How to Find Which Competitor Ads Are Actually Working

Finding which competitor ads are actually working is usually attempted by sorting for the ones that have been running longest, and that method is close to worthless. Duration costs an advertiser nothing. An ad sitting in an always-on campaign can run untouched for half a year because reviewing it is nobody's assigned job, and it will appear at the top of a longevity sort looking exactly like a proven winner. Last updated: September 2026.

Omniconvert brings CROBenchmark-scale evidence to this, drawn from 70,000+ experiments and 13 years in eCommerce. The lesson that transfers most directly from testing to competitor research is about cost asymmetry: what an organisation spends money on is a far better read of what it believes than what it merely allows to continue.

This article is about the individual creative. The four-layer survey of competitor research sits in the DTC competitor research stack, and the method for reading a creative's construction is in how to Reverse-Engineer competitor ads with the Meta ad library.

Why longevity is the weakest evidence in competitor ads research

Think about what you are actually observing when you see a long-running ad. You are observing an absence: nobody stopped it. Absences have many causes, and a strong performance report is only one of them.

The others are mundane and common. The person who owned the campaign left. The account is managed by an agency paid to maintain rather than to prune. The ad sits in an evergreen set that is reviewed quarterly at best. The spend on it is small enough that nobody notices it in a report.

Now consider what you observe when the same idea appears in a new execution. Somebody wrote a brief, commissioned or shot new material, edited it, approved it and put it live. That is a chain of deliberate acts, each costing time and money, and none of it happens by inertia.

This is the whole argument of this article compressed into one line: read what they built, not what they left alone.

The re-paste test

1. Group by idea, not by asset. Most competitor research counts assets, which inflates everything. Fifteen active ads frequently turn out to be three ideas in five aspect ratios each. Write down the concepts first, in your own words, then attach the assets to them.
2. Count the executions per idea. One idea with six genuinely different executions is a far stronger signal than six unrelated ads each running once. The count is the measurement.
3. Look for re-production. New footage, a new setting, a different presenter format, a rebuilt edit. Recycling an old asset into a new crop is not re-production. Reshooting is.
4. Check whether it spread. An idea that started in one placement and now appears in others has been adapted, and adaptation is another deliberate spend. Spread is the confirmation that the count on its own can lack.

An idea that passes all four is one the advertiser has invested in repeatedly. That is not proof it works, because organisations make mistakes and persist with them, but it is the strongest evidence available to somebody standing outside the account.

The four false positives

Always-on evergreen. Brand or category ads that exist to keep a presence rather than to perform against a target. They run for years, they are reviewed rarely, and their persistence is a policy rather than a result.

Retargeting creative. This is the most consequential mistake on the list, because retargeting ads reach people who already know the brand and therefore perform well on metrics that have nothing to do with the creative. Copying the approach for cold prospecting reproduces the execution without the warm audience that made it work.

Brand-safety filler. Inoffensive creative that exists to keep a campaign populated, often around a seasonal peak or a sensitive period. It is designed not to fail rather than to succeed.

Tests still in flight. A test that has been running three weeks has not yet been judged by the advertiser either. Reading it as a decision imports a conclusion nobody has reached.

Nielsen's long-running work on advertising trust has consistently found that recommendations from other people outrank brand-controlled messages, which is worth holding alongside all four. The competitor creative that appears to perform best is often the one carrying a customer voice, and what transfers from it is that structure rather than its script.

What each signal is worth

The bottom row deserves its position. Engagement is the most visible number in any ad library and the least connected to commercial outcome, and a creative optimised for comments is frequently a creative that entertained people who were never going to buy.

Isolating the element rather than copying the ad

This is the step that turns observation into something usable. Lay the versions of one concept side by side and mark what is identical in every single one. It might be an opening claim, a specific objection handled in the first few seconds, a format such as a side-by-side comparison, or an offer structure.

Then mark what changes between versions. Different presenters, different settings, different music, different colours. Those are the parts the advertiser is comfortable altering, which means they do not believe those parts carry the result.

The constant is the transferable finding. You cannot use their footage and you should not use their script, but the claim underneath it, or the structure of the argument, is a hypothesis you can test on your own audience with your own assets.

The Meta Ad Library makes this comparison practical because it shows active variants side by side, which is precisely the view a single-ad analysis throws away.

What this method cannot tell you

Nothing visible from outside touches margin, discounting, return rates or the cost of the sale. A concept can drive enormous volume at a price that destroys the unit economics, and from the library it looks like a triumph.

Nor does it account for how differently the same argument performs against a different audience. An advertiser with years of brand awareness can open an ad in a way a newer store cannot, because the viewer already knows who is speaking.

So use the finding for what it is. A hypothesis backed by somebody else's spending decision is a better place to start than an internal brainstorm, and it still has to be tested on your own traffic before it means anything. Where your store sits against competitors across creative, reviews, AI visibility, agentic commerce, competitor synthesis and CRO is what a free leaderboard score answers, and the paid audit report goes dimension by dimension.

What to do this month

• Choose five, not twenty. Direct rivals in your category and price band. Twenty produces a spreadsheet; five produces a finding.
• Write the ideas down in your own words. Before counting anything. If you cannot state the idea in a sentence, you are looking at an asset rather than a concept.
• Mark the re-produced ones. New material, not new crops. These are your candidates and there will be fewer than you expect.
• Extract the constant. Compare versions, find what never changes, write that as a claim rather than as a description of the ad.
• Test one claim, properly. Your own assets, your own audience, measured against a control. Where the constraint is that ad, order and margin data sit in systems that never meet, Nexus by Omniconvert is an AI for eCommerce growth engine that unifies commerce data, ranks experiments by True Profit and generates campaigns you approve before they go live.

FAQ: finding which competitor ads are working

Does a long-running competitor ad mean it is working?

Not on its own. Duration is cheap: an ad can run for months because it sits in an always-on campaign nobody reviews, or because switching it off is nobody's assigned job. What costs money is rebuilding the idea, which is why a creative that has been reshot, recut or re-versioned tells you far more than one that has simply never stopped.

What is the strongest signal that a creative is performing?

Deliberate reinvestment in the same idea. Several distinct executions of one concept, produced at different times, adapted into new formats and placements, means somebody decided to spend production budget on it more than once. That decision is the closest thing to a performance report you can observe from outside.

What are the false positives to watch for?

Four of them. Always-on evergreen that nobody reviews, retargeting creative that looks like prospecting but reaches a warm audience, brand-safety filler that exists to keep a campaign populated, and tests still in flight that have not yet been judged. Each can run for a long time and none is evidence of anything.

Should I copy a competitor ad that appears to be working?

Copy the claim, never the execution. What a successful ad proves is that a particular argument moves that audience, and the argument is transferable. The execution is tied to their brand, their offer, their price point and their audience history, so a replica typically underperforms the original and teaches you nothing you can build on.

How do I tell which element of an ad is carrying it?

Compare the versions against each other. Where an advertiser has produced several executions of one idea, whatever stays constant across all of them is what they believe is working, and whatever changes freely is what they consider incidental. That comparison isolates hook, offer or format far more reliably than studying any single ad closely.

How often should I run this check?

Monthly is enough, and more often produces noise. Re-production cycles run in weeks rather than days, so a weekly check mostly records the same state repeatedly. What matters is that the check is logged each time, because the signal is the change between readings rather than anything visible in one.