Competitor ad analysis: finding them is the easy part

Three layers, and only the last one changes anything

Competitor research has three layers. Most guides stop at the first two.

Layer 3 is the one that changes your account, and it is where the useful writing thins out. The ad-library material stops at layer 1 almost by definition - it is a guide to a tool. The analysis material stops at layer 2. The few people who do write about layer 3 write about doing it by hand, in a timeline, one ad at a time.

Which leaves the actual question untouched: how do you rebuild a structure often enough to matter, given that production is the constraint everyone in this space quietly assumes away?

A folder of screenshots feels like progress. It is not. It is layer 1.

What the ad library gives you - and what it withholds

Meta's Ad Library is free, needs no login, and covers Facebook, Instagram, Messenger, Threads, WhatsApp and the Audience Network. For any advertiser you can see the creative, the copy, the call to action, the landing page, the placements and the date each ad started.

What it does not give you is just as important.

**No end date.** Ads that stop running are removed rather than marked as stopped. The card tells you when an ad began; it never tells you when it finished. Which means the single most useful number in competitor research - how long an ad survived - cannot be computed from the interface at all. You only get it by taking a snapshot and coming back later.

**No variant count.** The interface says "this ad has multiple versions". It does not say how many. Two versions and six versions render as the same sentence.

**No arithmetic, no history.** Nothing sums, averages or compares across advertisers, and nothing tells you what changed since last week.

That is not a complaint about Meta. The library is a transparency tool, not an analytics product. But it explains why "just look at the ad library" is not a workflow - it is a place to start gathering.

Sort by run time, not by what looks good

This is the single most useful habit in competitor research, and it is the one people skip.

When you sort a competitor's ads by how long they have been live, oldest first, you are sorting by a signal you could not otherwise buy: **someone else's money**. Nobody keeps paying to run an ad for months out of sentiment. A long run means the ad is earning its place.

A short run tells you nothing. It might be three days old and dying, or three hours old and about to be a winner. Only duration separates the two.

So the first pass is mechanical:

• Pick three to five competitors your buyers actually compare you against.
• Resolve each one to its page ID rather than searching by brand name - keyword search matches the text of the ad, not the advertiser, so searching a brand name returns everyone who merely mentions it.
• Sort by run time and write down the handful that have been live longest.
• Do it weekly rather than once. A single snapshot is a photo; a series is a trend.

What survival time actually tells you

Duration alone is a number. Duration compared across competitors is a strategy.

A published study sampled four meal-kit brands in August 2026 and computed what the ad library interface cannot:

Read those two middle rows against each other and you get two completely different businesses. Hungryroot has a 91-day median and a live ad from eleven months ago: it found winners and is scaling spend behind them. HelloFresh has a 13-day median across 120 ads: it is in permanent test mode, writing, launching and killing on a two-week cycle.

Those imply opposite responses if you compete with them. You will not out-test HelloFresh cheaply - that takes a production budget. But you might out-last Hungryroot by finding a fresher angle it has not tried.

Blue Apron is the cautionary row: 35 ads, none of them active.

Variant depth shows where the budget went

The library groups near-identical creative into a collation - several versions of one concept. The count is a readout of how disciplined the advertiser is.

In the same sample, Hungryroot ran 77% multi-variant creative, topping out at six versions of a single concept. HelloFresh ran 2%.

That gap is not a style preference. Holding the idea constant and varying the execution is how you find out which execution works - the idea gets a fair test instead of being discarded alongside a bad hook. One ad per idea means every failure is ambiguous: you never learn whether the concept was wrong or the hook was.

Worth checking on your own account too. If your variant depth is low, that is usually a production constraint showing up as a testing strategy.

Turning a finding into a brief

The output of good competitor research is not a screenshot folder. It is a brief that someone can produce from. The difference is structure.

