How do you read cohort retention before an investor does?

A monthly churn number is an average of things that are not alike. It mixes a customer who joined last month with one who has been with you three years, and it reports a single figure that can stay flat while the business underneath it gets steadily worse.

Cohort analysis fixes that by refusing to average. It groups customers by the month they joined and follows each group forward on its own, so you can see whether what you sold in March behaves like what you sold in September.

This is the analysis Series A investors rebuild themselves, from raw transaction data, in the first week of diligence. Founders who have never built it are at an obvious disadvantage in that conversation.

What a cohort table is

Rows are joining months. Columns are months since joining. Each cell holds what that group was still worth in that month, as a share of what it was worth at the start.

So the January row shows January's customers in their first month, second month, third month and so on. The February row starts a month later and is one column shorter. The table ends up as a staircase, with the oldest cohort having the longest history.

Read down a column and you are comparing the same point in the customer life across different joining months, which answers the question that matters most: is what we are selling now better or worse than what we sold a year ago?

Read across a row and you see the shape of one group's life: how fast it falls, and whether it settles.

Build it on revenue, not just accounts

Most founders build the logo version first, which counts customers. It is useful and incomplete, because it treats a ₹4,500 customer and a ₹1,20,000 customer as the same loss.

The revenue version tracks what each cohort is worth each month. It captures three things the logo version cannot: customers leaving, customers downgrading, and customers growing. The third is what separates a good software business from an ordinary one.

Revenue retention above 100% means the cohort is worth more in month twelve than it was in month one, because expansion from the customers who stayed more than covered the ones who left. That is the single most persuasive number in an early-stage software pitch, and it cannot be seen in a churn percentage at all.

Logo retention tells you whether customers stay. Revenue retention tells you whether the relationship is worth more over time. Investors care much more about the second.

A worked example

Illustrative figures for a fictional company. Three cohorts, revenue basis, indexed to 100 in month one.

• ·January cohort: 100, 84, 79, 77, 76, 76, 78, 81 by month eight. Sharp early loss, then a flat floor, then growth from expansion.
• ·April cohort: 100, 79, 71, 66, 63, 62 by month six. A steeper fall and a lower floor.
• ·July cohort: 100, 72, 64, 59 by month four. Steeper again.

The blended monthly churn across the business barely moved during those months, because the January cohort was large and stable enough to hold the average up. The cohort table says something quite different: each new group is worse than the one before it.

That pattern has a small number of likely causes. The company may have moved down market into a segment that fits the product less well. A channel that produces cheaper sign-ups may have grown as a share of new business. Onboarding may have broken as volume rose. Or a competitor may have started taking the customers who would have stayed.

None of those is visible in a churn number. All of them are fixable if they are found in month four rather than in diligence.

The four patterns worth knowing

Almost every cohort table an early-stage company produces falls into one of four shapes.

1. 01The smile: a fall, a flat floor, then a rise as expansion outruns churn. The best outcome, and rare before Series A.
2. 02The flattening curve: a fall that settles at a stable level. Perfectly healthy. The question an investor will ask is where it settles and how confident you are that it holds.
3. 03The slide: a curve that keeps falling without ever flattening. This is the pattern that ends rounds, because it implies every customer eventually leaves and growth is permanently rented.
4. 04The decay across cohorts: each new cohort worse than the last, as in the example above. Often the most urgent, because it says something is currently going wrong rather than something went wrong once.

The floor is the number to watch. A business whose cohorts flatten at 70% has a fundamentally different economics from one that flattens at 35%, even if their first-month churn looks identical.

How to build one without a data team

Almost every company can produce a first version from a billing export and a spreadsheet, in an afternoon.

1. 01Export every invoice or subscription charge for the last eighteen to twenty-four months: customer, month, amount.
2. 02For each customer, find the first month they paid. That is their cohort.
3. 03Build a grid of cohort month against calendar month, summing revenue in each cell.
4. 04Convert each row to an index, dividing by the cohort's first month.
5. 05Do the same with a count of paying customers for the logo version.
6. 06Then look down the columns before you look across the rows.

