How We Check Data

The problem this method solves

Advertising network data has a circulation problem. A figure appears in a directory listing, gets copied into a comparison article, gets copied again into a listicle, and within a year the same number sits on twenty pages with no one having checked it since. When the company changes it, none of those pages update. The figure keeps circulating, gaining apparent authority from repetition alone.

Platform policy has the opposite problem: it is documented well but changes often, and the documentation is long enough that most writers summarise it from memory or from another summary. The result is content that describes last year's rules confidently.

Our method is designed against both failures. Nothing is taken from a page that itself took it from somewhere else, and nothing goes up without a date attached that tells you how stale it might be.

What counts as a source

Sufficient on its own

The company's own current help center, advertiser or publisher documentation, published terms and conditions, and official advertising policy pages. A hard figure — minimum deposit, payout threshold, blocked-market list, certification requirement — needs a source at this level or it does not get stated as fact. We record the exact URL and the date read alongside the figure.

Usable with attribution

The company's own blog posts, product announcements and marketing pages. These often contain figures the documentation omits, which makes them useful, but they are written to persuade and they are not maintained the way documentation is. When we rely on this tier we say so in the body text so the reader knows the figure is softer.

Usable for patterns only

Dated reports from users on independent platforms. These are attributed as reports, never converted into flat statements of fact, and we look for repetition across many dated accounts rather than treating any single account as evidence. One complaint tells you about one experience; the same complaint fifteen times tells you about a process.

Never a source

Network directories, comparison sites and competitor articles. We read them to find out what is being claimed and to build a list of figures worth verifying, and that is their entire role. A figure that appears only in this tier is treated as unverified regardless of how many places repeat it.

The checking process

1. Build the claim list. We assemble every figure the page will need — deposits, thresholds, pricing models, formats, country lists, certification requirements — before opening a single source. Working from a list prevents the common drift where a writer verifies the easy facts and improvises the rest.
2. Locate the primary document. For each claim we find the company's own page that should carry it. This is deliberately slower than searching for the number, because searching for a number surfaces the pages that already repeat it, which is exactly the trap the method exists to avoid.
3. Read the surrounding text, not just the figure. Deposit minimums are usually conditional on payment method; policy rules are usually conditional on market, user age or content category. Lifting a number without its conditions produces a fact that is technically present in the source and still wrong in practice.
4. Record source and date. Every confirmed claim gets the URL it came from and the date it was read. These records are what allow a later reviewer to re-check quickly and what let us tell a reader honestly how old a figure is.
5. Check the company against itself. Where a company publishes a figure in more than one place, we compare them. Marketing pages and documentation frequently disagree, and when they do the discrepancy is more interesting than either number on its own, so it goes into the text rather than being resolved silently.
6. Mark what could not be confirmed. If a figure is not published at Tier 1 or Tier 2, it does not become a verified fact. Either it is left out or the page states which weaker source supplied it. We never launder a directory number by placing it in a table next to verified ones without comment.
7. Adversarial re-read. A second pass challenges each claim: is the source the company's own, is it current, does the stated condition match the source, and would this sentence still be true if the company changed the policy tomorrow. Anything that survives only on plausibility is cut.
8. Date and schedule. The page ships with its verification date visible. Platform policy pages go on a shorter re-check cycle than network pricing because they change more often and their changes are announced.

A worked example

While building our comparison of networks that carry dating ads traffic, one figure appeared consistently across directory listings: a $50 minimum deposit for a well-known network. It was repeated widely enough to look settled.

The company's own help center gave a different number. Its published article on minimum deposits stated $100 for most payment methods and added that the account will not activate below that threshold — a detail no directory carried, and the detail that actually matters to somebody planning a first test. The $50 figure was not a typo somewhere; it was an old value that had propagated and then outlived the policy.

Two things follow. First, the published table uses $100 and cites the company's own page. Second, the case demonstrates why the directory tier is excluded on principle rather than case by case: the wrong figure was not marginal or obscure, it was the consensus, and consensus in this category is produced by copying rather than by checking.

A second case cut the other way. For another network, we initially concluded that no minimum deposit was published anywhere official, and said so. That was our error — the figure was in the company's advertiser documentation, and our search had simply missed it. The correction changed both the figure's status and a claim we had made about the company's transparency. It is the clearest illustration we have of why the process ends with an adversarial re-read: failing to find something is not the same as it not existing, and the difference is easy to miss when you are the one who did the looking.

What we do about staleness

Verification has a shelf life. A minimum deposit confirmed today can change next quarter without an announcement, and a policy page can be revised without a changelog entry. We handle this in three ways rather than pretending it is solved.

Every page states its verification date in a place you cannot miss, so you can weigh the figures accordingly. Every figure that came from a source weaker than the company's own documentation says so in the body text. And every table figure is linked or attributed to the document it came from, so checking it yourself takes one click rather than a fresh search.

The same method produced the platform policy breakdown on our page about dating ads traffic, where every country list and certification rule traces to the platform's own policy document. If you find a figure here that no longer matches the source, tell us — the correction route and our response times are described in the editorial policy. Reader corrections are the fastest signal we get that something has moved.

What this method does not give you

The person who does this checking is named on every page, and his background is on the author page. Document verification establishes what a company states, not what it does. We can confirm that a network publishes a $100 minimum and accepts a vertical; we cannot confirm from a document how its moderation behaves on a borderline creative or how quickly a payout actually arrives. Where those questions matter, we say that the answer requires operational testing rather than implying our research covers it.

We also do not present research as first-hand testing. Reading a company's terms is real work and produces real findings, but it is a different activity from running a campaign, and blurring the two is the most common form of dishonesty in this category. When a page describes what a document says, it says so plainly. More about who does this work is on the about page.