Prompts to Detect a Competitor Pricing Change

Save a competitor’s pricing page today, save it again in six months, and put the two in front of a model. You will get a list of differences. Most of them will not be pricing changes.

A pricing page is not a document. It is a rendered application that decides what to show you. It reads your location and picks a currency. It remembers which side of the annual toggle you left selected. It may be running an experiment. Whoever last rewrote the comparison table renamed six rows while they were in there.

Every one of those produces a difference between your two copies, and a model reporting differences will report them all. That is the correct behaviour. It was asked what changed and it answered honestly. The problem is that the output puts a renamed row and a new seat cap on the same footing, and the reader has no way to tell them apart.

Four things are worth writing down at the moment you save a pricing page, and none of them is on the page itself. The date. The country it was served to. The currency it displayed. Which side of the billing toggle was showing.

The middle two are the ones nobody records, and they are the ones that break a comparison. A page saved in the United States and the same page saved in Germany can show different numbers on the same afternoon. Put those two copies side by side six months apart and every price looks like it moved.

This is the same trap that catches a messaging comparison run across two vendors, where one page arrives in a different language and nothing in the pasted text says so. The price column is worse, because a number in the wrong currency still looks like a number you can subtract.

Three passes, and the second one runs before you read anything. Sorting the noise out first is what makes the rest of the output short enough to trust.

The rule about printing empty lists matters more than it looks. A run that returns no renames has almost certainly folded them into the real changes. An empty list is evidence the comparison was done; a missing list is evidence of nothing.

Four inputs, and two of them are lines you write rather than material you paste. The capture conditions take ten seconds and decide whether the price columns can be compared at all.

Paste the comparison table as well as the plan cards. The cards carry the prices, and the table carries the limits. A change to a limit is the kind that costs a buyer money without any figure on the page moving, and it never appears in a screenshot of three cards.

We took one CRM vendor’s live pricing page, pulled an archived copy from a year earlier, and ran the comparison twice. Both copies are public and both runs can be repeated.

Sixteen differences, one decision

The first run is accurate and nearly useless. Every line in it is really there in one copy and not the other. Read as a report, it says a vendor raised prices, dropped priority support, and reorganised their packaging. Two of those three did not happen.

The prices did not rise. The 2025 copy was served in dollars and the 2026 copy in euros, which is a fact about where each capture came from. Priority support was not dropped either. It moved off the plan card and into the comparison table, where it still sits against the Pro column.

Six more of the sixteen are a single row renamed, counted twice: once as something removed and once as something added. Email sequences became Sequences. Migration assistance became Migration service. SAML and SSO became Single sign-on. Nothing changed for any buyer in any of those.

The evidence that two figures really moved

One block of numbers is worth reading closely, because it shows how to get an answer out of a mixed-currency comparison. The page sells extra credits in four sizes. The two smallest carry identical figures in both copies. The two largest do not: 225 becomes 285, and 300 becomes 515.

If the page had simply swapped the currency symbol, all four would have kept their numerals. Two did. So the reading that fits is that the symbol changed for everything and the top of the ladder was repriced as well. That is a conclusion drawn from what stayed the same, which is the only lever these two captures give you.

The real change in that year sits in one cell. The middle tier used to say seats were unlimited. It now says up to ten. No price moved with it, no announcement went out, and it appears on the page as two words in a table row.

Work out what that does to a buyer. A team of twelve on the middle tier is now on the tier above it, or split across two workspaces. Nothing about the headline price predicts that, which is the whole reason a repackaging is the expensive thing to miss rather than a price rise.

The reason to keep the old copy is in that table. A change of this kind is only ever visible as a difference, and the page will not tell you it used to say something else. A pricing teardown that records what each tier contained on a date is how the second capture becomes readable at all.

Every comparison on this page is a comparison of two moments you happened to capture. The change itself happened on a day nobody recorded, somewhere between them. A quarterly diff tells you a seat cap appeared at some point in ninety days.

Being three months late on a repackaging has a specific cost. Your reps go on describing a competitor’s middle tier the way it used to be, in front of buyers who are reading the current page. The credibility goes before the argument does.

There is a limit no amount of watching removes. A published page is the list price, and most deals are not done at list. What a competitor actually discounts, and to whom, never appears on any copy of any page, and the only place it shows up is in your own lost deals.

Flares keeps a dated copy of every competitor pricing page and reads the tier contents rather than the headline figure. A seat cap appearing arrives as a notification with both versions attached, which is the older half of the diff you would otherwise have to have saved.