Customer Analytics (LENS Framework)
Design a customer analytics architecture that answers which customers, doing what, are driving (or breaking) the business - instead of dashboards full of vanity counts. LENS produces a defensible segmentation, a retention model, an engagement score, and a behavioral diagnostic loop that PMs and CS can act on weekly.
Core Principle
Customer analytics fails when it stops at "users went up." LENS forces decomposition into who, what, when, and why - the four axes a dashboard usually collapses into one number.
The LENS Framework
| Letter |
Stage |
The Question |
| L |
Lifecycle Mapping |
What are the named lifecycle stages and what does each one's "good" look like? |
| E |
Engagement Scoring |
What weighted score combines depth, breadth, and recency of value events? |
| N |
Net Retention Decomposition |
Where exactly is NRR coming from - new logo, expansion, contraction, churn? |
| S |
Segment Behavior |
Which segments behave differently, and which behavioral cohorts predict outcomes? |
Lifecycle Stages
| Stage |
"Good" Signal |
Diagnostic |
| New |
First value event within target window |
Activation rate by cohort |
| Activated |
≥ N value events / week within 30 days |
Stickiness (DAU/WAU or analog) |
| Habituated |
Multi-workflow + multi-user adoption |
Workflow coverage % |
| Expanding |
New seats / modules / use cases attached |
Expansion lead indicators |
| At-risk |
Engagement decay + stakeholder loss |
Churn-risk score |
| Churned / Contracted |
Logo or ARR loss |
Reason-coded post-mortems |
Engagement Scoring
Engagement is depth × breadth × recency, not raw event counts.
| Dimension |
Definition |
Example |
| Depth |
Frequency of core value events per active user |
Core actions / week |
| Breadth |
% of paid seats active + # of distinct workflows used |
Seat activation, workflow coverage |
| Recency |
Time since last value event, weighted exponentially |
Decay half-life of 14-30 days |
Combine into a 0-100 score; bucket into Engaged / Mixed / Disengaged for routing into CS plays.
Net Retention Decomposition
A single NRR number hides the truth. Always decompose:
| Component |
Formula |
What It Tells You |
| GRR |
(Starting ARR − Churn − Contraction) / Starting ARR |
Floor on the business |
| Expansion % |
Expansion ARR / Starting ARR |
Upside from existing book |
| NRR |
GRR + Expansion % |
Compound growth signal |
| Churn drivers |
Reason-coded, % of churned ARR by reason |
Where to fix the leak |
| Contraction drivers |
Seat reductions vs price reductions vs downgrades |
Where pricing/packaging is misaligned |
Segment Behavior
Segments must be decision-driving, not decorative. Two segmentation lenses:
| Lens |
Example |
Use For |
| Firmographic |
Industry × Size × Region |
GTM motion design |
| Behavioral |
Activation pattern, workflow mix, usage intensity |
Lifecycle interventions, expansion targeting |
The behavioral lens almost always predicts retention better than the firmographic one - most teams underuse it.
Output
Save to outputs/customer-analytics-[scope]-[YYYY-MM-DD].md
| Artifact |
Description |
| Lifecycle Model |
Named stages with entry/exit criteria + "good" definitions |
| Engagement Score Spec |
Dimensions, weights, decay, bucket thresholds |
| NRR Decomposition |
Waterfall: starting → expansion → contraction → churn → ending |
| Segment Behavior Matrix |
Behavior cohorts × outcome (retention, expansion, time-to-value) |
| Diagnostic Loop |
Weekly review template: anomaly → hypothesis → action → owner |
| Cohort Retention Curves |
M0-M12 retention by acquisition cohort and segment |
Process
- Map lifecycle stages with entry/exit criteria - agree with PM and CS before instrumenting
- Define engagement score with explicit weights; validate against historical churn
- Decompose NRR into a waterfall - every component reason-coded
- Build behavioral cohorts that predict outcomes better than firmographics
- Stand up the diagnostic loop - weekly anomaly review, owner assigned, action tracked
- Wire cohort retention curves into the executive cadence, not buried in a tool
Tips
- Avoid vanity engagement metrics - logins, page views, and DAU rarely predict retention
- Validate the score against churn before deploying - score that doesn't correlate is decoration
- Segment behavior beats firmographic targeting for retention plays
- Treat contraction separately from churn - different root causes, different fixes
- Anomaly without action is noise - every diagnostic must end with an owner
Pairs With
- journey-architect - Lifecycle stages map directly to journey gates
- customer-success - Health score consumes the engagement score
- growth-loop - Retention curves feed loop-strength analysis
- revenue-analytics - NRR decomposition rolls into revenue diagnostics
