Customer Health Scoring: The Playbook (Strategy-Level)

Most B2B SaaS founders meet "health scoring" as a Salesforce dashboard requirement around Series B: "we need to track customer health." The CS team builds a 1-10 score from a weighted formula of usage + NPS + tickets, displays it in a column on the customer list, and it sits there ignored — because nobody trusts it, nobody acts on it, and the score doesn't predict the things that actually matter (renewal, expansion, churn). Six months later, a different VP rebuilds it. The pattern repeats every 18 months.

A real customer health scoring program is something different: a measurement system that drives BEHAVIOR (CSM intervention, expansion plays, renewal forecasting) and is tightly tied to outcomes (NRR, churn, expansion, advocacy). Done well, health scores predict 80%+ of churn 60+ days in advance, drive 3-5x ROI on CSM time allocation, and become the operating system for customer success. Done badly, they're vanity metrics.

This is distinct from Customer Health Scoring (VibeWeek), which is the IMPLEMENTATION chat (data model, scoring engine, surfacing). This article is the STRATEGY playbook: deciding what to score, how to weight, how to act, who owns, and how to make the program produce ROI rather than dashboard theater.

What Done Looks Like

A working customer health scoring program produces:

A score (categorical green/yellow/red OR numeric 0-100) per customer
Score predicts churn ≥60 days ahead with 70%+ accuracy
Score predicts expansion likelihood with 60%+ accuracy
CSM time allocation tied to score (red customers get more attention; not less)
Renewal forecast tied to score
Quarterly score validation: did red customers actually churn? did green renew?
Score is reviewed weekly by CS team (not just at QBRs)
Trends over time visible (deteriorating vs improving)
Per-segment scoring (free / SMB / mid-market / enterprise calibrated separately)
Action playbooks per score state (red → escalation; yellow → outreach; green → expansion ask)
Sales-CS-Product alignment on what "healthy" means
Iterations every quarter based on what predicted vs didn't

This pairs with Reduce Churn (the action playbook for low-score customers), Customer Success Metrics Framework, Voice of Customer Program (NPS feeds health), Renewal Forecasting & Pipeline Management (uses health to forecast), Customer Onboarding Playbook (sets initial health), Customer Marketing Program (high-health customers feed advocacy), Quarterly Business Reviews, Activation Metric Definition, Customer Lifetime Value Playbook, and Customer Health Scoring (VibeWeek implementation).

When to Build Health Scoring

Don't build health scoring if:

You have <50 paying customers. With 30 customers, a CSM can hold every account in their head. Health scores produce noise.

You don't have an activation metric defined. Health depends on knowing what good usage looks like. Get Activation Metric Definition first.

You have no CSM team yet. Score without owner = nobody acts. Pre-CSM, founder is the implicit CSM and doesn't need a dashboard.

You can't track product usage. Health depends on usage data. If you don't have product analytics (PostHog / Amplitude / Mixpanel) yet, fix that first.

You're pre-PMF. What predicts retention isn't stable yet. Health scoring assumes a stable working motion.

Your sales motion is purely self-serve, low ACV. With $20/mo customers, intervention isn't economical. Trial-to-paid + churn-prevention via product UX is the path.

If any apply, fix the precondition.

Build health scoring when ALL are true:

1. 50+ paying customers with meaningful per-account economics ($1K+/yr ACV).

2. Defined activation metric — you know what "first value" looks like.

3. CSM team exists (or will exist within 90 days).

4. Product usage tracked in product analytics + warehouse.

5. Renewal cycle is observable — you can see who renews vs churns.

6. Sales-Marketing-CS-Product alignment exists or can be facilitated — health scoring is cross-functional.

If 5+ are true, you're ready.

The Five Pillars of Health

Different companies pick different pillars. The 5 most-load-bearing across B2B SaaS:

PILLAR 1: USAGE / ENGAGEMENT
- Are they using the product?
- Frequency: weekly active users in account; sessions per user
- Depth: features used per user / per account
- Volume: actions per week (whatever your "value action" is)
- Trend: increasing or declining over last 30/60/90 days

Why important: usage is the leading indicator. Stop using → churn 90 days later.

PILLAR 2: ADOPTION / ACTIVATION
- Have they reached the value milestones?
- Activation rate: did they cross the activation threshold?
- Feature adoption: which features have they used (and which haven't)?
- Multi-user: are multiple users active or just one?
- Multi-team: more than one team using?

