पाठशाला Pathshala · हिसाब Hisāb, Unit economics · Lesson 26 · Scale

Unit economics across business lines and geographies

A company with three products or six cities has one P&L and several businesses inside it. How to split cost so each segment’s numbers tell you which to feed, which to fix and which to cut.

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

Somewhere between the second city and the third product, the company P&L stops answering the questions the founders are asking. It says the company lost ₹40 lakh last quarter. It does not say whether Pune is paying for Jaipur, whether the enterprise line is paying for the self-serve one, or which of them deserves next quarter’s money. That answer lives in how shared cost is split, and most companies split it badly or not at all.

This lesson sorts cost into three kinds, sets out how to allocate each, works one company with three cities through three allocation methods, and ends with the rule for feeding, fixing or cutting a segment and the monthly review that applies it.

Why the blended P&L cannot answer the question

A blended P&L averages good segments with bad ones. A company growing in a profitable city and a loss-making city can show an improving margin for a year while the loss-making city doubles in size, simply because the profitable one grew faster. The numbers are true and the conclusion drawn from them is wrong.

Local businesses have a particular reason to look city by city. In 13 Metrics for Marketplace Companies, Andreessen Horowitz partners point out that local network effects reset in each new geography, so older markets should retain users better, and that for these businesses CAC should fall and the organic share of users rise market by market over time. A company that only reads its blended numbers cannot tell whether that is happening in each city or only in the first one.

The same is true of business lines. A software line and a services line, a subscription product and a marketplace, a domestic book and an export book have different margins, different costs to serve and different futures. Each deserves its own unit economics, built on the [contribution margin](/library/contribution-margin-first-number-to-know) the company already computes for the whole.

Direct, attributable and truly shared

Put every line of cost into one of three kinds. Direct costs exist only because the segment exists: product cost, delivery and fulfilment, the city team, local marketing, the dark store or warehouse lease in that city. They are charged to the segment in full and nobody argues about them.

Attributable costs are shared but have a measurable driver. Cloud cost follows requests or orders. Central customer support follows tickets. A shared warehouse follows units shipped. A finance team’s collections work follows invoices. These are charged to segments on the driver, and the choice of driver is the most important decision in the whole exercise.

Truly shared costs have no driver worth measuring: the founders, head office rent, the brand campaign, the audit fee. These can be spread on revenue so the segment P&L adds up to the company P&L, or held as a single unallocated line beneath the segments. Both are defensible. What is not defensible is spreading them on revenue and then cutting a segment because the spread made it look unprofitable.

Allocating what is truly shared

Revenue is the default key because it is easy, and it is usually wrong for attributable costs. A city with low order values generates more orders, more deliveries and more tickets per rupee of revenue than one with high order values, and a revenue key hides that. The fix is the one Robert Kaplan and Steven Anderson set out in Time-Driven Activity-Based Costing in the Harvard Business Review in 2004: estimate a cost per unit of capacity, such as the cost of a support minute, and the time each activity takes, then charge each segment for the activity it consumes. Their method needs two numbers per resource group rather than the employee surveys that sank earlier activity-based costing projects, and it shows unused capacity explicitly instead of hiding it in the rate.

For a startup the method fits on one page. Support costs ₹6 lakh a month and handles 3,800 tickets, so a ticket costs about ₹158. Technology costs ₹8 lakh and serves 37,000 orders, so an order carries about ₹22 of it. Charge each city its tickets and orders at those rates. Review the rates once a quarter. Precision beyond that is rarely worth it; the aim is to stop a revenue key from moving cost from expensive segments onto cheap ones.

Companies that list in India meet the same question in the accounts. Ind AS 108, the operating segments standard, requires segment results to be reported on the measures the chief operating decision maker actually reviews when allocating resources, with a reconciliation to the financial statements, and treats a segment as reportable when it passes 10 per cent of combined revenue, profit or loss, or assets. A company that already runs its segment P&L this way monthly has nothing to rebuild when it lists.