1---2name: customer-analytics3description: Customer analytics framework - cohort retention, lifecycle funnels, engagement scoring, segmentation, and behavioral diagnostics for product and CS teams. Use when: customer analytics, cohort analysis, retention curve, engagement score, customer segmentation, behavioral analysis, lifecycle funnel, RFM, activation diagnostics, usage analysis.4---56# Customer Analytics (LENS Framework)78Design a customer analytics architecture that answers *which customers, doing what, are driving (or breaking) the business* - instead of dashboards full of vanity counts. LENS produces a defensible segmentation, a retention model, an engagement score, and a behavioral diagnostic loop that PMs and CS can act on weekly.910## Core Principle1112**Customer analytics fails when it stops at "users went up." LENS forces decomposition into *who*, *what*, *when*, and *why* - the four axes a dashboard usually collapses into one number.**1314## The LENS Framework1516| Letter | Stage | The Question |17|--------|-------|--------------|18| **L** | Lifecycle Mapping | What are the named lifecycle stages and what does each one's "good" look like? |19| **E** | Engagement Scoring | What weighted score combines depth, breadth, and recency of value events? |20| **N** | Net Retention Decomposition | Where exactly is NRR coming from - new logo, expansion, contraction, churn? |21| **S** | Segment Behavior | Which segments behave differently, and which behavioral cohorts predict outcomes? |2223## Lifecycle Stages2425| Stage | "Good" Signal | Diagnostic |26|-------|---------------|------------|27| **New** | First value event within target window | Activation rate by cohort |28| **Activated** | ≥ N value events / week within 30 days | Stickiness (DAU/WAU or analog) |29| **Habituated** | Multi-workflow + multi-user adoption | Workflow coverage % |30| **Expanding** | New seats / modules / use cases attached | Expansion lead indicators |31| **At-risk** | Engagement decay + stakeholder loss | Churn-risk score |32| **Churned / Contracted** | Logo or ARR loss | Reason-coded post-mortems |3334## Engagement Scoring3536Engagement is **depth × breadth × recency**, not raw event counts.3738| Dimension | Definition | Example |39|-----------|------------|---------|40| **Depth** | Frequency of core value events per active user | Core actions / week |41| **Breadth** | % of paid seats active + # of distinct workflows used | Seat activation, workflow coverage |42| **Recency** | Time since last value event, weighted exponentially | Decay half-life of 14-30 days |4344Combine into a 0-100 score; bucket into Engaged / Mixed / Disengaged for routing into CS plays.4546## Net Retention Decomposition4748A single NRR number hides the truth. Always decompose:4950| Component | Formula | What It Tells You |51|-----------|---------|-------------------|52| **GRR** | (Starting ARR − Churn − Contraction) / Starting ARR | Floor on the business |53| **Expansion %** | Expansion ARR / Starting ARR | Upside from existing book |54| **NRR** | GRR + Expansion % | Compound growth signal |55| **Churn drivers** | Reason-coded, % of churned ARR by reason | Where to fix the leak |56| **Contraction drivers** | Seat reductions vs price reductions vs downgrades | Where pricing/packaging is misaligned |5758## Segment Behavior5960Segments must be **decision-driving**, not decorative. Two segmentation lenses:6162| Lens | Example | Use For |63|------|---------|---------|64| **Firmographic** | Industry × Size × Region | GTM motion design |65| **Behavioral** | Activation pattern, workflow mix, usage intensity | Lifecycle interventions, expansion targeting |6667The behavioral lens almost always predicts retention better than the firmographic one - most teams underuse it.6869## Output7071Save to `outputs/customer-analytics-[scope]-[YYYY-MM-DD].md`7273| Artifact | Description |74|----------|-------------|75| **Lifecycle Model** | Named stages with entry/exit criteria + "good" definitions |76| **Engagement Score Spec** | Dimensions, weights, decay, bucket thresholds |77| **NRR Decomposition** | Waterfall: starting → expansion → contraction → churn → ending |78| **Segment Behavior Matrix** | Behavior cohorts × outcome (retention, expansion, time-to-value) |79| **Diagnostic Loop** | Weekly review template: anomaly → hypothesis → action → owner |80| **Cohort Retention Curves** | M0-M12 retention by acquisition cohort and segment |8182## Process83841. **Map lifecycle stages** with entry/exit criteria - agree with PM and CS before instrumenting852. **Define engagement score** with explicit weights; validate against historical churn863. **Decompose NRR** into a waterfall - every component reason-coded874. **Build behavioral cohorts** that predict outcomes better than firmographics885. **Stand up the diagnostic loop** - weekly anomaly review, owner assigned, action tracked896. **Wire cohort retention curves** into the executive cadence, not buried in a tool9091## Tips92931. **Avoid vanity engagement metrics** - logins, page views, and DAU rarely predict retention942. **Validate the score against churn before deploying** - score that doesn't correlate is decoration953. **Segment behavior beats firmographic targeting** for retention plays964. **Treat contraction separately from churn** - different root causes, different fixes975. **Anomaly without action is noise** - every diagnostic must end with an owner9899## Pairs With100101- **journey-architect** - Lifecycle stages map directly to journey gates102- **customer-success** - Health score consumes the engagement score103- **growth-loop** - Retention curves feed loop-strength analysis104- **revenue-analytics** - NRR decomposition rolls into revenue diagnostics
varunk130