Customer Health Score: A Worked Model You Can Copy

Cover of the GrowthStackPulse guide Customer Health Score: A Worked Model You Can Copy

Customer Health Score: A Worked Model You Can Copy

Contents

A customer health score is one number from 0 to 100 for each account. It combines usage, support, sentiment and relationship signals to show who is likely to renew, expand or need help.

It turns scattered signals into one decision per account: who the CSM calls first. Customer success teams often call it a customer success health score and use it to plan the week.

Below you get a four-input, 100-point model and three sample accounts scored end to end, which double as a spreadsheet template. A back-test then sets the bands from your own churn.

What Is a Customer Health Score?

A customer health score is a single 0 to 100 rating of how an account is doing. It decides where CSM time goes before renewal. In B2B it works at account level, and user-level signals such as logins and seats roll up into it.

It is not a churn prediction, a single NPS reading or a customer health check. It sits alongside the SaaS retention metrics, not in place of them.

Teams also say customer health scorecard when the inputs sit side by side, and customer health scoring for the process. You will also see customer health index and client health score. B2B teams scoring at company level say account health score.

Teams use the score for three jobs:

  • Prioritize CSM time: the lowest and fastest-falling scores get the first calls.
  • Trigger a play before renewal: a band change starts a defined action instead of a hunch.
  • Share one view across CS, sales and finance: existing customers matter to all three. In a 2025 survey of 793 SaaS and technology leaders, 74% said most revenue comes from them.

Customer Health Score Metrics: What to Feed the Score

The inputs fall into five signal families, and each family lives in a different system.

The table shows each family, example metrics, where the data lives and whether the signal leads or lags renewal. Leading or lagging is our judgment.

Signal familyExample metricsWhere the data livesLeading or lagging
Product usageFeature adoption, active seats, time to valueProduct analyticsLeading
SupportTicket volume, escalations, time to first resolutionHelp deskLeading
SentimentNPS, CSAT, CESSurvey toolMixed; NPS lags
RelationshipSponsor engagement, QBR attendance, champion changeCRMLeading
FinancialRenewal date, payment status, expansion or downgradeBilling or CRMLagging

CSAT measures satisfaction with one interaction, NPS measures likelihood to recommend and CES measures how much effort a task took. In a survey of 105 software executives, AskNicely, Dec 2025 found that 87% track CSAT, 66% track CES and 47% track NPS.

Use 4 to 6 inputs and start with fewer. The rule for how to measure customer health is simple: pick signals that moved before your past renewals, not signals that are easy to export.

Takeaway: If you can score only one input, start with product usage. Billing status only says the account pays, not that it uses the product.

How to Calculate a Health Score in Five Steps

You calculate a health score in five steps.

  1. Pick four inputs, one from usage, support, sentiment and relationship; keep an input only if it moved before past renewals.
  2. Scale each input to a 0 to 100 score, such as active seats as a share of seats bought. Cap it at 0 and 100. For an input where lower is better, such as escalations, invert the scale so fewer escalations score higher.
  3. Set weights that add to 100, and give the most weight to the input that moved earliest.
  4. Multiply each input score by its weight and add the four results; the sum is the account’s score.
  5. Map the total to three bands, and treat the cut-offs as a guess until you back-test them.

Run these steps in a spreadsheet first, with one row per account, and you have a working customer health score template.

The Health Score Formula and Starting Weights

The customer health score formula is the sum of weight times input score for each input. Scale every input to 0 to 100 and make the weights add to 100%.

`score = sum of (weight x input score)`

With a year of renewal history, start with weights of 40, 25, 20 and 15. Then move the band cut-offs until the at-risk band catches most of last year’s churners.

The table gives each input’s starting weight and the reason it carries that weight.

InputStarting weightWhy it carries that weight
Usage40Usage moves first, before support tickets or survey answers
Support25Support friction shows early pain
Sentiment20Surveys are periodic and sparse
Relationship15It is the hardest input to measure

No standard or benchmark backs this split, and we found no published evidence for it. The back-test below shows how to check the weights against churn.

Worked Example: Three Accounts Scored End to End

Most customer health score examples stop at a formula, so here are three accounts scored end to end. The model uses four inputs with weights of 40, 25, 20 and 15.

Bands are 71 to 100 Healthy, 41 to 70 Watch and 0 to 40 At risk. The table gives each account’s four input scores, its weighted total, its band and the next move.

AccountUsageSupportSentimentRelationshipWeighted totalBandNext move
Northwind8580757280HealthyProtect the account and look for expansion
Brightloop9020423054WatchReview the support history and call the sponsor
Cobalt3045352033At riskReach out this week and name an owner for a save plan

Brightloop’s weighted parts are 36.0 for usage, 5.0 for support, 8.4 for sentiment and 4.5 for relationship. They add to 53.9, which rounds to 54.

Brightloop’s usage of 90 alone would read Healthy, but its three weak inputs cost it 46 of 100 points. That is why one input is never the score.

You can rebuild this table as a customer health score Excel template in about five minutes. Use one row per account and one SUMPRODUCT of the weights row and the input row.

