Ad Creative Analysis: Tag Every Ad on the Same Fields, Then Compare
Ad creative analysis for TikTok: tag each ad's hook, format, proof and CTA, compare them against view rate and CTR, then brief the next test.
Ad creative analysis means you break every ad into the same parts, such as the hook, the format, the proof and the call to action, and then check which of those parts turn up in the ads with the best results. Most analysis goes wrong at the first step. A strategist watches a winning ad, writes "strong hook, feels authentic" and moves on to the next one. Notes like that can't be compared with anything. Tag every ad on the same fields, include ads that did worse, and let the patterns come out of the comparison.
Start with ads that have results attached
Without performance data you only learn what a brand made, with no way to judge whether it worked. On TikTok, the Creative Center's Top Ads dashboard gives you ads with a performance signal next to them. According to TikTok's help article on the dashboard, you can:
- filter by region, industry, campaign objective, likes and time frame, plus ad language and ad format on desktop
- sort by reach, click-through rate (CTR), 2-second view rate or 6-second view rate, and on desktop also by conversion rate and likes
- click "See Analytics" on an ad to open a second-by-second graph of its relative performance
Log in before you start, because the same article says logged-out visitors get only five ads. TikTok also says every ad in Top Ads was authorized by its advertiser inside Ads Manager, so the collection holds ads that brands agreed to share.
Pull 15 to 30 ads for a single category, objective and region. Sort once by 6-second view rate and once by CTR, and save some ads from further down each list as well. Add your own brand's past ads to the same sheet, with the results from your ad account next to them.
The tagging sheet
Give each ad a row and fill every column for every ad. Keep the values to short labels from a fixed list, so you can filter and count them later.
| Column | What to write | Example labels |
|---|---|---|
| Hook type | The kind of opening in the first 3 seconds | product first, question, bold claim, before and after, problem |
| First line | The first spoken sentence, word for word, and the second it ends | |
| Format | How the ad was filmed | talking head, POV, skit, demo, unboxing, listicle |
| Product appears | The second the product is first shown or named | 0:01, 0:06 |
| Proof | The thing on screen backing the main claim | demo, result shown, review read out, none |
| CTA | The wording, and whether it is spoken, on screen or both | on screen only |
| Length | Seconds | 24 |
| Signal | The sort you pulled it from, or your own results | top 10 by CTR |
Write the first line out in full instead of summarizing it. Two ads tagged "question" can open with very different questions, and a creator will need the actual wording later when you brief them.
Read the drop-off graph against the words
The second-by-second graph in See Analytics marks where an ad performs better or worse from one second to the next. To work out why, put the graph next to a timed transcript and note what was said at each point where the line drops. A drop at 0:04 means one thing when the product appeared at 0:03 and something else when the speaker was still describing the problem.
You can type timestamps by hand while scrubbing through the video. When the ad also runs as a TikTok, Reel or Short and you have its link, UGC Script gives you every spoken word with timestamps and splits the video into timed beats with a format label, which fills the first line, format and product columns for that row.
Compare the top of the list with the rest
Once the sheet has 20 or more rows, filter it a column at a time. For each label, count how often it appears among the ads at the top of your sort and among the others. A hook type that shows up in most of the high-CTR ads and rarely further down is worth testing. When a hook type is common across the whole sheet, the category simply uses it a lot.
Match each metric to the columns it can explain. A 2-second view rate depends only on the first 2 seconds of the ad, so compare it against hook type and first line. A 6-second view rate also covers the setup that follows. CTR and conversion rate come at the end of the ad, so compare those against proof, offer and CTA.
Likes are one of the dashboard's filters, but a like is a reaction, and a viewer can like an ad without clicking it. When the campaign objective is sales, sort by CTR or conversion rate.
Turn a pattern into a brief
Write each finding as a test for the next round and change a single element per test, for example "product shown in the first second, against our current problem-first opening, same body". If the opening and the CTA change together, you have no way to tell which one moved the result.
Give the creator the structure and leave the wording to them, since the lines in a reference ad belong to the creator who filmed it. Put the hook type, the beat timings and the proof you need into the brief, and add the reference link so they can watch the pacing. The UGC brief template has a field for each of these, and the TikTok ad script template lays out several hooks over a single body, which suits a hook test.
Keep the sheet going
Add rows every time you pull ads, with the date and the sort you used. After a few rounds of testing, your own results will fill the signal column for more rows than the dashboard does. Weight those rows most, because they come from your audience and your offer.
Try it on a video
Paste up to 50 links at once, or upload your own raw footage. Free, 5 videos a day, no account.