When you save an ad worth studying, capture the same fields every time, in the same vocabulary, so that ads can be compared rather than merely collected:

That last group is what turns a library into a signal. A single ad tells you nothing about strategy; the same ad tracked across four weeks tells you whether the advertiser is scaling it or replacing it.

Then the brief writes itself: "this structure has been live for five months for a brand in our category, it leads with a conceded downside and pays it off with a comparison, and it has run six variants of the concept. Build me our version."

• **Format and placement** - what it is and where it runs.
• **The hook, rewritten in one or two lines** - so you can search it later and see when a competitor reuses it.
• **The visible elements** - UGC, founder face, product demo, text overlay, before and after, testimonial.
• **The offer**, if there is one - what kind, and how strong.
• **The proof mechanism** - ratings, quantified outcomes, guarantees, awards.
• **The call to action**, verbatim.
• **The dates** - first seen, still live, and how fast fresh cuts of the same idea appear.

Four ways this analysis goes wrong

**Mistaking a long run for a good ad.** Duration proves the ad is being paid for, not that it is any good. A mediocre ad in an uncontested audience can run for months. Read duration as "this is working for them", then check whether their situation resembles yours.

**Reading one snapshot as a trend.** A single capture tells you what is live today. Whether a competitor is scaling video, flooding discount creative, or quietly pulling back only shows up across weeks. One-off digs do not tell a story.

**Counting ads instead of concepts.** A brand running forty ads that are all the same concept with different text has one creative strategy, not forty. The number that matters is how many distinct arguments are live.

**Assuming their winners will be yours.** A structure that works at their price point, with their proof and their audience, is a hypothesis for you - not a result. It becomes a result when you test your version, which is the whole reason layer three exists.

The mechanics, and the one mistake everybody makes

**Do not search your competitor's brand name.** Keyword search matches the text of the ad, not the advertiser. Searching a well-known sportswear brand in the US library returns an anime league, a phone-widget app and a musician alongside the actual brand, because each of those mentions it in the body copy. The same search for a meal-kit brand returns a retailer and a cookware company.

The fix is to resolve each competitor to its numeric Page ID once, and query by that from then on. It takes a few minutes per competitor and it is the difference between a competitor library and a pile of ads that merely mention your competitor.

**Filter by country.** Ad libraries are not global. Mixing markets means comparing creative that was written for different audiences, in different languages, at different price points. Pick one market per pass.

**Snapshot on a cadence, not once.** Because stopped ads are removed rather than marked, a single capture cannot tell you how long anything lasted. A daily or weekly snapshot is the only way to build the history the interface declines to keep. The absolute cadence matters less than the consistency - the first useful readout arrives after about a month.

**What you get when you do this.** For each competitor: which ads are still live, how long each has run, how many versions of each concept exist, and how fast new cuts of the same idea appear. Those four numbers are not available anywhere in the interface, and together they are a decent picture of how a competitor spends on creative.

**What you still do not get.** Anything about performance. You cannot see their CTR, their CPA or their ROAS, and no external tool can show you those. Duration is a proxy for "this is working" - a good one, but a proxy. Treat every conclusion as a hypothesis to test on your own account, not as a fact about theirs.

What a weekly workflow looks like

Competitor research fails when it is a project. It works when it is a small recurring habit - the value is in the series, not in any single session.

**Once, at the start.** Resolve your three to five competitors to Page IDs. Build the capture template: the fields from the section above, in the same order every time. This is a one-off hour.

**Weekly, 30 to 45 minutes.**

**Monthly, an hour.** Step back and read the direction rather than the snapshots: is a competitor shifting toward video, flooding discount creative, or pulling back from a placement they used to dominate? That is the layer where arcs become visible.

The discipline that makes this work is writing one brief instead of collecting forty screenshots. A file of images produces no decisions; a single specified structure produces an ad.

• Pull the current library for each competitor.
• Diff against last week: what is new, what disappeared, what is still running.
• Update survival and variant counts for anything already in the file.
• Flag anything that has crossed into "long-running" - those are the ones worth a brief.
• Write one brief, not five. One structure, properly specified.

