How to Verify a Seller's Numbers Before Buying an Online Business

Every listing shows numbers. Revenue, profit, traffic, subscribers. The question that decides whether you overpay is not what the numbers say, but what stands behind them, and in the AI era that question has become harder, not easier. A convincing dashboard screenshot can now be generated in seconds, and even honest sellers curate. Verification is the work of moving a claim from "asserted" to "evidenced," and it is more difficult, more expensive, and more judgment-dependent than most buyers expect. This guide explains what that work involves, where doing it alone hits structural limits, and what a professionally verified analysis provides instead.

Does this sound familiar?

You have shortlisted a deal. The P&L looks clean, almost too clean, and you cannot articulate why that bothers you. You have spent two evenings cross-referencing a spreadsheet against screenshots, and you are less certain than when you started. You pasted the listing into a chatbot; it was enthusiastic, then you asked again and it was cautious. You found a number that seems high, and you have no idea whether it is high for this niche or normal. You are about to commit five or six figures, and everyone in the transaction, the seller, the broker, the platform, gets paid when you say yes. Every one of these is the same underlying problem: you have numbers, and you do not yet have evidence.

Why can't I just trust the listing?

Because a listing is a sales document, and it does the job sales documents do: it presents the figures most favorable to the asking price, over the window most favorable to the asking price. That is not an accusation of dishonesty. It is selection, and it is exactly what a good listing should do. The month a decline started, the customer that churned, the ad spend behind the growth curve: these live in the underlying records, and summaries are not underlying records. The buyer's job begins where the listing ends.

Marketplace verification helps here, and it is a genuinely valuable filter: it confirms the asset is real, the seller controls it, and headline figures have support. Buy on platforms that verify. What remains yours is the layer above it: whether the numbers hold for your purchase decision at this price. Trend quality, concentration, transferability, the questions specific to you as the buyer. Verification of the listing and diligence on the deal are two different jobs, and the first being done well does not complete the second.

What does "verified" actually mean?

A useful way to grade any number is by what stands behind it. Seller-asserted: claims and screenshots. Marketplace-verified: confirmed under the platform's process, a real filter and a meaningful step up. Source-verified: read from the system of record itself, or reconciled against independent data the seller does not control. Each tier upward removes a category of ways a number can be wrong, and each tier upward costs more access, tooling, and skill to reach. Most listings live in the first two tiers. Most purchase decisions deserve to be made from the third, and reaching the third is where buyers underestimate the work.

Why is independent verification hard to do yourself?

Four barriers, and they compound.

Access. The decisive evidence sits inside systems only the seller controls. Getting to it means knowing exactly what to request, in what format, over what period, and recognizing when what arrives has been trimmed. A vague request gets a curated answer; a precise, professional request gets the truth or a very informative refusal. Most buyers get one chance to make that request well.

Tooling cost. Cross-checking claims against independent data requires licensed datasets: traffic intelligence, search data, market and trend records. These are professional subscriptions, often hundreds of dollars a month across several vendors, priced for people who run many deals through them. Buying them for one acquisition inverts the economics; the free alternatives are estimates with error bars wide enough to be dangerous at negotiation time.

Benchmarks. A verified number is half an answer. A 62 percent margin, 4 percent churn, one page carrying a third of revenue: good or bad? That depends on the asset class and the niche, and the comparison data is accumulated across many deals, not looked up in an afternoon. Without benchmarks, verification tells you a number is real without telling you what it is worth.

Calibration. The hardest part is knowing which deviation matters. Every business has anomalies; most are noise, a few reprice the deal. Telling them apart is pattern recognition built from watching deals go wrong, which is precisely what a buyer on their first or second acquisition does not have yet, and cannot subscribe to.

Can't an AI chatbot verify the numbers for me?

A chatbot is a reasonable first pass and a poor verifier. It works from what you paste plus what its model learned in training, which may be months or years old, so it cannot check a claim against live, independent data, and recency is exactly what decides questions about trends and comparables. It tends to take pasted material at face value and agree with the thesis you bring, which is the opposite of a verifier's job. And it is nondeterministic: the same listing pasted twice can return two different verdicts with equal confidence. Verification needs the one property a chat transcript does not have: numbers that trace to a source.

What can I do myself?

Two things, and they are worth doing on every deal. Ask for exports, not screenshots: structured files from the platform itself, covering at least twelve months, are harder to curate and faster to check. And reconcile one month deeply rather than twelve superficially: pick a single month and trace it from platform record to P&L to bank; discrepancies rarely confine themselves to the month you chose. These steps catch the crude problems. What they cannot give you is source access done right, independent cross-checking, benchmarks, or the calibration to know which finding reprices the deal. That is the distance between a first pass and diligence.

What does a professionally verified analysis provide?

Everything above, as a service, for one deal, at a fixed price. A Dealytix report grades every material input by evidence quality and states what stands behind each number. Where the analysis warrants it, we arrange view-only source access with the seller directly: the request made precisely, the data read at origin, your relationship with the seller kept clean for the negotiation. Claims are cross-referenced against independent data sources, including licensed datasets we pay for, queried at the time of your analysis, not recalled from a model's memory. What cannot be verified is scored as exactly that, never quietly assumed. The result is not a promise that nothing is wrong; it is a map of how much weight each number can bear, which is what a negotiation actually runs on, and what turns "the listing says" into "the evidence shows."

Common questions

A clearer picture before you commit

Send us the listing URL. Independent analysis delivered within 48 hours — before you bid, before you offer, before you sign.

Independent · Evidence-graded · 48-hour delivery · Not financial advice