The short answer
A useful ad swipe file has three things a screenshot folder doesn't: an entry rule, tags and a one-line note per ad. Tag three dimensions (hook pattern, angle, awareness) and patterns appear: on our board, one combination of the three reaches the top tier 29% of the time, against about 10% overall.
Why most swipe files die
Almost every media buyer has one: a folder of 400 screenshots, a Notion page with links that no longer work, a Slack channel called #inspo. Nobody opens it when writing a brief, because finding the right ad takes longer than writing from scratch.
A swipe file stays useful when it answers a question in under a minute. "Show me first-person story hooks for problem-aware buyers in pet." That needs three things the screenshot folder doesn't have:
- An entry rule, so only ads with evidence get in.
- Tags, so you can filter.
- A note, so future you knows why it was saved.
Rule 1: only ads that proved something
The fastest way to ruin a swipe file is to save ads because they look good. Pretty ads that ran for four days teach you nothing about what sells.
Set a bar and stick to it. A workable one: the ad has been live at least 30 days, and either has multiple versions or a solid spend estimate. Every ad on the Ad Radar board clears $50k in estimated spend and 30 days live, and the median one has run 120 days. The full set of signals is in how to find winning Facebook ads.
Save your own ads too, winners and losers, with results. Your losers are the only ads in the file where you know the real numbers.
Rule 2: tag with a fixed vocabulary
Free-text tags drift. One person writes "UGC", another "creator", a third "testimonial video", and the filter breaks. Use a fixed list of values per dimension.
Ad Radar tags every ad on ten dimensions. You can copy the system, or use a subset. Here is how often the most common value of each dimension shows up across the 2,443 ads on our board (October 2026):
| Dimension | What it captures | Most common value | Share of all 2,443 ads |
|---|---|---|---|
| Hook pattern | The shape of the first line | Problem/Benefit Opener | 29% |
| Hook mechanism | Why the first line works | Problem-to-Solution Bridge | 64% |
| Angle | The argument of the ad | Root Cause | 34% |
| Awareness | What the viewer already knows | Unaware | 34% |
| Proof level | How much evidence it shows | Medium | 60% |
| Production format | How it was made | Footage + Voiceover | 7% |
| Video type | The video's structure | Demo / Mechanism | 5% |
| Landing page type | Where the click goes | Product page | 36% |
| Niche | The problem or category | Pet | 13% |
| CTA type | Buy or lead | Buy | 95% |
Two lessons from the table. First, some dimensions are dominated by one value (CTA, hook mechanism), so they filter less. Second, the most useful dimensions for writing are hook pattern, angle and awareness, because they have several common values with real differences between them. If you only tag three things, tag those. The awareness scale comes from Eugene Schwartz; we explain it in awareness levels for ads, and the difference between the two hook dimensions in hook pattern vs hook mechanism.
Why three tags beat thirty screenshots
Tags turn a pile into a dataset. Combine just three dimensions (hook pattern, angle and awareness) and our board splits into 190 combinations. Most are small. A few are big, and they don't perform the same:
| Hook pattern + angle + awareness | Ads | Reach the top tier | Median estimated spend |
|---|---|---|---|
| Problem/Benefit + Root Cause + Unaware | 115 | 7% | $216K |
| First-Person Story + Root Cause + Unaware | 88 | 8% | $222K |
| Problem/Benefit + Root Cause + Problem-Aware | 67 | 15% | $316K |
| First-Person Story + Story/Case Study + Unaware | 60 | 13% | $228K |
| Specific Number + Systematic Breakdown + Solution-Aware | 48 | 29% | $227K |
About 10% of the board reaches the top tier. The specific-number, systematic-breakdown, solution-aware combination reaches it almost three times as often (48 ads, so treat it as a strong hint rather than a law). Without tags, you would never see that. With tags, it is one filter.
Here is one ad from that combination:
Yeah, so I think I need a refund on your water fountain.
- Est. spend
- $476K
- Days live
- 68
- Format
- Video, 59s
- Variants
- 2
Opens like a complaint, then breaks down point by point why the fountain costs more than a cheap one. The buyer already knows fountains exist; the ad explains why this one.
Rule 3: write a note before you save
Every ad gets three lines of text. This is the part people skip, and it is the part that makes the file reusable.
- Hook, verbatim. For video, transcribe the first sentence. The library won't search spoken words, so if you don't write it down, it is lost.
- Why it works, in your words. One sentence. "Names the smell nobody mentions" is useful. "Great hook" is not.
- What you would steal. The structure, not the words. "Confession of a result, then the mechanism" is a brief. The competitor's sentence is not.
The peeing finally stopped, and all I did was plug this in.
- Est. spend
- $716K
- Days live
- 523
- Format
- Video, 70s
Example note: "Opens on the result, names the embarrassing problem plainly, product is a plug-in so effort is near zero. Steal: result-first opener plus a low-effort fix."
Rule 4: boards by use, not by brand
Organizing by competitor feels natural and is rarely how you search. Organize boards by what you will do with them:
- Hooks to test this month. Ten openers, rewritten for your product.
- Angles by awareness level. One board per level, for briefing different funnel stages.
- Offers and promos. Saved by season, so Q4 planning starts with real examples. See Black Friday ads for 2026.
- Landing pages. The page behind each winning ad, because ads and pages win together.
- Client boards. If you run an agency, one board per client that you can share.
This is Riley.
- Est. spend
- $1.5M
- Days live
- 109
- Format
- Video, 58s
Three words, a dog's name, and the story starts. A creator-style Page selling a GPS dog collar. It belongs on a "story openers" board, not a "competitor X" board.
Rule 5: a review cadence
| When | What | Time |
|---|---|---|
| Weekly | Add new ads that cleared your entry rule, with tags and notes | 20 minutes |
| Monthly | Remove ads that stopped early or never proved anything; promote the best ten to a "use now" board | 30 minutes |
| Quarterly | Count tags: which hook patterns and angles dominate the ads that ran 90+ days in your niche | 1 hour |
The quarterly count is where the file pays for itself. It tells you which combinations your market rewards, and which ones nobody is using yet.
A starter template
Columns for a spreadsheet version:
Brand | Page | Library ID | Date saved | Start date | Days live | Variants | Est. spend | Format | Length | Hook (verbatim) | Hook pattern | Angle | Awareness | Proof | Landing type | Offer | Why it works | What to steal | Board
If you outgrow the spreadsheet, Ad Radar comes with the ten tags already applied, transcripts searchable, and boards with share links and CSV export. Our comparison of swipe file tools covers the other options, and the weekly research process shows how the file fits into a routine.
Figures marked as estimated spend come from Ad Radar's model of engagement on public Meta Ad Library ads. They are estimates, labeled as such, and are best used to rank ads against each other.