HTML MASTER CLASS /// LEARN TAGS /// BUILD STRUCTURE /// SEMANTIC WEB /// HTML MASTER CLASS /// LEARN TAGS ///

Reading the Leaky Bucket

Measuring who's leaving, why, and how to tell a real trend from a noisy month.

⚡ Total XP: 0|💻 management XP: 0

Skill Matrix

UNLOCK NODES BY LEARNING NEW TAGS.

Select an unlocked node to view details root

🚀 LEVEL UP TO SENIOR:Unlock 500+ Advanced Practical Challenges & Exercises.

🎓 COURSERA PARTNER:Earn professional Google, Meta, and IBM certificates to supercharge your resume.

Growth in new signups can hide a shrinking business if retention is bad enough. Here's how to measure churn in a way that actually points at the fix.

1Why Blended Retention Numbers Lie

If you report 'our retention is 80%' as one number for the whole customer base, you're averaging brand-new users who haven't had a chance to churn yet with two-year veterans who've already proven they stick. Cohort tables — rows are signup month, columns are months since signup — expose the real curve, and let you compare whether a product change actually improved retention for cohorts that signed up after it shipped versus before.

2Net Revenue Retention Changes the Conversation

Logo churn (percentage of accounts lost) treats every customer as equal, which is fine for consumer apps but misleading for B2B SaaS with wildly different account sizes. NRR folds in expansion revenue from upsells and seat growth, contraction from downgrades, and churn from cancellations into one number relative to starting revenue. An NRR of 115% means the existing customer base alone is growing the business, independent of new sales — the single most-watched numbers on SaaS earnings calls for exactly this reason.

3Step-by-Step Breakdown

Introduction. Churn rate is the percentage of customers (or revenue) you lose over a period. Retention rate is what's left: 1 minus churn. A product can grow bookings every month and still be dying underneath if churn is quietly eating the base — retention is the metric that catches that.

Cohort Retention Curves. Blended, company-wide retention hides the truth because new cohorts dilute old ones. Cohort analysis groups users by signup month and tracks what percentage of each cohort is still active in month 1, 2, 3... A healthy product's retention curve flattens into a plateau; an unhealthy one keeps sliding toward zero.

Logo Churn vs. Revenue Churn vs. NRR. Logo churn counts lost accounts equally, but losing a $200/month customer and a $20,000/month customer are not the same event. Net Revenue Retention (NRR) tracks revenue from the existing base including upgrades and downgrades — NRR above 100% means expansion is outpacing churn even with zero new sales.

Knowledge Check. Why is cohort-based retention analysis more trustworthy than a single blended retention number for the whole user base?

→Because a blended number mixes old and new users, letting a surge of new signups mask declining retention in older cohorts
→Because cohort analysis is required by GAAP accounting standards for public companies

Summary. Retention work isn't just measurement — it's diagnosis. Once you know which cohort, which plan tier, or which onboarding path churns fastest, you have a specific problem to fix instead of a vague instruction to 'improve retention.'

Pascual Vila

Frontend Instructor // Code Syllabus

Common Pitfalls & Errors

The Error //

Reporting a single blended retention number instead of cohort curves

// Wrong
"Overall retention this quarter: 78%." (mixes week-old and year-old users)

// Correct
"Q1 cohort: 55% retained at Month 3. Q2 cohort: 61% retained at Month 3 — improving."

The Solution //

A blended retention figure improves the moment you have a burst of new signups, even if long-term retention is getting worse — the new users haven't had time to churn yet. Always break retention down by signup cohort before declaring a trend healthy or unhealthy.

The Error //

Treating all churned accounts as equal (logo churn only)

// Wrong
"Churn is stable at 5% of accounts, nothing to worry about."

// Correct
"Logo churn is 5%, but revenue churn is 14% because two enterprise accounts left — NRR dropped to 91%."

The Solution //

Losing one $50/month starter-plan account and one $15,000/month enterprise account both count as '1 churned logo,' but they are not equally urgent. Track revenue churn and NRR alongside logo churn so a handful of large losses don't hide inside a flat-looking logo number.