How we verify

Every guide on this site is built from published testing, manufacturer specifications and retail listings rather than from products we own. That choice puts a burden on us: if we’re not the ones measuring, then everything we repeat has to be checked against the people who did. This page describes how that checking actually works, because a claim about rigour is worth nothing without the method attached.

Every product goes through a second pass that assumes it is wrong

A draft is written first, from sources. Then a separate review reads that draft with one instruction: find the problems. Not confirm the good ones, find the problems. It pulls the manufacturer page, the review that was cited, and a retailer listing, and it checks the product name, every number, every attribution and the current price against what those pages say today.

This isn’t a formality. Across the reviews run in the past week, the average product block came back with three to seven real problems, and the writing that generated them read perfectly well. That’s the point: fluent prose isn’t evidence of accuracy, and the only way to tell the difference is to check.

What the review is looking for

• Is this product still the product? Manufacturers replace models quietly. One pick here was discontinued in 2020 and its listing had quietly become a different camera. Another had been renamed by its own manufacturer with no other change.
• Does the number match the source? Refresh rates, capacities, warranty lengths and battery figures all drift between generations, and a specification from the newer model is one of the easiest errors to make.
• Did the review we cited actually say that? A quote here was once traced back to an aggregator page that had invented it and attributed it to a well-known publication. We now follow attributions to the original.
• Is the price real today? Not the list price, not a launch price, and not an expired sale. See below.
• Does the page contradict itself? A pro that argues with the prose beside it, or two picks that both claim to be the cheapest, is a correction even when both halves are individually true.

Prices are re-checked on a schedule, not when we happen to notice

Every product we recommend carries a current price range, and those ranges are swept weekly: every tracked product is re-checked against its retailer listing, anything that has drifted outside its published range is flagged, and a human judges every flag before a single number changes. A sweep reads the live retail pages and reports what a buyer would actually pay, which is regularly not what the manufacturer says. Recent sweeps moved a purifier’s range by more than two hundred dollars, corrected a robot vacuum that had been listed at roughly double its real price, and caught a console-class handheld whose maker had raised prices by three hundred dollars mid-year.

The price index is the public version of that dataset. It’s dated, and it’s recomputed from the guides themselves rather than maintained by hand, so it can’t quietly fall out of step with them.

Since August 2026 that recomputation is automatic rather than a job someone remembers to do. Finishing a sweep writes a verification date onto every guide it covered, and that write schedules the index to rebuild itself a few minutes later, once the batch has settled. The figures on that page are therefore never older than the last sweep.

Two rules in that rebuild are worth stating, because both are the kind of detail that decides whether a number can be trusted. Every price is parsed rather than copied: a price string that doesn’t contain an actual number is reported and the rebuild refuses to publish, instead of quietly skipping it and leaving the totals looking plausible. That check exists because two price bars once went live reading “~–42” after a scripting error ate the amount, and no other check on the site caught it, since a price is just a string to everything else. And each price is counted once, in its guide’s main department, so the per-department column adds up to the sitewide total printed above it — a handful of gift guides also sit in a topic department, and counting them twice would make the table contradict its own headline.

What we do not do

We don’t test products in a lab, and we never write as though we have. When a guide describes how something performs, that performance was measured by someone else and the guide says who. We don’t accept payment for placement, and no manufacturer sees a guide before it publishes. We earn money through affiliate links, which are marked, and which pay the same whichever product in a guide you choose, so there’s nothing to gain from steering you toward one.

We also don’t pretend the checking is perfect. It catches a great deal and it will still miss things, which is why the corrections page exists and why it’s public. If you find something wrong, the fastest way to get it fixed is to tell us.

Who does this

Before You Buy Guide is a small independent operation. There is no corporate test lab behind it: the sourcing, the price sweeps, the second-pass reviews and the corrections described above all follow the written process on this page — which is exactly why it is written down rather than taken on trust. No product on this site has been supplied by a manufacturer for review — every guide is built from public sources, as described above — and if that ever changes for a specific product, the guide itself will say so. Mistakes still happen; they’re logged on the corrections page, and the fastest way to report one we haven’t caught yet is to contact us.

Last updated August 22, 2026 · Rewritten for the new page design; the checks themselves are unchanged.