Two practical warnings. Use the month of first payment rather than first contact or trial start, so the table is about paying customers. And keep partial months out of the newest column, because a cohort measured mid-month will look like a fall that has not happened.

Where founders get it wrong

• ·Mixing plans or segments in one table. Enterprise and self-serve customers behave nothing alike, and blending them produces a curve that describes neither. Split by segment if you have enough volume.
• ·Reading one cohort as the truth. A single month can be distorted by one large customer or one unusual campaign. Look at the shape across several.
• ·Counting annual contracts as retained in month two. They cannot churn until renewal, so the interesting column is month twelve, and the intervening flat line means nothing.
• ·Building it only on logos, which hides both downgrades and expansion.
• ·Ignoring cohorts that are too young to read. A cohort with three months of history says very little about the floor.
• ·Building it once for a pitch. The value is in watching it move quarter by quarter, which is also what makes the answer credible when an investor asks how long you have known.

Turning the floor into a number you can use

A cohort table is interesting on its own and far more useful once it produces a figure that feeds other decisions.

The step is straightforward. Take the monthly contribution a cohort produces, not its revenue, and follow the curve until it flattens. Sum the contribution across the months the cohort actually survives, and you have an estimate of what a customer from that cohort is worth.

Set that against what it cost to acquire them, fully loaded: marketing, the sales team, commissions and the founder's own selling time. The ratio between the two is what tells you whether growth makes the business better or just bigger, and the payback period tells you how long the company is out of pocket before it finds out.

Two cautions. Use contribution rather than revenue, or the number will overstate by whatever the cost to serve is. And do not project a curve that has not flattened yet: a cohort with five months of history cannot tell you what it is worth over three years, and an estimate built on an optimistic extrapolation is worse than admitting the data is young.

Where it pays off is in the segment comparison. Do this by segment and the answer usually changes which channel gets next quarter's budget, which is a much more valuable outcome than a single company-wide figure.

What a cohort table cannot tell you

It is a description of what happened, not an explanation. The table shows that the July cohort fell faster; it says nothing about why.

The why comes from the four or five conversations that follow: with the customers who left, with the salespeople who sold to them, with support about what those accounts asked for, and with whoever changed the pricing or the onboarding in the intervening months. Founders who treat the table as the whole analysis end up with a well-presented chart and no action.

It also has a blind spot around timing. A cohort that joined during a discount promotion will retain differently, and a cohort acquired just before a product change will show the effect of the change rather than of its own quality. Annotating the table with what happened in each joining month, one line per row, makes it far easier to read a year later.

Non-software businesses

The technique is not specific to subscriptions, and it is underused everywhere else.

A clinic can cohort patients by first visit and track repeat visits and spend. A restaurant group can cohort by first order through each channel and see whether delivery customers ever return. A services firm can cohort clients by the quarter they signed and track billings over the following two years.

The insight is usually the same in every case: the customers acquired through the cheapest channel come back least, and nobody had measured it because the average looked acceptable.

In these businesses the definition of retention needs a decision first. A patient who returns after fourteen months has not churned in the way a lapsed subscriber has. Pick a window that reflects how the business actually works, write it down, and use it consistently.

What an investor will do with it

Expect three specific moves in diligence.

They will rebuild the table from your raw data rather than accept yours, which is one reason the definitions you use should be written down and defensible. They will compare cohorts before and after any change in pricing, channel or segment, which is where a decay pattern shows up. And they will use the floor to build a lifetime value estimate, which becomes the denominator in every efficiency ratio they calculate about you.

The founder who has been watching the same table for four quarters can explain what happened in each of those moves, including the bad quarters. That is a much stronger position than having the analysis presented to you.

About Simplify

Simplify is a finance clarity and investment readiness practice working with founders across India, built on six years inside startups. We write about the questions founders bring before a decision, not after it.