Agree on the boundaries first
Write down which costs change with an order, which relate to customer acquisition and which support the company as a whole. Delivery cost may include the operator and packaging but not the office. That can be appropriate for an order contribution calculation if the office remains in the next level of analysis. The remainder after some costs cannot be called net profit.
Acquisition cost also needs a denominator. Total advertising budget divided by registrations differs from the budget divided by new paying customers. If spending includes brand development and retention of existing buyers, mechanical division by new customers mixes different tasks. Without a breakdown, show a range rather than pretending to have a precise metric.
One order and the customer relationship
A hypothetical service receives 1,000 roubles per order. Fulfillment and other variable costs are 700, leaving a contribution of 300 per order. Acquiring a new paying customer costs 900. Excluding fixed costs and the value of time, three orders merely cover initial acquisition. The first positive contribution does not by itself prove that the customer has paid back acquisition.
Suppose that of 100 new customers, 40 place only one order, 30 place two and another 30 place four. This produces 220 orders and 66,000 roubles of contribution. Acquisition for the group cost 90,000. The result before common costs is −24,000. Every order has a positive contribution, but the customer group has a negative total. That distinction disappears when only delivery margin is shown.
Why a cohort is needed
A cohort groups customers by when or under what conditions the relationship began. Compare them at the same age: the first three months of a new group with the first three months of an older one. Do not compare mature customers’ lifetime total with an incomplete recent result and call it falling efficiency. Nor should every new customer be treated as a future long-lived customer merely because the most durable customers remain among the old group.
Check retention, order frequency, discounts and direct costs separately. Average payment can rise because customers with small orders leave, not because the product improved. A customer’s “lifetime value” is a forecast of future contribution, not cash already received. Its assumptions must be visible: relationship duration, repeat-purchase probability and servicing costs.
From a good unit to bad scale
Even positive cohort economics does not automatically cover shared infrastructure, taxes and investment. Check whether increased volume requires a new warehouse, contact center or liquidity buffer. If the model includes only variable costs, call it that. A positive unit result is not proof of positive free cash flow for the firm.
Management-presentation data should be checked against disclosures about the business, expenses and result drivers. The SEC suggests reading management’s discussion of financial condition together with the business and risk descriptions. This is a guide to working with primary documents, not confirmation of a particular acquisition calculation.[101](#source-cv-10k) If the company changes its definition of “active customer,” recalculate the historical series or explicitly call it non-comparable.
Finish with a table of three levels: order contribution, cohort result after acquisition and business result after infrastructure. For each level, state which costs remain excluded. This structure permits a discussion of growth limits without promising that a larger customer base must create value. The example sets no payback standard and does not recommend buying profitable or unprofitable companies on one metric.
Reverse-check an advertising claim
If a presentation says that a customer pays back quickly, reconstruct the sequence of cash events. Advertising may be paid before the first order, while a return may occur after the stated contribution is recognized. A positive cumulative result at the end of observation does not show the maximum deficit that required financing during the period. That requires a calendar, not only a cohort total.
Also ask how many sales the campaign actually added. A customer might have arrived without it; dividing the whole advertising budget among all new buyers then does not measure the causal effect of advertising. Without an experiment, do not assign the calculation more precision than the data allow. Show the descriptive metric and the campaign-impact hypothesis separately: they are useful but answer different questions.
Related foundations: [Fundamental analysis: check a business history with numbers](/guides/fundamental-analysis/); [Why profit is not cash](/guides/profit-vs-cash).