Why important: lonely accounts (one user) churn at 3-5x the rate of broad accounts.

PILLAR 3: SENTIMENT
- Do they like us?
- NPS / CSAT scores
- Support ticket sentiment (any "very frustrated" tickets?)
- Engagement with comms (email open rates; webinar attendance)
- Champion engagement (replies to CSM email?)

Why important: sentiment leads usage; if they're frustrated, they'll stop using soon.

PILLAR 4: BUSINESS HEALTH (the customer's business)
- Are they doing well as a company?
- Layoffs / news signals (downsizing)
- Industry health (e.g. fintech companies in 2023-2024)
- Funding signals (raised? running out?)
- Champion still employed?
- Executive sponsor still engaged?

Why important: customer's business problems → first thing they cut.

PILLAR 5: COMMERCIAL HEALTH
- Are we set up to renew?
- Contract end date
- Payment status (any failed payments?)
- Recent expansion / contraction
- Multi-year discount in place?
- Auto-renewal enabled?
- Procurement signal (renewal evaluation initiated?)

Why important: commercial signals predict actual contract decisions.

The right number of pillars: 3-5. More than 5 over-fits + adds noise.

How to Weight (this is the hard part)

Naive weighting:
- 25% usage + 25% adoption + 25% sentiment + 25% commercial = 1-100 score

Why naive weighting fails:
- Usage trends matter MORE than absolute usage (declining usage is the predictor)
- Different segments need different weights (SMB vs enterprise)
- Sentiment alone doesn't predict churn (NPS detractors often stay)
- Some pillars are leading; some are lagging

Better weighting approaches:

A. SEGMENT-DRIVEN
- SMB: heavily weight usage + activation
- Mid-market: balance usage + sentiment
- Enterprise: weight commercial + sponsor engagement higher
- Different formulas per segment

B. TIME-WEIGHTED
- Recent activity (last 14 days) > older (last 90 days)
- Trend matters more than absolute level
- Velocity-of-change > value

C. TIER-BASED CATEGORICAL
- Don't compute a single score; compute pillar-level scores
- Then apply rules: any pillar red → overall red; all green → overall green
- Easier to explain; easier to act on

D. ML-DRIVEN (advanced; later stage)
- Train classifier on historical churn / renewal data
- Features: all pillar inputs + meta (segment, ACV, tenure)
- Output: P(churn within 90 days)
- Best at scale; requires 12+ months of data

DEFAULT FOR FIRST HEALTH SCORE:
- Tier-based categorical (option C)
- Manageable; explainable; debuggable
- Don't start with ML; you don't have the data yet

Sample tier-based rules:

Score = RED if any:
- No login in 30+ days
- NPS detractor (0-6) in last 90 days
- Champion left company
- Open priority-1 ticket > 7 days
- Failed payment unresolved

Score = YELLOW if any:
- 14-30 days since last login
- NPS passive (7-8) in last 90 days
- Usage declined > 30% over 60 days
- Activation never reached
- Open ticket without resolution > 14 days
- Champion engagement low

Score = GREEN if all:
- Active in last 14 days
- Activation reached
- NPS promoter or no recent score
- No open priority tickets
- Champion engaged in last 30 days

Scoring Pillar Examples (concrete)

Per pillar, define inputs + thresholds:

USAGE PILLAR (per account):

Inputs:
- weekly_active_users (last 4 weeks): count
- sessions_per_user (last 4 weeks): average
- key_actions_per_week: count

Thresholds (varies by product):
- Green: WAU/total_seats ratio >= 60%
- Yellow: 30-60%
- Red: <30%

Trend modifier:
- If trend is declining > 20% over 60 days: bump down one tier

ADOPTION PILLAR:

Inputs:
- has_reached_activation: bool
- features_used_count: int (out of N total)
- multi_user: bool
- integrations_connected: int

Thresholds:
- Green: activated, 5+ features, multi-user
- Yellow: activated, 2-4 features, single-user
- Red: not activated OR 0-1 features used

SENTIMENT PILLAR:

Inputs:
- last_nps_score (0-10)
- last_csat_score (1-5)
- support_ticket_priority_1_count_last_90d
- support_ticket_resolution_avg_satisfaction