Move to a CRM or customer health score software once the model holds.

The three accounts and their input scores are invented for illustration; the weighted totals and bands were computed by script from the starting weights above.

Takeaway: A busy account is not a healthy account, so read the weak inputs, not only the total.

What Is a Good Health Score?

A good health score is one above the cut-off your own churn history supports, and no number is universal. In the example model, 71 or above is Healthy.

The table compares four display formats by who each suits and what to watch.

FormatBest forWatch out for
Percentage, 0 to 100Data-mature teams and forecastingOver-precision can hide the cause, and the weights need validating
Color: green, yellow, redLarge account volumes and playbook alertsBroad bands need review
Letter grade, A to FInternal and executive reviewsLittle granularity
RankPortfolio and resource allocationRanks shift with the account mix and carry no context

The three bands used here, 71 to 100, 41 to 70 and 0 to 40, are a common three-band split and illustrative only. The back-test below shows how to move them.

Benchmarks to Calibrate Your Inputs

Published medians check your input scores, not your band cut-offs.

The table puts Survicate’s NPS and CSAT medians Survicate, Dec 2025 and Customerscore.io’s vendor study Customerscore.io, Jun 2026 beside the decision each one changes.

BenchmarkValue and scopeDecision it changes
Median NPS, B2B software29, against 47 for B2C software; software segment of Survicate’s 2025 datasetAn NPS cut-off of 50 for “healthy” sits 21 points above the B2B median, so a typical B2B software company would fall short; set the sentiment input relative to 29
Median CSAT80.0%; half of 1,259 surveys from 370 companies fall between 64.8% and 90.0%; all industries, Survicate’s CSAT benchmarkA CSAT of 80 is normal, not a strong signal
Disengaged share of the “active” base23% at the median B2B SaaS company, some above 35%; 44,000-user study, vendor definition“Active” billing status hides nearly one account in four with no product activity; weight usage over billing

How to Set Your Bands from Your Own Churn History

Band cut-offs are a guess until you compare them with real renewals, and most guides stop at “validate with history”. A three-step back-test turns the guess into a check:

  1. Score last year’s renewed and churned accounts as they stood 90 days before renewal.
  2. Compare the churn rate by band.
  3. Move the cut-off until the at-risk band catches most churners without flagging most of the base.

The worked example scores 80 accounts, and 17 of them churned.

BandAccountsChurnedChurn rate
Healthy (71 to 100)4436.8%
Watch (41 to 70)22627.3%
At risk (0 to 40)14857.1%
Total801721.2%

The at-risk band catches only 8 of the 17 churners, or 47.1%. Inside Watch, scores 41 to 55 churn at 55.6% (5 of 9) and scores 56 to 70 at 7.7% (1 of 13).

Raising the at-risk cut-off from 40 to 55 flags 23 of 80 accounts (28.7%) and catches 13 of 17 churners (76.5%).

If the middle band churns like the bottom band, merge them or raise the cut-off. Take a different back-test where Watch churned 11 of 22 (50.0%) and At risk churned 8 of 14 (57.1%). Merged, the band churns 19 of 36 (52.8%).

Then re-check the weights the same way: an input that does not differ between renewed and churned accounts loses weight. None of this shows that a score predicts churn at a known rate; it shows how your own bands behaved.

The 80 accounts are invented; band counts, churn rates and the effect of moving the cut-off were computed by script. Run the same steps on your own renewal list.

How to Segment the Score by Customer Stage and Size

One weighting does not fit every account, so keep the same four inputs and shift the weights by customer stage.

Churn does not peak at onboarding. In a Customerscore.io vendor study of 44,000 users, it dips around months 3 to 6 and peaks at 6 to 12 months.

The table shows which inputs weigh more at each stage and why.

StageWhich inputs weigh moreWhy
OnboardingSetup and adoption inputsEarly value decides the rest of the relationship
AdoptionDepth of usage, seats used against seats boughtDeep use shows the product is part of the workflow
RenewalRelationship and support, such as sponsor engagementThe renewal decision rests on the sponsor and recent support experience

In B2B, flag the account when the main admin or champion drops out. An account can look healthy on average while its one champion has already left.

Keep one score until the back-test shows segments churning differently by band. Then split by stage or size.

Our take on size: high-touch accounts with a named sponsor can carry more relationship weight, while low-touch accounts with no sponsor contact lean on usage. We found no sourced account-count threshold for when to split, so let your own churn data decide.

How to Turn the Score into Action

Turn the score into action by mapping each band and each sharp fall to a play, an owner and a time limit.

Band or trendPlayOwnerTime limit
HealthyProtect the account, watch for drift, ask for advocacyCSMQuarterly review
WatchProactive outreach to surface blockersCSMWithin two weeks
At riskSame-week call, save plan, executive sponsor if neededCSM with CS leadThis week
Score fell 10+ pointsReview the reasonCSMWithin a few days
Flat for two cyclesDiagnose stalled adoptionCSMNext cycle
Healthy and usage growingExpansion conversationAccount manager or AEThis quarter

Owners and time limits are starting values to adjust, and the 10-point fall is a design choice, not a finding.

