पाठशाला Pathshala · वृद्धि Vṛddhi, Growth · Lesson 15 · Build

Retention is the growth engine nobody budgets for

Every growth plan has a line for acquisition and almost none has one for keeping customers. Measure retention by cohort, find the three places customers leak and fix them before buying more.

Pathshala, The Founder Library · 11 October 2026 · 7 min read

Rows of tea bushes cover rolling hills under a misty morning sky. — Photograph: Duc Nguyen · Pexels

A Bengaluru subscription company adds a thousand paying customers a month and loses six per cent of its base every month. In its first year it grows fast and the board is pleased. In its fourth year it is still adding a thousand a month, spending more to do it, and the base grows by about six per cent across the whole year. Nothing broke. The company hit the ceiling its churn had set on the first day.

Growth plans in India are written as acquisition plans: a budget for Meta and Google, a target for new users, a line for the sales team. Keeping customers appears nowhere, or appears as a support cost. This lesson is the case for treating retention as the growth engine it is, with the arithmetic that proves it, the way to measure it by cohort, benchmarks for good and great, and the three leaks that account for most lost customers.

Churn sets the ceiling

The arithmetic is short. If a business adds a fixed number of customers each month and loses a fixed share of its base, the base grows until the number leaving each month equals the number arriving, and then it stops. That level is new customers divided by monthly churn. At a thousand new customers and six per cent churn the ceiling is about 16,700 customers, and the company reaches most of it in under three years. Ron Gill, then CFO of NetSuite, quoted in David Skok’s SaaS Metrics 2.0, puts it in one sentence: the churn rate combined with the new revenue added defines the maximum size the business can reach.

Skok’s own example shows the size of the effect. A company booking a constant $6,000 of new monthly revenue with 3 per cent monthly churn reaches about $140,000 of monthly revenue after forty months and is flattening out. The same company with 3 per cent negative churn, where expansion from existing customers more than covers the losses, reaches about $450,000, more than three times as much, from the same acquisition. Skok calls negative churn the ultimate solution to the churn problem and names the two routes to it: pricing that rises with usage, and upselling customers to larger plans or added modules.

Measure it by cohort

An average churn figure hides everything that matters. Sequoia’s data science team, in its essay on retention, sets out the method: group customers by the month they started, plot the share still active in each month after, and read the shape. A curve that flattens has found a group of customers who stay, and the height of the plateau is the health of the business. A curve that keeps declining is weak product-market fit, and spending on acquisition before fixing it makes growth a leaky bucket. A curve that smiles, rising again as lapsed users return, is rare and valuable.

Two decisions make the measurement honest. Define active as the action that reflects real value, not an app open: Sequoia’s example is a completed ride, not a launch. For a subscription it is a renewal paid; for a marketplace a repeat order; for a B2B tool the core workflow run in the week. Then read the cohort table, which Sequoia calls a triangle chart, in three directions. A pattern along a row points to that cohort, such as a festival campaign that brought the wrong customers. A diagonal points to an event that hit every cohort at once, such as an outage or a price rise. A vertical points to the calendar, such as annual renewals. The [cohort analysis](/library/cohort-analysis-for-founders-not-analysts) lesson builds the table step by step.

What good looks like

Lenny Rachitsky’s survey of operators and investors, What is good retention, gives benchmarks for user retention at six months. Consumer social: about 25 per cent is good and 45 per cent great. Consumer transactional, which covers most commerce and delivery: 30 and 50. Consumer subscription software: 40 and 70. SMB and mid-market software: 60 and 80. Enterprise software: about 70 and 90. For net revenue retention over twelve months, which counts expansion as well as loss, bottom-up software companies are good at about 100 per cent and great at 120, and enterprise companies good at 110 and great at 130.

Use the benchmark for your category and treat it as a direction, not a pass mark. A food subscription in Hyderabad and a payroll product for Pune factories are not in the same table. What matters more than the level is the shape: whether your curve flattens, at what month, and whether each new cohort flattens higher than the one before. The figure below shows why the level matters so much for growth.

Acquisition decides how fast the bucket fills. Retention decides how full it can ever get. Most companies spend on the first and inherit the second.

The three leaks

The first week. Most customers who leave, leave early. Andrew Chen’s analysis of Quettra’s Android data, covering more than 125 million devices in 2015, found the average app lost about 77 per cent of its daily users within three days of install and more than 90 per cent within thirty. The top ten apps kept about 75 per cent on day one against 29 per cent for the average, and about 60 per cent on day thirty against under 10. Their curves fell at roughly the same speed after day one; they won on the first day. Sequoia’s method for finding this leak is to break retention into ratios, day one over day zero, day seven over day one, and fix the worst step first. Its example is Poncho, whose seven-day retention rose from 60 to 80 per cent after it removed questions from sign-up. The [activation lesson](/library/activation-first-session-that-decides) is the work on this leak.

An old wooden bucket rests on the rim of a countryside well. — A bucket that leaks is filled in vain however deep the well. Find the leak before drawing more water. Photograph: Marek Ruczaj · Pexels

The habit that never forms. Some customers activate and still drift away because the product never became part of their week. The diagnosis is feature-level retention: which actions do the customers who stay perform that the leavers do not? Sequoia points to finding the super users and the moment that hooked them, and cites Facebook’s finding that users who connected with seven friends in ten days were far more likely to stay. Build the product and the onboarding to push every new customer towards that action.

The payment that fails. In subscription businesses a share of churn is customers who never decided to leave: an e-mandate that was not renewed, a card that expired, a bank debit that failed, an annual renewal invoice that sat in a finance inbox. In India, where recurring payments run through mandates that customers must authorise and renew, this leak is often larger than founders think, and it is the cheapest to fix: retry failed debits, send a renewal reminder thirty and seven days ahead on WhatsApp and email, offer a UPI link when a mandate fails, and give a human the list of every failed renewal each Monday.

Fix the bucket before filling it

Sequoia’s advice is blunt: if the curve is declining, fix the product before spending on acquisition. The economic case is old. Fred Reichheld of Bain & Company, in Prescription for cutting costs, reported that in financial services a 5 per cent increase in customer retention produced more than a 25 per cent increase in profit, because retained customers cost less to serve, buy more and refer others. Every rupee of acquisition spent into a declining curve buys customers who will be gone before they pay back.

So budget for it. Give retention a named owner with a number, the six-month retention of the latest cohort or net revenue retention, reported beside new-customer numbers in every [weekly metrics review](/library/weekly-metrics-review-one-page-one-hour). Give it a line in the plan: onboarding, customer success calls for the first ninety days, payment recovery, and the product work on the first-week and habit leaks. Then, when planning next quarter’s growth, ask the question the figure asks: what would it cost to bring a quarter more customers in, and what would it cost to keep a quarter more? The second answer is usually smaller.

The monthly retention review

First working week of each month, one hour, the founders, product and whoever owns customers. Add last month’s cohort to the table and update every older row. Read the newest three cohorts against the three before them: is month-one retention rising, and are curves flattening earlier and higher? Compute the ceiling, new customers divided by monthly churn, and write it next to the current base; if the base is within a fifth of the ceiling, growth will stall without a retention change. Count the month’s churned customers by leak: left in the first week, drifted after activating, payment failed. Pick the largest leak, name one change to make this month and the number it should move. Finally, read ten cancellation reasons in the customer’s own words before closing the meeting.

The benchmarks are the sources’ and vary by category; your own cohort table is the only number that counts. Start it this month.