Thresholds:
- Green: NPS 9-10 OR no recent score + no escalations
- Yellow: NPS 7-8 OR resolved P1 ticket recently
- Red: NPS 0-6 OR open P1 OR unresolved escalation

BUSINESS HEALTH PILLAR:

Inputs:
- champion_employed: bool
- exec_sponsor_engaged: bool
- public_layoff_signals: bool
- recent_news_negative: bool

Thresholds:
- Green: champion + sponsor engaged
- Yellow: champion engaged but sponsor cooled
- Red: champion left OR sponsor unreachable

COMMERCIAL HEALTH PILLAR:

Inputs:
- days_to_renewal: int
- payment_status: ok / overdue / failed
- contract_health: standard / stretched / contested
- recent_expansion: bool

Thresholds:
- Green: payment current; 90+ days from renewal; recent expansion
- Yellow: 60-90 days from renewal with no engagement
- Red: failed payment OR <30 days to renewal with no expansion talk

Acting on Scores (the part that matters)

A score that doesn't drive action is theater. Define playbooks per state.

RED CUSTOMERS:

Owner: senior CSM + manager involvement
Cadence: daily monitoring + weekly action
Plays:
- "What's wrong?" 1:1 outreach within 48 hours of red
- Diagnose root cause: product, sentiment, business, commercial
- Action plan with customer (joint document)
- Senior leader involvement (CSM Director or VP)
- Optional executive escalation
- 90-day save plan
- Track: did the save attempt work? Conversion: red → yellow / red → churn

If save fails:
- Graceful churn (don't burn bridges)
- Capture exit interview
- Feed learnings back to product / GTM

YELLOW CUSTOMERS:

Owner: CSM
Cadence: weekly check-in
Plays:
- Proactive outreach: "I noticed [signal]; how are things?"
- Re-engagement: training, advanced features, integration help
- Renewal prep: 60+ days before renewal, surface value, address concerns
- Move toward green within 30-60 days

GREEN CUSTOMERS:

Owner: CSM + AE (if expansion possible)
Cadence: monthly check-in
Plays:
- EXPANSION CONVERSATIONS (often missed!)
- Reference / case study ask
- Power-user training
- Cross-team expansion
- Multi-year renewal locks
- Advocacy program enrollment
- Don't ignore green: they're your growth

Anti-pattern: spending all CSM time on red, ignoring green
- Green customers are who renews + expands; ignoring them is leaving 30-50% growth on the table
- Allocate: 40% red intervention; 30% yellow improvement; 30% green expansion + advocacy

Operational Cadence

Weekly:
- CSM team reviews red list
- Manager reviews escalations
- Action items captured + tracked

Bi-weekly:
- Score audit: any score that "shouldn't be"? Investigate
- Pillar-level review: which pillar is failing most?

Monthly:
- Cross-functional review (CS + Product + Sales)
- Identify systemic issues (too many reds because of feature X)
- Feed back to product roadmap
- Adjust thresholds if too noisy

Quarterly:
- Validation: did scores predict actual outcomes?
- Cohort analysis: red 90 days ago → outcome today
- Score model adjustments
- New pillar / signal additions

Annually:
- Major model overhaul if needed
- Comparison with industry benchmarks
- Strategic review

Common Failure Modes

Failure 1: Score nobody acts on
- Pattern: built; displayed; ignored
- Fix: explicit playbook per score; CSM accountability

Failure 2: Lagging indicators only
- Pattern: scoring NPS + ticket count + payment status (all lagging)
- Fix: add USAGE trend (leading); product activity must be > 50% of weight

Failure 3: Same weights for all segments
- Pattern: SMB and enterprise score with same formula
- Reality: SMB usage matters more; enterprise sentiment + sponsor matters more
- Fix: segment-specific formulas

Failure 4: Score not validated against outcomes
- Pattern: built once; never validated
- Fix: quarterly cohort analysis; adjust based on what actually predicts churn

Failure 5: Score bouncing wildly
- Pattern: account is green Mon, red Tue, yellow Wed
- Reality: thresholds are too tight; actions impossible at this volatility
- Fix: smooth with rolling-window averages; minimum 7-day stickiness

Failure 6: Score = product analytics only
- Pattern: ignoring sentiment, business signals, commercial signals
- Fix: multi-pillar; usage alone isn't enough