Where the Score Should Live

The score belongs in three places, each with its own job:

  • A CRM account property: the number and band sit on the record the CSM opens first.
  • A Slack alert on band change: the owner hears about a move the day it happens.
  • A customer health score dashboard: the team reviews every account’s band in the weekly meeting.

Write the reason beside the score on the CRM record, next to the suggested play. An example reason is that the main admin has not logged in for two weeks.

Rules First, AI Later

Start with rules your CSMs can explain, and add an AI customer health score only when you have the history to check it. Only 22% of organizations use AI for health scoring today, per TSIA, May 2026.

Compare it with the rules score on the same accounts, using the back-test above. AI lead scoring covers the same choice for leads. A model nobody can explain gets ignored, just like a bad rules score.

Five Mistakes That Make a Health Score Useless

Five mistakes make a health score useless, and one review routine keeps them fixed.

The table gives each mistake, why it fails and the fix; the sixth row is the routine.

MistakeWhy it failsFix
Logins counted as engagementAn “active” account can do nothing in the product; the Customerscore.io figure above shows how many sit idleScore feature adoption and seats used, not logins
One threshold for every segmentOnboarding and renewal accounts behave differentlyShift the weights by stage, as in the segmentation section
Too many inputsEach extra input dilutes the ones that matterKeep 4 to 6 inputs
A static snapshotA 75 that was 90 last month is a warningTrack the change and apply the 10-point rule
Weights never checked against churnIn our worked back-test, one cut-off move changed how many churners the at-risk band caughtRun the back-test before you trust the weights
The routine: no review cadenceBands and weights drift as products and customers changeRe-run the back-test every quarter

Takeaway: Check the weights first: the stage shifts, the 10-point rule and the band cut-offs all assume a model that has been back-tested.

Build the Score, Then Check It Against Your Churn

Build the four-input customer health score in a spreadsheet this week, and let last year’s renewals set the bands. Count the lost accounts in each band in the churn calculator before you trust a cut-off. Which account would you score first, the one you are sure of or the one you are worried about?

Frequently Asked Questions

How to calculate customer health score in a spreadsheet?

Pick four inputs, scale each to 0 to 100, set weights that add to 100, multiply and add, then map the total to bands. Cap each input at 0 and 100 first, with MAX(0, MIN(100, raw)) in a spreadsheet. Uncapped, a usage input at 130% of target adds 12.0 points (69.9 against 57.9). With every other input at 100, one input at 150 can push the total to 120, off the scale.

What is a CX scorecard?

A CX scorecard is a set of customer experience metrics shown side by side, such as NPS, CSAT and CES. A health score does something different: it rolls chosen inputs into one number per account. The two work together, because a scorecard can carry the health score as one row, next to the metrics that feed it.

What is a good health score number?

No number is universal because the right cut-off depends on your own churn, and even published example bands differ. One vendor’s example puts Healthy at 71 to 100, At Risk at 31 to 70 and Critical at 0 to 30. Our example ends At risk at 40. That gap shows bands are a choice, so back-test yours against your own renewals instead of copying someone’s cut-offs.

What is a good CES score?

There is no confirmed public median for customer effort score in B2B software, so there is no number to aim for. CES measures the effort a customer spent, so lower effort is better. Read it against the account’s own trend and the scale your survey uses.

What are the top 3 KPIs for customer success?

In a 2024 vendor survey of over 1,000 customer success leaders, respondents ranked churn rate first, then NRR and GRR. GRR shows leakage from churn and downgrades, and NRR adds expansion. The share of accounts in the Healthy band is a leading indicator that tends to move before all three in a scored base.

What is the difference between a customer health score and a churn score?

A churn score predicts whether an account will leave, while a customer health score describes how the account is doing now and drives the play. Check either one with the back-test: compare each account’s score 90 days before renewal with what happened at renewal.

What is the difference between a customer health score and a customer health check?

A customer health check is a direct conversation with a high-value customer to make sure they are happy and do not need help. A customer health score is a number built from account data. The score tells you who to call, and the check finds out why they are struggling. Use the score to choose which accounts get a check first.

Is a customer health score worth it for a small team?

Yes, once you have renewal history. Without past renewals, the weights are guesses and the bands have nothing to be checked against. A small team can start with a spreadsheet of accounts it knows renewed or churned. Score them as they stood 90 days before renewal, and see whether the scores separate the two groups. If they do not, change the inputs before you add software.

What are the 5 key CX metrics?

Five common CX metrics are NPS, CSAT, CES, churn rate and time to first resolution. NPS measures likelihood to recommend, CSAT measures satisfaction after an interaction and CES measures effort. Churn rate is the share of customers lost in a period. Time to first resolution is how long a support request waits for its first answer.

J

Who writes this

Jakub Zachveja

Founder and editor

Marketing specialist working in B2B SaaS, with a bias toward work that shows up in revenue rather than in visibility metrics. Covers the whole go-to-market stack, from what the tools cost to how the pipeline is run.