One company, three cities

A quick-service food brand runs kitchens in Bengaluru, Pune and Jaipur. Bengaluru makes ₹60 lakh of revenue a month at a 25 per cent contribution margin, Pune ₹30 lakh at 20 per cent and Jaipur ₹10 lakh at 10 per cent, after food cost, delivery, kitchen staff and local marketing. Orders run at 20,000, 12,000 and 5,000, and support tickets at 1,500, 1,400 and 900; Pune’s aggregator mix generates more complaints per order. Shared cost is ₹18 lakh a month. The company makes ₹4 lakh.

Allocated on revenue, Pune earns ₹60,000 and Jaipur loses ₹80,000. Allocated on orders, Pune earns ₹16,000 and Jaipur loses ₹1.4 lakh. Allocated on activity, technology by orders, support by tickets and head office by revenue, Pune is at breakeven and Jaipur loses ₹1.9 lakh. The same company, the same month, and the middle city goes from profitable to breakeven depending on a choice made in a spreadsheet.

Now use the last slider. Jaipur loses ₹1.9 lakh after its activity allocation, but most of that allocation is head office and technology that will not shrink if the kitchen closes. If 30 per cent of it goes away, closing Jaipur removes ₹1 lakh of contribution and saves ₹87,000 of shared cost, and company profit falls from ₹4 lakh to ₹3.9 lakh. The city that looks like the problem is paying a little of the rent.

Allocate shared cost to see which segment is expensive to serve. Never cut a segment for a cost that will stay behind when it goes.

Feed, fix or cut

Three numbers decide what to do with a segment. The first is contribution after direct and attributable costs: if it is negative, the segment loses money on every order before head office costs a rupee, and growth makes it worse. The second is the trend in that contribution over the last two or three quarters, by cohort of customers or by month since launch; a new city usually starts behind and should be judged against where older cities were at the same age, which the [city launch lesson](/library/expanding-to-new-cities-launch-playbook) covers. The third is fully allocated profit, which says whether the segment pays its share of the company.

Feed a segment whose contribution is positive and rising and whose fully allocated profit is positive or close. Fix one whose contribution is positive but whose activity cost is high: Pune’s tickets per order are the problem to solve, not Pune. Cut one whose contribution is negative with no improving trend, and when cutting, count only the shared cost that truly leaves. Pause expansion into new segments while any existing one sits in the cut category without a plan, because the next city will usually look like the weakest current one for its first year.

The monthly segment review

On the fifth working day of each month, after the books close, publish a one-page segment P&L: revenue, contribution, attributable cost on its drivers, contribution after attributable cost, allocated shared cost and profit, one column per segment and one for the company, with the last three months beside each. Below it, the driver rates for the quarter: cost per ticket, cost per order, cost per server hour.

The founders spend the review on the segment that moved most and ask which of the three numbers moved it. Once a quarter, put each segment into feed, fix or cut, write down the one action for each, and check the previous quarter’s actions. Keep the allocation method the same from month to month; a company that changes its keys every quarter is choosing the answer before it reads it. The segment P&L should sit beside the company P&L in every board pack, built as the [P&L lesson](/library/p-and-l-a-founder-can-read-and-defend) describes.

Nothing here is legal, tax or investment advice. The three-city company is illustrative; its figures are the figure’s defaults so that every number in the text can be reproduced.

Sources

1. Robert S. Kaplan and Steven R. Anderson, Time-Driven Activity-Based Costing, Harvard Business Review, November 2004 — Two inputs per resource group: cost per unit of capacity and time per activity; accounts for unused capacity.
2. Institute of Chartered Accountants of India, Board of Studies, Ind AS 108 Operating Segments — Segments reported on the measures the chief operating decision maker reviews; 10 per cent thresholds for revenue, profit or loss and assets.
3. Jeff Jordan, Li Jin, D’Arcy Coolican and Andrew Chen, 13 Metrics for Marketplace Companies, Andreessen Horowitz, February 2020 — Local network effects reset in each geography; unit economics should improve market by market.