First, read watch time beside CTR
Open the ad-level report for the ad group and add the video watch columns: average watch time, the share of views that get past the opening and the share that reach the end, alongside impressions, clicks and the click-through rate (CTRShare of impressions that result in a click.). Compare each ad only with its siblings in the same ad group and with the account's earlier ads for the same objective, because tap rates differ widely by objective and market.
Then sort the weak ads into three patterns. Early exit: few views pass the opening, so the tap rate is low because almost nobody stayed. Watched but untapped: views pass the opening and many reach the end, yet few tap. Decay: the ad started near its siblings and has drifted down over the weeks while FrequencyAverage times a person saw an ad. rose. Each pattern points at different rows of the table. The content formats and hooks guide explains why the opening matters so much on this feed and holds the catalogue of hook patterns; this page uses the numbers to say which layer failed, and leaves taste out of it.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Few views pass the opening; the first frame is a logo, title or slow shot | First frame and first spoken line against the ad's siblings | Opening seconds that read as an advert | Re-cut so the first frame shows the result or the problem |
| Good watch time, few taps; premise or offer text near the bottom or right edge | The ad on a phone in the feed, with the interface showing | Captions or key text under the interface | Move text into the clear area of the frame |
| Viewers leave once the music starts or the voice drops; captions carry everything | The edit played with sound on, the way the feed is built to play | Edit built for sound-off on a sound-on feed | Lead with voice or native sound; keep captions as support |
| Rate falling week on week; frequency climbing | Ads per ad group; weeks each ad has run | Too few creatives in rotation, so fatigue sets in | Add genuinely different ads to the same ad group |
| Watched to the end, few taps; ad runs under a custom name and avatar | Identity setting against Spark versions of similar content | Custom identity instead of a real account | Test Spark posts from the brand or a creator |
| Watched to the end, few taps; the closing line is a generic prompt | Last spoken line and on-screen text against the button and page | Call to action that does not say what the page offers | Name what the page holds, out loud and on screen |
Fix, verify and prevent each cause
An opening that announces an advert
Early exit is the most common pattern. A brand logo, an intro animation, a greeting or a slow establishing shot tells the viewer this is an advert before any reason to stay has appeared, and the thumb moves on.
Fix: re-cut the video so the first frame shows the result, the problem or the product in use, and the first spoken line states the point, using the layers the content formats and hooks guide describes.
Verify: the share of views passing the opening rises against the old cut in the same ad group, and taps follow.
Prevent: review the first frame and first line of every ad before upload, as a separate check from the rest of the edit.
Words the interface covers
TikTok documents that its interface overlays parts of the screen, the caption area, the side buttons and the sound ticker, so text placed there cannot be read. An ad whose offer or premise sits at the bottom edge loses viewers who watched precisely to find out what was on offer. The TikTok Ads hub lists keeping text out of those areas among the creative requirements.
Fix: move every line of text into the clear area of the frame and recheck the placement on a phone.
Verify: viewers who reach the end tap at a higher rate than before, with watch time unchanged.
Prevent: keep a safe-area overlay in the editing template and check each export against it.
An edit made for silence on a feed that plays sound
TikTok is built as a sound-on feed, and the content formats and hooks guide treats sound as both a creative choice and a signal, while noting that a large share of viewing also happens without sound. An edit built for silent feeds elsewhere, captions doing all the work over generic music, feels foreign here, and viewers leave when the soundtrack tells them they are watching an advert. Captions remain essential for people watching without sound, so the fix is balance, not removal.
Fix: lead with a voice, the product's own sound or an eligible native sound, and let captions follow the spoken line.
Verify: average watch time rises against the silent-first version, and the tap rate with it.
Prevent: brief TikTok creative as sound-on from the start rather than adapting assets made for other placements.
One video asked to carry the ad group
Ads fatigue quickly in a dense feed. An ad group with one or two creatives shows the same videos to the same people until frequency climbs and the tap rate decays. The campaign structure guide covers creative volume and rotation.
Fix: add a handful of genuinely different ads, different in hook, format or angle rather than in a caption word, to the existing ad group, and pause the tired one, rather than rebuilding the ad group.
Verify: the ad group's tap rate recovers and frequency per viewer levels off.
Prevent: keep a queue of new creatives so each ad group always has replacements ready.
A face nobody recognises
An ad under a custom identity is a stand-alone video with a display name and an avatar, and no profile, history or comments behind it. A Spark Ad runs as a real post, from the brand's account or a creator's, with its engagement visible, as the Spark Ads guide explains. When viewers watch but do not trust the ask, identity is a candidate.
Fix: test Spark versions of similar content in the same ad group, holding audience and bidding constant.
Verify: the Spark ads earn a higher tap rate than the custom-identity ads they sit beside.
Prevent: brief organic posts and creator partnerships with promotion in mind, so proven posts are available to authorise.
A closing ask that names nothing
The button comes from a fixed list, so the video has to say what tapping leads to. "Link below" or "check it out" gives no reason to leave the feed; "the full recipe is on the page" or "see every size" does. The TikTok traffic setup recommends naming what the page holds near the end and where the button appears.
Fix: rewrite the closing line and on-screen text to state the specific thing on the page, matching the page itself.
Verify: among viewers who reach the end, the tap rate rises, and landing page views hold their share of taps.
Prevent: write the call to action from the landing page's first screen, not from a generic list.
Prevention: a creative pipeline and rotation calendar
Low tap rates are cheapest to prevent upstream. Keep a pipeline of new creatives in production at all times, each reviewed for its first frame and first line, text inside the safe area, sound-on audio with captions, a specific closing line, and the identity it will run under. Add new ads to existing ad groups on a calendar rather than when the old ones have died, so rotation happens before decay. Weekly, read average watch time, the share passing the opening and the tap rate per ad against its siblings, and flag any ad that has slid for several weeks while frequency rises. Monthly, compare custom-identity and Spark ads in the same ad groups. To prove which opening deserves the next round of briefs rather than leave it to rotation, the guide to testing TikTok ad creative sets out when to use an organic pre-test, rotation or a split test. Price follows engagement, so a falling tap rate usually shows up later as a rising cost of a thousand impressions (CPMCost per 1,000 impressions.); acting on the tap rate first is the cheaper order. The TikTok Ads troubleshooting hub covers the next problem once taps recover: traffic that does not convert.
A pet food brand whose ads were watched but not tapped
A pet food brand runs an ad group with 3 uploaded ads under a custom identity. Over a month the ad group delivers 1.2 million impressions at a CPM of £5.10 and a click-through rate of 0.38%, against 0.9% for the brand's earlier Spark ads. Ad by ad, the watch metrics differ. Ad A opens on the brand logo and only 21% of views pass the opening. Ads B and C hold 48% past the opening and reach the end in about a quarter of views, but their offer, "first bag half price", sits along the bottom edge under the caption, and the closing line is "check it out". Ad A is re-cut to open on a dog finishing a bowl; B and C get the offer moved to the upper third and a closing line saying "first bag half price on the page"; and 2 Spark Ads from a pet creator are added. Four weeks later the ad group's click-through rate is 0.86%, with the Spark ads at 1.1%, and frequency has fallen from 3.4 to 2.2.
Common mistakes
Judging a tap rate against another platform's benchmark. Changing the audience when the opening is the problem. Stripping captions to make room for sound. Rebuilding the ad group instead of rotating ads within it. Chasing tap rate with curiosity hooks the page cannot pay off.
