Why do blended margins mislead?

Here is a pattern that repeats across businesses that look nothing alike. Revenue grows, the gross margin in the monthly pack holds steady at a respectable number, and cash gets tighter rather than easier. Nobody can explain it from the P&L, because the P&L says the business is doing fine.

The explanation is almost always the same. The company is not one business, it is three or four, and the average hides the one that is losing money. Worse, the losing one is often the one growing fastest, because low-margin business is usually the easiest to win.

This is the most common finding in any first look at a company's numbers, and it is entirely avoidable.

Two problems with the number in your P&L

The first is that gross margin in most startup accounts is calculated on a narrow idea of cost of sales: hosting for software, food cost for a kitchen, consumables for a clinic, delivery staff for a services firm.

It leaves out the costs that exist only because a sale happened: payment gateway fees, aggregator commissions, sales commissions, delivery partners, onboarding effort, the support team that grows every time customers do, and unrecoverable GST on any of those. Put them back and you have contribution margin, which is what actually remains to pay for the fixed costs of the business.

The second problem is the average. A blended contribution margin of 60% might be one segment at 78% and another at 31%. The first is the business you want more of; the second may be losing money once the cost to serve is counted properly.

An average margin is a statement about the past mix of your business. It tells you nothing about which part to grow.

A worked example

Illustrative figures for a fictional software company with three plans. The founder sees a blended gross margin of 71% and considers the business healthy.

• ·Starter: ₹4,500 a month, ₹2,100 of variable cost to serve, so ₹2,400 of contribution, a 53% margin. Customers stay about 22 months.
• ·Growth: ₹18,000 a month, ₹4,300 of cost, so ₹13,700 of contribution, a 76% margin. Customers stay nearly five years.
• ·Enterprise: ₹1,20,000 a month, ₹41,000 of cost, so ₹79,000 of contribution, a 66% margin, with very low churn.

The averages were true and useless. Starter, which produces the most sign-ups and therefore attracts the most marketing budget, has the thinnest contribution and the shortest customer life. Its fully loaded acquisition cost of ₹38,000 takes nearly 16 months to repay against a customer who stays 22.

So the company's growth engine is the segment closest to losing money, and nobody could see it, because the blended margin looked fine and the sign-up chart looked excellent.

Where to split, by business

The right cut is whatever genuinely differs. In practice one or two of these matter in any business, and the rest are noise.

• ·Software: by plan, by customer size, and by whether the customer is on annual or monthly billing. Add usage-based costs by customer where the product calls paid models or APIs, because the heaviest users can be the least profitable accounts.
• ·Services firms: by client and by project type. Fixed-price work and time-based work have completely different risk, and one bad fixed bid can absorb the margin from three good ones.
• ·Food service: by channel first, then by outlet. A dine-in order and an aggregator order at the same menu price are not the same sale, and the gap is usually more than half the contribution.
• ·Healthcare: by service line and by payer. Cash, corporate, insurance and scheme patients pay differently, get discounted differently, and arrive as cash at very different speeds.
• ·Marketplaces: by category and by side of the market, with subsidies counted as a cost of the transaction rather than as marketing.

The costs people forget to allocate

Most of the difference between a useful analysis and a misleading one sits in a handful of costs that get left in overheads.

• ·Support and account management. Usually the largest hidden difference between segments, and the reason cheap plans lose money. Estimate it by asking the team to track where their week goes for a fortnight.
• ·Onboarding and implementation. Real effort, usually delivered free, often concentrated in the segment with the longest sales cycle.
• ·Payment costs. Gateway fees on small tickets are a meaningful share of contribution, and aggregator commissions can exceed food cost.
• ·Unrecoverable GST. For exempt healthcare services and restaurants on 5% without input credit, GST on inputs is a cost that never comes back, and it belongs in the unit economics rather than in a tax line.
• ·Delivery people's time in services businesses, at fully loaded cost including employer contributions rather than at salary.
• ·Discounts, which are a reduction in price rather than a marketing expense, and should be visible against list price.

None of these needs perfect precision. An allocation rule that is roughly right and written down beats an exact number nobody can reproduce next quarter.

What the split usually reveals

Three findings come up again and again, in businesses that have never done this before.

