Measuring Retention And Expansion

Why this matters

For an AI Product Manager, growth compounds when customers stick around and spend more. Measuring retention and expansion tells you if the product delivers recurring value, which segments are thriving, and where to invest in activation, adoption, and pricing. Real tasks you will do:

• Define activation and retention events for an AI assistant or API product.
• Build cohort tables to monitor logo and revenue retention by signup month.
• Calculate Gross Revenue Retention (GRR) and Net Dollar Retention (NDR).
• Identify expansion levers: seat growth, usage consumption (tokens/requests/credits), upsells.
• Set targets and alerts for churn risk and contraction.

Concept explained simply

Retention answers: do customers come back and keep paying? Expansion answers: do existing customers increase spend over time?

• Logo retention: percent of customers who remain customers.
• Revenue retention: percent of revenue that remains (and can increase with expansions).
• GRR (Gross Revenue Retention): existing revenue kept, after churn and downgrades, before expansions.
• NDR (Net Dollar Retention): existing revenue kept plus expansions, minus downgrades and churn.

Common formulas (monthly or annual, but be consistent):

• Logo retention = retained customers / customers at start (excluding new)
• GRR = revenue from existing customers at end (no expansions) / revenue from same customers at start
• NDR = (start revenue - churn - contraction + expansion) / start revenue

Mental model

Think of a leaky, self-filling bucket. Retention reduces leaks. Expansion turns the faucet on inside the bucket. Your job: fix the holes (onboarding, product value, support) and turn up the internal faucet (upsell paths, usage growth, pricing/packaging).

AI product specifics

• Usage units: tokens, requests, compute minutes, seats accessing AI features.
• Activation: first successful inference that meets latency and quality criteria.
• Leading indicators: weekly active users, feature adoption, prompts per active user, API success rate, time-to-value.
• Lagging indicators: logo retention, GRR, NDR.

Worked examples

Example 1: GRR and NDR

Start-of-month revenue from existing customers: 100,000.

• Churned revenue: 8,000
• Contraction (downgrades): 5,000
• Expansion (upsell/usage): 18,000

GRR = (100,000 - 8,000 - 5,000) / 100,000 = 87%.

NDR = (100,000 - 8,000 - 5,000 + 18,000) / 100,000 = 105%.

Interpretation: You lost revenue but expansions more than offset losses; the base is growing.

Example 2: Logo retention

Start-of-quarter customers: 500 (exclude any new customers acquired this quarter). End-of-quarter customers from that same set: 455.

Logo retention = 455 / 500 = 91%. Logo churn = 9%.

Example 3: Cohort retention (AI API)

January cohort: 300 new accounts activated (first successful API call). Active by month after signup:

• Month 1: 210 active (70%)
• Month 2: 180 active (60%)
• Month 3: 150 active (50%)

Interpretation: 70/60/50% month 1/2/3 retention. Investigate drop between M1 and M3: quota limits, latency, or unclear value?

How to measure (step-by-step)

1) Define events and units

• Activation: first successful AI output meeting latency/quality bar.
• Active: at least N successful outputs per week (choose N relevant to value).
• Revenue unit: seats, tokens, requests, or plan price.

2) Build cohorts

• Group users by signup or activation month.
• Track percent active and revenue per cohort over time.

3) Calculate GRR and NDR

• Exclude new customers from the period when computing GRR/NDR.
• Break down losses (churn, contraction) and gains (expansion).

4) Segment

• By plan, industry, company size, use case, geography.
• By model type or feature adopted (RAG, summarization, copilot).

5) Instrument leading indicators

• Time-to-first-value, weekly active users, prompts per user, success rate.
• Set alerts when leading indicators dip for healthy cohorts.

Who this is for

• AI Product Managers working on SaaS, AI assistants, or AI APIs.
• Founders and growth PMs responsible for monetization and retention.
• Analysts supporting AI product growth.

Prerequisites

• Basic spreadsheet and SQL comfort.
• Understanding of product events (signup, activation, usage).
• Familiarity with pricing models (seat-based, usage-based).

Learning path

1. Define activation and active usage for your AI product.
2. Build monthly cohorts for logo and revenue retention.
3. Compute GRR and NDR; set targets (e.g., GRR 90%+, NDR 105–130%). Varies by country/company; treat as rough ranges.
4. Segment cohorts to find drivers of churn and expansion.
5. Design experiments to improve activation and upsell paths.

Common mistakes and how to self-check

• Mixing cohorts: comparing users who started in different months without cohorting. Self-check: every rate references one cohort at a time.
• Including new customers in GRR/NDR. Self-check: verify numerator/denominator exclude new logos.
• Counting reactivations as retained without marking a gap. Self-check: track reactivation separately.
• Currency or billing period mismatches. Self-check: normalize to one currency and one consistent period.
• Using vanity active definitions (e.g., any login). Self-check: active must reflect value (e.g., successful AI outputs).

Practical projects

• Spreadsheet cohort table: rows = cohorts by month; columns = retention by month N. Add GRR and NDR per cohort.
• Define and implement activation event: first successful AI output under X seconds and quality above threshold.
• Expansion playbook: identify 3 upsell triggers (e.g., hit 80% of quota, team collaboration features, compliance add-on) and map in-product nudges.

Exercises

These mirror the exercises below. Try them here, then check the solutions.

1. Exercise 1 (Metrics): You start the month with 1,200 customers generating 240,000 in revenue. During the month, 30,000 churns, 12,000 contracts (downgrades), and 48,000 expands (upsells/usage). Compute GRR and NDR.

2. Exercise 2 (Activation & leading indicators): For an AI writing assistant with free and Pro plans, define a measurable activation event and two leading indicators that are predictive of 4-week retention. Propose one A/B test to improve the leading indicators.

• I excluded new customers when computing GRR/NDR.
• I used a clear, value-based activation definition.
• I identified at least one actionable leading indicator.

Mini challenge

Your NDR is 98% despite 93% GRR. Segmentation shows startups have 120% NDR, enterprises 88% NDR with high contraction. In one paragraph, choose where to focus next month and what leading indicators you will track to confirm you are improving.

Next steps

• Instrument event tracking for activation and active usage.
• Automate monthly cohort and NDR calculations.
• Set team-level targets for GRR and NDR; review weekly leading indicators.

Quick test info

There is a short test for this subskill. Anyone can take it for free; only logged-in users get saved progress.

Scroll to the Quick Test section below.

Tips for clean measurement

Retention windows

Pick monthly or quarterly windows and stick to them. For usage-heavy AI products, monthly is common.

Revenue recognition vs. invoicing

For long contracts, use recognized revenue for GRR/NDR where possible to avoid one-time spikes.

Define contraction clearly

Contraction includes seat reductions, plan downgrades, or sustained lower usage in usage-based models.

Glossary

• Logo retention: percent of customers retained.
• GRR: percent of revenue retained excluding expansions.
• NDR: percent of revenue retained including expansions.
• Cohort: group of customers that started in the same period.

Quick Test

Take the test to check your understanding. Anyone can take it for free; only logged-in users get saved progress.