likely an account is to renew and expand. Customer Success owns it and uses it to trigger the renewal playbook, the QBR agenda, and the exec briefing.
Six signals cover almost every B2B SaaS book: product usage depth (weekly active seats and feature adoption), support ticket pressure (volume, severity, age), NPS or CSAT trend, invoice and payment status, executive engagement (QBR attendance and sponsor meetings), and time to renewal. Pick the handful that correlate to your historical churn and expansion outcomes and cut the rest. Fewer sharp signals beats a long noisy list every time.
Daily is the right default for mid-market and enterprise books because it catches drift inside the week a CSM can still act on it. Weekly is enough for very small customer counts or when usage data lands on a slow pipeline. Every refresh should timestamp the computation and store the component subscores alongside the composite so CS can audit why the color is what it is and compare trends across releases.
Customer Success Operations owns the model end to end: the signal definitions, the weighting, the thresholds, the daily pipeline, the dispute path, and the quarterly retro. The CCO or VP of CS owns the playbooks that fire off each band and the compensation implications. Finance co-signs the invoice signal. One owner, two co-signers, no committee, and the score lives on the CRM account record where CSMs already work.
Measure three signals. First, precision and recall against the last churn and expansion cohort: did the churned accounts trend red in the ninety days before they left, and did the expansion accounts trend green. Second, playbook execution rate: did every red account have a named save-plan owner and a documented next action. Third, CS trust: are CSMs opening the score before pipeline reviews or are they still running private spreadsheets. If any slips, run a short retro before quarter-end.
Not in the first two quarters. Tying compensation to a model that has not been tuned yet guarantees the model gets gamed before it gets accurate. Pay CSMs on retention and expansion outcomes while the score matures, then once precision and recall are stable, consider tying a small portion of compensation to playbook execution against the score, not to the color of the score itself. Pay on outcomes, coach on activities, and keep the score as a diagnostic.
Strkr captures product usage, support pressure, NPS, invoice status, exec engagement, and renewal timing on the same account record, so CS Ops can design the score once and every CSM reads it from the account card where the save plan lives.
No sales call, no migration consultant, no four-month implementation. Enter your card, get 14 days of the full Pro tier, cancel any time before day 14 with zero charge. Spin up a workspace, import your CSV, and have something useful before lunch.