The first is that one segment is subsidising another, and nobody chose it. Often a large customer negotiated hard years ago and the terms never changed, while smaller customers pay list price and get less support.

The second is that growth is coming from the wrong place. Marketing spends where sign-ups are cheapest, which is usually where contribution is thinnest and churn is highest, so the business grows and gets no better.

The third is that a segment everyone assumed was strategic is simply unprofitable. Enterprise customers with long implementations and heavy support are the usual candidate, and the answer is rarely to abandon them: it is to price implementation, cap support, or accept the economics deliberately because those logos win other business.

How the mix drifts without anyone deciding

Segment economics rarely go wrong in one move. They drift, and the drift is invisible in an average.

A marketing team optimises for cost per sign-up, which pushes spend towards the cheapest segment to acquire. Sales, paid on revenue rather than contribution, discounts where discounting closes deals fastest. Support absorbs whatever arrives without anyone costing it. Product adds features that increase usage costs on plans that were priced before those features existed.

Every one of those decisions is reasonable on its own terms. Together, over four quarters, they can move a company from a healthy mix to one where the growing half of the business is the unprofitable half, and the blended margin barely moves because the good segment is carrying the average.

The fix is not to stop any of those teams doing their jobs. It is to give them a number that reflects the whole picture: contribution by segment, reported monthly, so cost per sign-up is never the only measure anyone optimises against.

The two-hour version

This analysis has a reputation as a project. The first useful pass is an afternoon, and it usually answers the question well enough to act on.

1. 01Export revenue by customer for the last quarter from your billing system.
2. 02Tag each customer with a segment: plan, size, channel, location or payer. Three or four segments, no more.
3. 03List every cost that varies with a sale from the ledgers, and allocate each to segments with one written rule. Where a rule is a guess, write that it is a guess.
4. 04Ask support and delivery, for two weeks, roughly what share of their time goes to each segment. Use that split for their cost.
5. 05Divide contribution by revenue for each segment, and put the segments side by side.
6. 06Then add one column: how much of last quarter's new business came from each. That column is usually the finding.

A rough version done this quarter beats a precise version done next year, because the decisions it changes, where marketing spends and what the thin segment is priced at, are decisions being made every week in the meantime.

What investors do with this

At seed, a blended margin is usually accepted. From Series A onwards it is not, and a founder who has never split it is at a disadvantage in a room where the analyst has already tried.

Diligence teams rebuild contribution by segment from raw transaction data, and they look for two things: whether the mix has been shifting towards worse economics, and whether the company knew. Discovering it themselves, in a company that reports only averages, changes the tone of the whole process.

The opposite is also true. A founder who opens with the split, names the weak segment and explains what is being done about it, moves the conversation to strategy rather than suspicion. It costs an afternoon a quarter to be that founder.

What to do with the answer

1. 01Move acquisition spend towards the segments with the best contribution relative to what winning them costs. This is usually the single largest gain, and it costs nothing.
2. 02Reprice or repackage the thin segment. A higher price, a smaller support entitlement, or an annual prepaid option all improve it without losing the customers who value it.
3. 03Attack the cost to serve, which is often a product problem: better onboarding, fewer support tickets, self-service for things that currently need a human.
4. 04Review your largest customers specifically. Concentration and poor margin frequently travel together.
5. 05Put contribution by segment in the monthly pack so the mix is visible every month rather than once a year.
6. 06Then decide, deliberately, whether any segment should be retired, and what that would do to fixed costs.

Doing it without perfect data

Founders often postpone this analysis because the data feels inadequate. It rarely is.

Revenue by customer exists in every billing system. Direct costs are usually identifiable from the ledgers. The uncertain part is allocation of shared costs, and that can be estimated from a two-week sample of how the support and delivery teams spend their time.

Write the allocation rules down, apply them consistently, and mark the estimates as estimates. An analysis that is 90% right and reproducible each quarter changes decisions. One that waits for perfect data never gets done, and the average keeps hiding what it hides.

One warning: do not let the exercise become a cost accounting project. The goal is a decision about pricing, spending and mix, not a perfectly attributed P&L by segment. Three or four segments, a handful of allocation rules, and a monthly line in the pack is the version that survives contact with a busy company.

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.