Which customers actually make money?

A company with a blended gross margin of 71% sounds healthy. Split it and you often find one plan at 53% that churns quickly, one at 76% that stays for years, and an enterprise tier whose implementation eats most of its first year. The averages tell you nothing about which of those to sell more of.

That's what this template is for. It puts up to six segments beside each other, applies the same arithmetic to all of them, and shows the two numbers that decide where money should go: contribution per customer and how many months it takes to earn back what winning them cost.

Segments can be plans, product lines, customer sizes, locations, channels or payer types. Whatever cut actually differs in your business is the right one.

The exercise is worth doing the moment a business has more than one of anything: more than one plan, more than one location, more than one type of customer. Before that, the P&L tells you what you need. After it, averages start covering for whichever part of the business is being carried by the rest.

The last row on the Segments sheet compares payback with how long a customer stays. If payback is longer than the average life, the segment loses money on every customer won, and more marketing makes the loss bigger.

In the example that ships with the file, the Starter segment contributes ₹2,400 a month against a fully loaded CAC of ₹38,000, so payback is 15.8 months. At 4.5% monthly churn the average customer stays about 22 months. It passes, but only just, and only if churn doesn't get worse.

The Growth segment contributes ₹13,700 with a CAC of ₹1.1 lakh: payback in 8 months against a life of nearly five years. Every rupee moved from Starter acquisition to Growth acquisition is worth several times more, even though Starter produces more sign-ups and therefore looks better in a weekly marketing report.

1. 01Price: what the segment actually pays after discounts, from billing rather than the price list.
2. 02Infrastructure and usage: hosting, third-party APIs, model usage, data costs, attributable to that segment. Heavy users often sit in one segment.
3. 03Payment and platform fees: gateway charges, aggregator commissions, marketplace cuts, and the GST on them where you cannot claim it back.
4. 04Support and account management: the fairest split is usually by time. Ask the team to estimate what share of their week goes to each segment for a fortnight.
5. 05Other variable costs: delivery, implementation, consumables, referral fees.
6. 06Fully loaded CAC: all sales and marketing costs including salaries, commissions, tools and founder selling time, divided by customers won in that segment.
7. 07Churn: customers lost divided by customers at the start of the month, averaged over several months.

The support line is where most people give up and enter zero. Don't. A rough estimate from the team is far more accurate than pretending the cost is nil, and support is usually the reason a cheap plan is unprofitable.

Software

A customer, segmented by plan or by size. Variable cost is infrastructure, model or API usage, payment fees and support. Watch for heavy users on flat pricing, which is where AI-powered products lose money quietly.

Services firms and agencies

A project or a client. Variable cost is the delivery team's time at fully loaded cost, so realisation and utilisation are the economics. Acquisition cost includes proposal and pitch time, including the ones you lost.

Restaurants and cloud kitchens

An order, split by channel. Variable cost is food, packaging, aggregator commission and the GST on it. The same dish through three channels is three different unit economics.

Healthcare services

An episode: a consultation, test or procedure, split by payer. Variable cost includes consumables, clinician share where it's a revenue share, and GST on inputs that exempt services cannot recover. Repeat rate matters more than almost anything else.

Marketplaces

A transaction, with your take rate as revenue rather than gross booking value. Variable costs are payment fees, support and any subsidy you fund on either side.

A blended acquisition cost is an average of channels that behave nothing like each other. The Channel CAC sheet adds allocated sales cost to media spend for each channel and divides by customers won, which usually produces a spread wide enough to change how the budget is set.

Two cautions. Attribution is imperfect: a customer who read your content, attended an event and then clicked a paid ad gets credited to one of the three. And some channels take longer to convert, so this month's spend produced customers who will appear next quarter. Use a quarterly view when the sales cycle is long.

Even with those caveats, the ranking is usually stable and usually surprising. Referrals and organic almost always win on cost. The question they raise is whether they can be made to produce more volume, which is a different kind of work from buying more clicks.

The Cohorts sheet asks one question: is the product getting better at keeping customers? Each row is a joining month, and each column is how many of that group were still active after one month, two months and so on.

Read across a row and you see decay, which every business has. Read down a column and you see whether each new cohort is holding up better than the one before it. That second reading is the one that matters, because it shows whether the changes you have made to onboarding, pricing or the product are working.

It is also the analysis investors ask for most often at Series A and beyond, and the one that takes longest to produce if nobody has been keeping it. Building the grid now, with whatever history exists, costs an hour and saves a week later.

One practical note: count customers, not revenue, in the grid first. Revenue retention can look healthy while customer retention is poor, because a few accounts that expand mask many that leave. Both matter, and the customer view is the one that tells you whether the product is working for the people who bought it.

A full rebuild once a quarter is usually right. Prices change, support costs drift, channel mix moves, and a segment that paid back in eight months last year may not now.

Between rebuilds, carry two lines in the monthly pack: contribution margin by segment, and payback by channel. Both take minutes once the template exists, and both move slowly enough that a change in either is worth a conversation.

Rebuild immediately, rather than waiting for the quarter, when you change prices, launch a plan or product, sign a partnership that changes the channel mix, or see churn move in one direction for two months.

The template calculates lifetime contribution as monthly contribution divided by churn, because that's the conventional formula. It also shows contribution over 24 months, and that's the figure to plan with.

The reason is arithmetic. At 0.9% monthly churn the formula implies an average customer life of over nine years. Almost no early-stage company has the history to support a claim like that, and small errors in a small churn number produce enormous swings in the result. A lifetime value figure that drives a spending decision should always be capped at a horizon you would defend out loud.