I rebuilt an entire ad set around a competitor's hook once. Their line had been running since March, the copy was tight, and my version read well in the doc. It lost money for two weeks straight.

Then I read past the hook. Ninety-day returns, shipping free both ways, a price about 20% under ours. The hook was never the thing doing the work. It was the trailer for a deal we were not running.

That is the swipe file's quiet failure mode. You copy the message and inherit none of the terms underneath it.

Short answer: Analyzing a competitor's offer means reading their ads for commercial terms rather than for angles. Pull the price, the guarantee length, the free-shipping threshold, the trial, the financing and the bundle shape out of the ad copy and the on-image text, then compare those terms across the whole competitor set in one table.

The takeaways

Most of the deal is stated in the ad itself. Price points, guarantee lengths and shipping thresholds sit in the primary text or burned into the creative, so you can compare ten competitors without opening a landing page.
An ad running past 30 days is survival evidence for the whole package. Someone can sustain that price at that guarantee at that margin. It says nothing about whether you can.
A terms table beats a folder of screenshots. One row per competitor, with price, guarantee, threshold, trial and destination, shows which lane is crowded and which term nobody has claimed.

What counts as an offer in a competitor's ad?

An offer is the bargain in numbers: what the buyer gives up and what they get back. The angle is how that bargain gets dressed. "Stop overpaying to heat your house" is an angle. "First service €49, cancel any time, 12-month parts warranty" is an offer.

In ad copy the offer shows up as a small, finite set of terms. Price point or price band. Discount depth, and whether it needs a code. Guarantee length and what triggers it. Free-shipping threshold. Trial length, and whether a card is required up front. Financing, plus the monthly figure it reduces to. Bundle shape, meaning how many units at what per-unit price. Delivery promise.

Short enough to be a spreadsheet header. Also the part you cannot swipe before checking your own margin. If you are unsure whether the deal or the creative failed in your own account, the spread between your angles answers that first: is it my offer or my ads.

What can you pull out of an ad library listing?

Three things, reliably. The offer as stated in the copy and in the text burned into the image, which is where most e-commerce deals sit. The destination the ad points at, which tells you whether this is a site-wide promotion or a single product push. And the date the ad started running, stamped on every listing, which tells you how long that exact set of terms has been live.

What you will not get is anything past the click. The library holds ads. It does not hold the landing page, the checkout, the upsell path, or the version of the offer that appears after a first-visit popup fires.

So treat the library as the first half. Keep the second half manual: open the destination yourself for the three competitors who matter, and note where the stated deal and the landed deal disagree.

How do you turn 40 saved ads into something you can act on?

Build a terms table. One row per competitor per distinct offer, columns for the terms above, plus first-seen date and destination. Ten competitors takes an afternoon. Then read it column by column.

That is where it pays. A column where everyone sits at 30-day returns tells you 30 days is table stakes here and buys you nothing. A column where free-shipping thresholds cluster at €50 with one outlier at €25 shows what the aggressive end of your market costs. An empty column is the interesting one: nobody offering financing in a category with a €400 average order is either an open lane or a bad idea somebody already dropped.

A screenshot folder answers none of this. Screenshots are organised by ad, and every question worth asking here is organised by term.

Does a long-running offer mean the offer works?

It means somebody can sustain it. An ad still running after 30 days or more has cleared a bar most ads fail, and because the ad carries the terms, that durability covers the whole package: the price, the guarantee, and whatever margin sits behind them. Useful evidence. It is also the same signal you use to find a competitor's winning ads.

The part people skip is whose economics it proves. A competitor with better cost of goods, a longer customer lifetime, or investor money to burn can keep a 40% discount alive for a year while the same discount puts you under water in a quarter.

Copy a price and you have copied someone else's unit economics without ever seeing the spreadsheet behind it. That is a bigger bet than copying a headline, and it gets made with less thought.

What will the ad libraries never show you?

No spend, no ROAS, no margin, no return rate. That is the honest limit, worth saying plainly, because a lot of competitor-research advice implies more than it delivers. You can see that an offer exists and that it persisted. You cannot see what it cost or what it brought back.

So give the terms table two jobs. First, a constraint check: before you swipe a message, confirm the deal underneath it is one you could run at your own margin. Second, a gap finder: the term nobody in your set offers is a hypothesis worth a small test with your own numbers attached.

Everything past those two jobs is guesswork with better formatting.

Where does tooling help here?

Adscalr pulls three ad libraries (Meta, TikTok, Google Ads Transparency) into one dataset, updated daily, and vision AI decodes each static ad into roughly 20 structured fields plus design, copy and strategy scores from 1 to 10. Winner detection flags ads still running past 30 days. The market map clusters competitors on five axes (Price, Quality, Speed, Trust, Innovation), and those read as offer dimensions as easily as messaging lanes, since Price is a terms question before it is a positioning question.

What it does not do is worth stating just as clearly: it reads ads. Landing pages, checkouts and anybody's cost base stay outside its view, so the margin call stays yours. More on that library side in competitor intelligence.

How to analyze competitor offers from ads