What is an unprofitable online business worth?

A profit multiple is sold price ÷ trailing-twelve-month profit. At zero profit that division is undefined. At a loss it returns a negative number, and a "−3× multiple" is not a small multiple or a cheap business — it is arithmetic that has stopped describing anything. There is no repair for this. You cannot value a loss-making business on a profit multiple, however the listing chooses to present it.

What has not gone away is the sale price. Businesses at break-even and worse change hands constantly, for figures buyers are willing to state. So the honest move is not to force the multiple; it is to change what you are dividing by, or to stop dividing and look at prices directly.

The rule is one line of the code and worth stating plainly: the estimate is computed on profit when the profit you enter is above zero, and on revenue otherwise. Zero, blank and negative all take the same branch. If neither figure is positive, no estimate is produced and the API says Provide a positive ttm_profit or ttm_revenue. rather than returning something.

Two consequences worth knowing before you use the calculator on a loss-making business:

• A loss is not partially counted — it is not counted. Entering −$8,000 of profit and $40,000 of revenue produces exactly the same estimate as entering nothing in the profit field. The size of the loss does not push the estimate down, because the revenue basis has no way to see it.
• The only signal that this happened is the basis label. The estimate card names the basis it used — "0.9× revenue" rather than "0.9× profit". There is no separate warning, so that word is the thing to read.

That is a deliberately blunt design: a fallback that quietly blended a loss into a profit-based range would produce a number nobody could reconstruct. A basis that is named on the output can at least be checked.

Loading live revenue-basis coverage from /api/stats…

Coverage is measured over the paid report's window — the most recent 200 priced sales of that type — and both bases apply the same five-comp floor. Where a category shows too few revenue-basis comps, the fallback has nothing to fall back to and you will get no estimate at all. How many sold comps a valuation needs covers that floor.

When a marketplace page states a loss, we store it. Flippa prints one with the minus ahead of the currency code — Monthly Profit −USD $6 /mo — and until recently our parser could not see that minus, so the row came out looking like a page that stated no profit at all. It now reads the figure as the negative it is.

That fact then stops where it should. A multiple is only computed from a profit above zero, so a loss-making sale contributes its price to the record and no multiple to any median. Both things matter: suppressing the loss would have thrown away something the source page actually says, and letting it into a multiple would have corrupted every median in its category. Where such a row exists, a Pro comps export carries the negative figure beside an empty multiple_profit, so the loss is visible as a loss rather than as missing data.

One asymmetry, for completeness: a negative revenue is read and then dropped rather than stored, because a negative revenue is not a reading worth guessing at.

If not earnings, then what? In practice the price of a break-even or loss-making asset rests on one of a short list of things, and it is worth being explicit about which one you are buying, because they carry very different risk:

• Revenue that a cheaper operator can keep. The most common case: the revenue is real and the cost base is the seller's, not yours. This is a bet on your own operating costs, and it is the bet the revenue basis loosely prices.
• An audience or traffic. Priced on the assumption you can monetize it better. Ask what has already been tried, not what is theoretically possible.
• The asset itself. A domain, a codebase, a mobile app's install base, a brand. Here comps on price are the right tool and always have been — the entire domain market is priced this way, with no earnings anywhere. See what is a domain name worth.
• Time already spent. Occasionally legitimate — an integration, a licence, an approved app-store listing — and often just the seller's sunk cost, which is not a value to you.

Two framings that stop working, and knowing they have stopped is the point: payback period is undefined when there is nothing to pay you back, and "N× profit" cannot be sanity-checked against anything. What remains is the price distribution for similar assets, plus your own written assumption about what changes after the sale.