Where our data comes from

How does Rival Radar check a competitor's website?

You choose which of a competitor's public pages to watch: pricing, features, homepage, changelog, careers and so on. Rival Radar visits each one on a schedule set by your plan, always from the same country, so a page that shows different prices in different places doesn't look like it changed. It keeps the text of each version, compares it with the last one, and ignores noise like cookie banners and rotating adverts. When something real changes, you get a plain-English summary and can see exactly which words changed.

How does Rival Radar read a competitor's prices?

For pricing pages, Rival Radar also reads off each plan and its price, so you can see what they charge today and how that's changed. Three rules keep this accurate:

• Every plan name and price has to appear word for word on the page. Anything that doesn't is thrown away.
• A new price only counts once it's been seen on two checks in a row. Companies often test prices on some visitors, and this stops a test from showing up as a real change.
• If a page can't be read reliably, we say so and show nothing, rather than show prices that might be wrong. You still hear about the page changing.

Every price links to the page it was read from. There's more on why a price you only saw once can't be trusted in how to tell if a competitor is testing prices.

Where does the history from before I signed up come from?

The Internet Archive runs the Wayback Machine, a free public service that has been saving copies of websites for many years. When you add a competitor, Rival Radar looks for saved copies of the pages you're watching from the last two years, one per month at most, and turns the differences between them into history. We fetch them one at a time, so we're not a burden on a free service.

What the archive has depends on what it happened to save. Well-known sites are saved often; smaller ones might have a few copies or none at all. Sometimes a page was saved before its content finished loading, and those copies are skipped. So some competitors will have plenty of history and some won't have any. We always tell you which, and every historical change links to the exact archived copies it came from. The archive is also sometimes slow or offline. When that happens we wait and try again later, for up to a few days.

Where do the hiring signals come from?

Many companies post their jobs through a job board such as Greenhouse, Lever or Ashby, and link to it from their careers page. If the careers page you're watching does that, Rival Radar reads that board once a day and shows open roles by team, what's new, and what's closed. We only ever use a board the company links to itself, never one guessed from the company name, since different companies can have similar names. A role only counts as closed once it's been missing from two daily checks in a row, so a glitchy day on their board can't make it look like they've stopped hiring. If they don't use one of these boards, we say so. More on what hiring tells you is in what a competitor's job postings tell you.

How are the monthly competitor reads written?

On paid plans, you get a short monthly read on each competitor: what they appear to be doing, what you could do about it, and points for your sales team. It's written by AI, but only from what Rival Radar actually recorded that month. Every point has to cite the specific changes behind it, and you can open each one to see the change and follow its link to the source. Any number in the read has to appear in the changes it cites, so it can't invent figures like “a 20% price rise”. If too little happened, the read says it was a quiet month rather than padding it out.

How is AI used, and how do you stop it making things up?

AI does three jobs in Rival Radar: it writes the one-line summaries of changes, it reads plans and prices off pricing pages, and it writes the monthly reads. It isn't trusted on its own. Everything it produces is checked against the source before you see it: prices and plan names must appear on the page, summaries can't state numbers that aren't in the change, and monthly reads must cite real recorded changes. Anything that fails is dropped. The AI only ever sees public page text, never your account details. The privacy policy has the full detail.

What we don't do

• Log in to anything, or look at anything that isn't publicly visible.
• Guess. If we can't read something reliably, we tell you.
• Show you a change without a way to check it. Everything links back to where it came from.

If something looks wrong, please tell us. We'd much rather fix it than have you rely on it.