The gap is layer three

Collecting is solved. The ad library is free and good. Analysing is solvable - sort by run time, compute survival, count variants, and you know more about a competitor's creative strategy than most of their own agencies will tell you.

Then you try to act on it, and the process stops.

*"I pulled up the Meta Ad Library, saved some screenshots, and dropped them into a deck. It looked neat. But when I had to decide which creative to test next, those slides had nothing to offer me."*

That is a practitioner describing their own competitor-research workflow, and it is the most honest sentence in this entire topic. The slides were fine. The problem was what came next: to act on the finding you need to produce the ad, and producing one more ad is exactly the thing that is expensive.

So the analysis is not where the value leaks. **The value leaks between "I know what to make" and "I have made it".**

The line: take the template, not the ad

This is the part that makes the whole practice legitimate, and it is worth being strict about.

Every ad is two things stacked. There is the **skin**: their product, their photography, their model, their logo, their brand colours. And there is the **template**: the layout, the structure, the rhythm of the copy, where the headline sits, the kind of claim it leads with.

The skin is theirs. Copying it is both a legal problem and a pointless one, because it still sells their product. The template is a format, not an asset - "open by conceding a downside, then turn it into the reason to buy" is not anyone's property.

**The 15/85 rule.** Take roughly 15% as inspiration - the structural move, the format, the shape of the argument. Make the other 85% genuinely yours: your product, your evidence, your voice.

The failure mode is the inverse, and it is common. Brands lift a proven ad wholesale - same premise, same rhythm, sometimes near-identical wording - and swap in a different category. It reads as a copy, and worse, the borrowed claim does not hold up, because they took the format without earning the argument.

So when you find an ad that has been running for eight months, what you take is the move it makes. Then you rebuild that move around your product, and you have to be able to defend the claim in it.

How BrandU closes layer three

BrandU takes the layer-three problem directly: it makes turning a competitor's structure into your own ad cost a fraction of producing it from scratch.

**Start from the ad that is working.** Paste a link to it - a page URL is fine, it does not need to be a direct video file. BrandU watches the whole thing frame by frame, works out what makes it work, and rebuilds that structure with your product, your claims and your voice. What comes back is not their footage. It is the argument, rebuilt for you - which is exactly the 15/85 split, done in one step rather than by hand in a timeline.

**A video is generated as parts, so a change costs the change.** Independent segments mean:

**Variant depth becomes affordable.** That study found one brand running six versions of a single concept and another running one. Holding the idea and varying the execution is the discipline; the reason more brands do not do it is that each version looks like a full production run. When a variant costs the render rather than the rebuild, running three hooks against one concept stops being a budget decision. A run generates up to 10 variants at a time, and you can run it again for the next 10 - the cost does not climb as you go.

• Rewriting a segment's line re-runs that segment's model and its voiceover; the other segments are kept.
• Restyling, reordering or swapping an asset only re-renders - the model is not called again.
• Change the hook, a line, the language or a component and the timeline and related elements re-time automatically.

What we do not do

If what you want is continuous competitor monitoring - alerts, spend estimates, ad history across dozens of brands - tools exist for that, and BrandU is not one of them. We do not track anyone's ads and we do not build swipe files for you.

We do one thing: take one ad, and turn it into your version, with variants. If your problem is "I have found the ad and I need to ship mine", that is the problem we solve. If your problem is "I need to watch forty competitors every week", you want a monitoring tool and we would rather tell you than waste your trial.

What BrandU does not do

• It does not find competitor ads. That is the ad library's job, and it does it well.
• It does not tell you which competitor structure will work for you. It produces the version; the auction decides whether it wins.
• It does not remove the need to check the claim. If a competitor's ad argues something your product cannot back up, the rebuild will still be wrong - the 15/85 rule is about more than wording.
• It does not measure results. Producing variants is not measuring them.