Skip to main content

​
Cohort Intelligence

The Cohort Intelligence engine groups accounts into cohorts by vintage (sign-up quarter), segment, territory, or rep, then computes per-cohort revenue metrics (NRR, GRR, churn, expansion), generates vintage retention curves, and detects anomalies like accelerating churn or health score decline.
Cohort Intelligence consumes data from the Accounts and Renewals tables. It does not modify entity-level scores — it produces aggregate cohort metrics for trend analysis and board reporting.

​
How It Works

Accounts + Renewals data
  → Cohort definitions (vintage, segment, territory, rep)
  → Cohort snapshots (NRR, GRR, churn, expansion per period)
  → Vintage curves (month-by-month retention from cohort start)
  → Cohort comparison (ranked by NRR with best/worst flags)
  → Anomaly detection (low NRR, accelerating churn, health decline)

​
Cohort Types

TypeGrouping FieldExample
Vintagecreated_at quarterVintage 2025-Q3
SegmentsegmentSegment: Enterprise
TerritoryterritoryTerritory: Northeast
Repowner_idRep: sarah-jones
CustomUser-defined filterCustom criteria

​
Cohort Snapshots

Each cohort receives a periodic snapshot with the following metrics:
MetricDescription
Total ARRSum of ARR across all cohort members
Starting ARRARR at the beginning of the snapshot period
NRR(Starting + Expansion - Contraction - Churn) / Starting x 100
GRR(Starting - Contraction - Churn) / Starting x 100 (capped at 100%)
Churn CountNumber of accounts with CHURNED disposition in the period
Churn ARRARR lost to churn
Expansion ARRARR gained from expansion
Contraction ARRARR lost to downsell
Avg Health ScoreAverage health score across cohort members
Avg Risk ScoreAverage risk score across cohort members
Avg Products per AccountAverage product count
Avg Adoption %Average product adoption percentage

​
Vintage Curves

Vintage curves track month-by-month retention from the cohort’s start date for up to 36 months. Each data point includes:
FieldDescription
Months Since Start0, 1, 2, … up to 36
Retention %(Starting ARR - Cumulative Churn + Cumulative Expansion) / Starting ARR x 100
Accounts RemainingStarting account count minus cumulative churned accounts
Cumulative Churn ARRRunning total of ARR lost to churn
Cumulative Expansion ARRRunning total of ARR gained from expansion
Vintage curves can exceed 100% retention when expansion ARR outpaces churn — this is a sign of strong net revenue retention.

​
Anomaly Detection

The engine flags three anomaly types:
AnomalyDescription
Low NRRNet revenue retention below healthy thresholds, with escalating severity
Accelerating ChurnChurn count increasing period-over-period, with severity based on acceleration rate
Health DeclineAverage health score declining significantly from prior period
Exact anomaly thresholds and severity classifications are configurable per organization and available in the PILLAR Implementation Guide provided to active customers.

​
Cohort Comparison

Cohorts are ranked by NRR with explicit best/worst indicators:
IndicatorMeaning
bestHighest NRR cohort
worstLowest NRR cohort
normalAll others

​
Data Model

PILLAR stores cohort definitions (grouping criteria and member counts), periodic cohort snapshots (revenue metrics, health scores, churn/expansion data per period), and vintage retention curves (month-by-month retention tracking for up to 36 months).
Detailed data model schemas are available in the PILLAR Implementation Guide provided to active customers.

​
API Endpoints

GET  /api/cohorts/definitions
POST /api/cohorts/definitions
See the Cohorts API reference for full endpoint documentation.

​
Access

Available to: CRO/CEO, VP Sales, VP CS, RevOps