Define the unit before calculating the multiple
“EV per user” is incomplete. A user may be registered, active during the month, paying or a unique payer after duplicates are removed. One person with several accounts is not necessarily several independent sources of revenue. Record the definition, the database snapshot date and the rules for excluding trial, free and suspended subscriptions. If the methodology changed, recalculate the history or mark an explicit break in the series.
The numerator must use the same perimeter. The value of an entire group cannot be divided by customers from only one segment without an adjustment. To value operations before servicing debt, EV is usually examined rather than share price alone. If excess cash was subtracted from EV, it should not be counted again as a resource that automatically improves every customer. Separating operating and financial claims matters even when the denominator is not monetary.
Registrations change the appearance of cheapness
A fictional service A has EV of 120 million roubles, 60,000 registrations and 20,000 active paying customers. That gives 2,000 roubles per registration but 6,000 per paying customer. B has EV of 90 million, 30,000 registrations and also 20,000 paying customers. That gives 3,000 per registration and 4,500 per payer. The apparent cheapness order changes solely because the unit definition changed.
The comparison of 6,000 and 4,500 is not finished. Assume A receives 800 roubles per payer per month and direct servicing costs are 500, giving a monthly contribution of 300. For B, payment is 600, costs are 400 and contribution is 200. Dividing customer value by current contribution gives 20 and 22.5 months respectively. This is a scale diagnostic, not an acquisition payback period: it still ignores churn, acquisition, fixed costs, taxes, investment and the time value of money.
Churn and the age of the base
In a hypothetical month, A loses 1,000 of its initial 20,000 payers; churn on the starting base is 5%. B loses 400, or 2%. New acquisitions must be shown separately: growth in the total does not erase losses among existing customers. Measure the same object—customers or revenue—throughout; revenue retention may include expansion of contracts with remaining buyers and answer a different question.
If no new customers are added and monthly retention is mechanically held constant, about 10,807 customers from A’s starting base and 15,694 from B’s remain after twelve months. This is not a forecast of a real cohort: constant churn is an exercise assumption and age differences are ignored. Compare cohorts of the same age and contract conditions. A free trial must not become a paying long-lived customer merely because a registration button was pressed.
Capacity also needs a specification sheet
A fictional power plant C has EV = 600 million roubles and installed capacity of 100 MW. The measure is 6 million roubles per MW. D has EV = 560 million at 80 MW, or 7 million per MW. C appears cheaper. But nameplate capacity does not show availability, load, tariff, fuel cost or operating life.
For a denominator-only check, suppose both plants have the same 8,760 hours in the hypothetical year. At a 40% capacity factor, C produces 350,400 MWh, while D at 70% produces 490,560 MWh. EV divided by annual generation is approximately 1,712 and 1,142 roubles per MWh of annual volume. This is not the selling price of electricity: the numerator is a stock of business value and the denominator is a flow for a particular year. Do not label the ratio simply “roubles per electricity.”
Quality, costs and boundaries
Before transferring a comparable multiple, determine whether the asset is operating, what investment is required for commissioning, its remaining life, network constraints and completion obligations. Two MW of identical capacity can require different spending to produce the same free cash flow. Services have analogous differences in customer solvency, discounts, servicing and replacement cost for departing users. A natural unit is a bridge to economics, not a substitute for it.
With zero selected units, the ratio is undefined. Incompatible customer bases or capacities cannot be repaired by averaging multiples. The final note should show the original natural metric, normalized base, date, quality and the next bridge to revenue, contribution and cash flow. If the data do not permit that bridge, the honest result is limited comparability. Customer count or capacity alone does not guarantee value, payback or a future market price.