Behind the sentence "we did a competitor analysis" there is usually this: someone spent an afternoon looking at competitor ads, screenshotted the three they liked, and pasted them into a deck.
That is not an analysis. It is an impression. And impressions have one problem: you do not remember what you saw or when, so you cannot tell what changed.
A single list says nothing
You learned that a competitor is running 40 ads today. On its own, what does that number mean?
Nothing. Is 40 a lot or a little? Was it 12 last month, or 80? You do not know. The same goes for creative: "they use a red background" is an observation; "they used red for three months and switched everything to blue two weeks ago" is a finding.
The difference: the first is a state, the second is a decision. And your competitor's decisions are the only real thing you can learn from them.
The ad library does not tell you what your competitor thinks. It tells you what they do. But measure the same thing twice and it tells you what they changed — and a change is always the result of a thought.
Four signals you can actually measure
The ad library carries no spend and no performance data. But these four, tracked over time, stand in for performance:
| Signal | How to read it | What it tells you |
|---|---|---|
| Lifespan | Days between launch date and today | A long-lived ad is probably winning |
| Version count | "This ad has N versions" | A high number means active testing |
| In / out | What was added and dropped between two scans | Where the budget is moving |
| CTA mix | Ratio of Shop Now to Learn More | Which end of the funnel they are loading |
The fourth is the least used. If a brand's CTA mix leans on Learn More they are building an audience; if it shifts toward Shop Now they have moved into harvest mode. That shift is usually the signature of a campaign calendar.
The method: two dates, one difference
What turns competitor research into measurement is simple and boring.
- Fix the list. 5–10 competitor pages. The list has to stay identical every round or the comparison breaks.
- Fix the country. Same country, every round. Change the country and you change what you are measuring.
- Scan completely. Go to the end of the infinite scroll. A half list does not give you half the data — it gives you wrong data.
- Save it. Ad id, launch date, first line of copy, CTA. Not screenshots — a table.
- Wait two weeks. Repeat.
- Take the difference. Split it into three sets: survivors, dropped, new.
Step six carries all the value:
- Survivors are your competitor's winning creative. The pattern worth copying is here.
- Dropped are the ones that did not work. What you learn from them is that they are not worth trying.
- New are your competitor's next bet. This is where you first see where the category is heading.
Who counts as a competitor
Most teams build the competitor list wrong: they write down the competitors already in their heads. In the ad market, your competitor is anyone showing ads to the same person at the same time — a competitor for the audience, not for the industry.
A better way to build the list is to start from keywords: search the offer phrases your customers use ("free trial", "money-back guarantee") and collect the advertisers that come back. Brands you never expected will show up — and they are usually the ones you learn most from.
One warning: copying is not analysis
The most common bad outcome of competitor research is a direct imitation of the winning ad. It fails for two reasons:
- You cannot see the context. The offer, the price, the landing page and the audience behind the ad are invisible. The same creative behaves differently bolted onto a different funnel.
- Being second is expensive. You are the second person showing that message to that audience; the first already took the memory.
What transfers is not the creative but the pattern: what the first line promises in this category, where the objection is handled, which CTA appears at which stage. Patterns travel. Creative does not.
How often
Two weeks is a good starting point. Looking more often produces noise — most campaigns do not change meaningfully in under two weeks. Monthly is too late: you miss the middle of a two-month campaign.
The point is not the frequency but the regularity. Irregular measurements cannot be compared, and a measurement that cannot be compared is just an impression again.
To do this with the extension: Competitor creative research · Shopify dropshipping product and creative research