Failure 7: Score becomes a CSM stick (punitive)
- Pattern: leadership: "why are 12 of your accounts red?"
- Reality: score becomes a CSM-blame metric; CSMs game it
- Fix: score is a SYSTEM measurement; CSM responsibility is action, not score itself

Failure 8: Owner unclear
- Pattern: data team builds score; CS team doesn't trust; product team doesn't update inputs
- Fix: ONE DRI (CS Ops or RevOps); cross-functional inputs

Failure 9: Score doesn't predict expansion
- Pattern: only built for churn
- Reality: green customers ARE expansion targets
- Fix: green-customer plays explicit; expansion is a health-score outcome

Failure 10: Customer-facing
- Pattern: showing customer their health score
- Reality: customer game it / get offended
- Fix: keep internal; share insights, not raw scores

Failure 11: Too granular too soon
- Pattern: shipping 1-100 score with 12 pillars in v1
- Fix: start with green/yellow/red; 3-5 pillars; iterate

Failure 12: Activation not in score
- Pattern: scoring usage but not activation
- Reality: a customer using product daily but never reached activation will churn
- Fix: activation is a non-negotiable pillar input

Failure 13: Score not trended
- Pattern: showing current score; no historical
- Reality: trend matters MORE than current level
- Fix: 30-60-90-day trend in CSM dashboard

Tooling

What you need:

- Product analytics (PostHog / Amplitude / Mixpanel) — usage data
- Data warehouse (Snowflake / BigQuery / Postgres) — central aggregation
- CRM (Salesforce / HubSpot) — where score lives + who owns
- Customer Success Platform (optional, mid-market+):
  - Vitally — modern; popular at growth stage
  - Catalyst — strong on segmentation
  - Gainsight — enterprise standard
  - ChurnZero — mid-market
- Notification tool (Slack) — alert CSMs on score changes
- Dashboard (Looker / Hex / Mode) — internal reporting

Stack patterns:

Indie / early ($1-5M ARR):
- DIY: Postgres-based scoring + Slack alerts + Salesforce / HubSpot CRM column
- No dedicated CS Platform yet

Growth-stage ($5-30M ARR):
- Vitally or Catalyst
- Health score automated
- CSM playbooks integrated

Enterprise ($30M+ ARR):
- Gainsight (enterprise default)
- ML-driven scoring
- Cross-functional dashboards

Note: customer-success-platforms are reviewed in the [Customer Success Platforms (Reference)](../../VibeReference/content/product-and-design/customer-success-platforms.md).

Don't:
- Buy Gainsight at $3M ARR (overkill)
- Build custom dashboards instead of using your CRM (loses CSM workflow)
- Run scoring in 5 different places (one source of truth)

Hire Signal: CS Operations DRI

Health scoring is a CS Ops discipline. Owner profile:

When to hire CS Ops:
- $5-10M ARR with formal CSM team
- Growing complexity (more segments, more products)

Profile:
- Background: Salesforce admin / RevOps / Customer Success Manager → Operations
- Skills: SQL, Salesforce admin, dashboards, process design
- Comp: $90-150K base depending on stage

Responsibilities (related to health scoring):
- Define + maintain health scoring model
- Operate weekly review with CSM team
- Validate scores against outcomes quarterly
- Iterate on model
- Cross-functional sync (CS / Product / Sales / Data)

What Done Looks Like (recap)

A working health scoring program:

Mistakes to Avoid

Building before activation defined. Health depends on usage; usage signal needs activation framework.
Showing customers their health score. Internal-only; share insights not numerics.
Single score; no pillars. Aggregate scores hide where the problem is.
No segment differentiation. One formula for all = noise for all.
Ignoring trend. A green customer trending down is more concerning than a yellow customer trending up.
Score without playbook. Theater; fix by defining red / yellow / green plays.
Punishing CSMs for red customers. Score is system measurement; action is CSM responsibility.
Forgetting expansion plays for green. 30-50% of growth comes from green customers; don't neglect.
No quarterly validation. Without it, the model never improves.
Lagging indicators only. Add usage trends (leading) to predict 60+ days out.
Score volatility. Smooth with rolling windows; weekly stability minimum.
Building dashboards in wrong tool. Score must live where CSMs work (CRM / CS Platform).
Overweighting NPS. NPS detractors often stay; NPS alone is weak signal.
Not feeding learnings to product. Health-score data is product feedback; close the loop.
Confusing customers' product health with customers' BUSINESS health. Both matter; track both.