Find out which customers actually make money.

The year was good by every measure the team celebrates. Revenue grew by 40%. Customer count nearly doubled. The gross margin in the monthly MIS held steady. And yet the bank balance at the end of the year looked a lot like the one at the start, the founder is still raising sooner than planned, and nobody can quite say where the growth went.

Usually it went into the customers who cost the most to win and serve and pay the least. They're often the fastest-growing segment, because they're the easiest to sign. A blended margin hides them perfectly.

A unit economics review takes the business apart by customer type, product and channel, and works out what each one actually contributes after every cost that moves with it. It's not complicated maths. It's mostly careful allocation of costs that are currently sitting in the wrong place, and a willingness to look at the answer.

Gross margin in most startup P&Ls is revenue less a narrow idea of cost of sales: hosting for software, food cost for a kitchen, consumables for a clinic. It usually leaves out costs that scale directly with each sale, like payment gateway fees, aggregator commissions, onboarding time, delivery, discounts and the support team that grows every time customers do.

Contribution margin puts those back. It's what each unit of revenue leaves behind to pay for the fixed costs of the business and, eventually, a profit. If contribution is thin, growth makes the company bigger without making it any closer to paying for itself.

Acquisition cost has the same problem in reverse. CAC is often calculated as ad spend divided by new customers, which leaves out the sales team's salaries, the tools, the agency retainer and the founder's own selling time. Fully loaded CAC is almost always higher, sometimes more than double, and payback periods move with it.

An illustrative software company with three plans. The founder sees a healthy blended gross margin, and the marketing team has been putting more budget into the Starter plan because it produces the most sign-ups. The company and every figure are invented.

Starter looks cheap to acquire, and it is, in absolute terms. But its support load is heavy relative to its price, so contribution is thin, and customers leave quickly. Payback takes nearly 16 months against a typical life of about 22. Across that life each Starter customer contributes around ₹53,000 against ₹38,000 spent winning them. That's barely positive, and it assumes every other cost in the business is paid by someone else.

Growth customers pay back in eight months and stay for years. Every rupee moved from Starter acquisition to Growth acquisition is worth several times more, even though Growth sign-ups look slow in a weekly marketing report.

The options that come out of a review like this are usually unglamorous. Raise the Starter price, or limit the support it includes. Offer an annual prepaid Starter plan so churn and payback both improve. Move marketing budget towards the channels that produce Growth customers. Stop celebrating sign-ups as a number on its own.

Founders often avoid raising prices because they fear losing customers. It helps to know exactly how many you can lose before the change costs money. The arithmetic is simple, and it depends almost entirely on contribution margin.

If your contribution margin is m and you raise prices by p, you can lose a share of volume equal to p divided by m plus p before total contribution falls. The same logic runs in reverse for discounts: a price cut of p needs volume to rise by p divided by m minus p just to stand still.

Two things stand out. Low-margin businesses, like food service, have the most room to raise prices and the least room to discount. And a 10% discount in a 25% margin business needs two-thirds more volume to break even, which is why festive offers in thin-margin businesses so often end with a record month and a worse quarter.

Where your prices include GST, as they usually do for consumers, do the arithmetic on the price excluding tax. A menu price rise of ₹20 is not ₹20 of extra contribution.

• ·GST on inputs in businesses that can't claim it back. Clinical healthcare services are exempt, so GST paid on rent, equipment and software becomes a real cost. Restaurants charging 5% GST without input credit carry the same problem.
• ·Aggregator commissions and the GST charged on those commissions, which for a delivery-heavy kitchen can outweigh the food cost difference between two menu items.
• ·Payment gateway fees plus GST on the fees, which look small per transaction and add up to a meaningful share of contribution on low-ticket products.
• ·Discounts given by sales teams at quarter end and never reported against list price.
• ·Dollar pricing with rupee costs, where margin moves with the exchange rate and gets read as a pricing problem.
• ·Onboarding and implementation time on larger customers, booked as salary overhead instead of cost to serve.

Not every company needs one right now. A business with a single product, one customer type and a clear margin can usually see its unit economics in the P&L. The review earns its keep when the business has become mixed enough that averages hide things.

• ·You sell more than one plan, product line, location format or service type, and nobody is sure which is carrying the others.
• ·Revenue is growing faster than contribution, or cash isn't improving as revenue grows.
• ·Marketing spend has risen and cost per sign-up looks fine, but payback feels slow.
• ·You're about to change prices, launch a new tier or introduce a discount scheme.
• ·An investor has asked for CAC payback or contribution by segment, and the numbers you have wouldn't survive a follow-up question.
• ·A large customer is asking for a discount and you don't know what they currently contribute.

The last one comes up more than you'd expect. Large customers negotiate hard because they know they're large. Sometimes the contribution they bring justifies the discount easily. Sometimes they're already the least profitable account in the business once implementation and support time are counted, and the right answer is a polite no.

It starts with the data you already have, not a new system. Invoice exports, the cost ledgers, payroll by team, marketing spend and a customer list with start and end dates are usually enough. The first job is agreeing which costs are variable and how shared costs like support should be allocated, and writing those rules down so the numbers can be rebuilt next quarter the same way.

Then the model gets built, the first results come back, and there's a conversation where the founder and the relevant leaders test them against what they know about their customers. That conversation matters. Allocation rules that look sensible on paper sometimes produce an answer everyone in the room knows is wrong, and it's far better to fix the rule than to defend the number.