The measurement that keeps this feed off sale

What this page is4,736 rows held · newest sealed read 2026-08-30

• 61 of the 249 company-and-address pairs with at least 60 stored copies produced a different copy on every single read. They cover 49 companies.
• 2 produced the identical copy every time. The remaining 186 sit somewhere between the two, which is what a page that genuinely changes now and then looks like.
• A price page does not change every day. Software vendors reprice a few times a year. So a run of reads where no two copies match is measuring the page furniture — a rotating quote, a session id, a cache-busting number — and not the prices on it.
• We hold 4,736 dated copies behind this page, 10 Jun 2026 to 30 Aug 2026. Every one is kept. When the detector can tell a price from the furniture, these same copies get re-read; nothing has to be collected again.

Real rows out of our sealed copies

These are rows we read out of dated copies we keep ourselves. The live source shows today only, so once a row moves, what it said before is gone from the place you would go to look.

What this page cannot tell you

• This feed cannot yet tell you that a price moved. The detector behind it compares a fingerprint of the whole page, so a rotating customer quote, a session id in a script tag or a cache-busting number counts exactly the same as a price changing. Nothing on this page is a count of price changes, and no row here says a price changed. That is also why there is nothing to buy: a monthly charge is a promise, and we will not make it until we can tell the two apart.
• “Differs” is not “changed”. The fourth column counts how many of the stored copies are not byte-identical to each other. It is not a count of edits, it is not a count of price moves, and a page that differs on every read has almost certainly not been rewritten every day.
• A pair needs 60 stored copies to appear here. Three reads that all differ is not evidence of anything, so shorter runs are left off rather than counted. That means this page understates the problem, and we would rather it did that than the other way round.
• We guess two addresses per company and most companies only have one. We ask every site for a pricing page and a plans page. A company that publishes one of those and not the other gives us a 404 for the other one forever, and that is our guess being wrong, not the company hiding anything. Every count on this page is per COMPANY and needs both addresses to have failed, so a wrong guess on its own never puts anyone on it.
• Some stored copies are cut short. A page bigger than our size cap is stored up to the cap. Where the cut lands can move between reads, which is one more way two copies of an unchanged page fail to match.

More from this feed

• Up one level: B2B price-page changesThe whole feed, its price, and how the file arrives.
• What is and is not in this feedEvery source we read for it, how often, and the ones we refuse to collect.