Hook rate, hold rate and ThruPlay: the short-form ad metrics that matter

Ads Manager can show a play, a view, a ThruPlay and a click for the same video, but those words do not describe the same event. This guide is for the person deciding what to change in the next cut: it separates the platforms' defined counts from the ratios practitioners calculate, then walks an invented funnel from impression to action. You can use it to read a dashboard without pretending that a single “view rate” diagnoses your ad.

Start with the event the platform counted#

A count is an event under a platform's reporting rule. A rate divides one count by another; its meaning depends on both the event and the denominator. Meta publishes definitions for three-second plays, ThruPlays and link clicks. TikTok publishes definitions for two-second views, six-second views and destination clicks. “Hook rate” and “hold rate,” by contrast, are practitioner formulas, not standardized names in those platform metric definitions.^{[1][2][3][4][5][6]}

That distinction changes your creative decision. If the counted event includes an engagement, a supposedly strong view rate might be partly an interaction rate. If a milestone includes skips to a point, it is not a pure retention curve. If the denominator is reach instead of impressions, the result can move even when no creative changed. Write the metric definition beside every chart before comparing two variants.

Meta's public help pages distinguish starts, timed plays and time watched. A “video play” starts when the video plays, whether by autoplay or click, counts per impression and excludes replays.^[7] A three-second video play counts a play of at least three seconds, or nearly all of a shorter video, with replay time excluded; a two-second continuous play needs two continuous seconds and most such plays have at least half the pixels in view.^[1] Those are different gates on the same viewing journey, not interchangeable synonyms for “someone watched my ad.”

Meta says timed duration metrics such as ThruPlay use unique seconds watched rather than adding rewatched segments, while its average-play-time calculation includes replay time.^[10][9] For example, replaying the opening cannot be treated as reaching the later threshold just because the summed seconds are long enough. A rising average with flat ThruPlays might mean some people are replaying a portion; it does not by itself prove they reached the ask. Check the platform's actual column definition instead of reverse-engineering intent from a pleasing number.

TikTok and Shorts count different thresholds#

TikTok defines a two-second video view as at least two seconds in an impression session, excluding replays. A swipe away and return counts as a new impression session.^[3] Its six-second video view is more surprising: it can count six seconds watched, full play for a shorter clip, or an engagement within the first six seconds.^[3] When you use six-second views as evidence of attention, an early interaction can qualify without a literal six seconds of watch time.

TikTok also has six- and fifteen-second focused view metrics in focused-view products, with interaction alternatives. Do not substitute those counts for the ordinary video-view columns without checking which product your report uses.^[3] The platform's “Clicks (all)” include social interactions; “Clicks (destination)” lead to a specified destination, and the corresponding cost-per-click definition uses destination clicks.^[4] For a landing-page diagnosis, those destination clicks answer a different question from likes or profile interactions.

Shorts adds a third vocabulary. Google says a Shorts ad impression counts when the ad starts playing in the Shorts feed, and a swipe back to the same ad does not create another impression or view. Its ad TrueView view threshold is ten seconds or the full video if shorter, with CTA clicks and some social actions also qualifying; an engaged-view uses five seconds or those actions. The public view counter can advance as soon as playback starts.^[11][12] Google renamed the ad “Views” label to “TrueView views” in October 2025.^[11] A Shorts public view, an Ads TrueView view and a Meta three-second play therefore measure different things even when the screen says “views.”

There is no useful universal table converting them. Keep platform and campaign product in the chart title, and compare like with like: Meta three-second plays with Meta three-second plays on similar placements; TikTok two-second views with the same TikTok metric; Shorts engaged-views with the same Google definition. Even quartile columns may differ in handling of skips or replays. The apparent precision of a percentage does not erase the difference in how its numerator was collected.^[1][3]

Hook rate: name the gate and denominator#

For Meta creative work, a common SECONDARY practitioner convention calls three-second plays ÷ impressions × 100 the “hook rate” or “thumb-stop rate.” Motion's glossary uses these names for the same formula.^[5] Meta supplies the counts; the label and ratio are the analyst's construction. Some practitioners use “thumb-stop” for a reach-based ratio instead, so seeing that label in a slide is not enough to know what was calculated.^[13]

Why impressions? They describe opportunities delivered, including ones where the video never gets a meaningful play. Meta's explanation of its impression counting says a video impression does not require playback.^[14] With three-second plays in the numerator, this ratio asks: of displayed opportunities, how many became at least a three-second play under Meta's counting rule? It does not say how many unique people were persuaded, and repeated impressions to the same person can matter. Divide by reach instead and you are answering a different question about unique accounts exposed.

On TikTok, many analysts use two-second views ÷ impressions for an analogous early gate, a SECONDARY convention rather than a TikTok-defined “hook rate.”^[15][3] Because the thresholds differ, a 28% Meta three-second ratio and a 35% TikTok two-second ratio cannot be ranked as if they were the same experiment. On Shorts, the start-of-play impression definition makes a Meta-style impression-to-play comparison especially unhelpful; use the campaign's Google-defined view or engaged-view alongside its own impression rule.^[11]

Do not borrow a widely circulated “good hook rate” target without its audience, placement, metric definition and creative goal. Motion offers an industry-dependent practitioner range, but it is SECONDARY advice, not a platform benchmark or a pass/fail standard for a new account.^[5] Your more useful baseline is your own comparable cuts with stable denominators. If a cut gets more early plays but fewer downstream clicks at comparable delivery, the opening may be attracting the wrong expectation rather than improving the ad.

Hold rate: three formulas, three questions#

Hold rate is even less standardized. Motion's SECONDARY practitioner formula is ThruPlays ÷ three-second plays × 100 on Meta.^[6] Other practitioners use fifteen-second plays ÷ three-second plays or plays at 75% ÷ three-second plays.^[16] These are not three spellings of the same statistic: the first uses Meta's completion-or-fifteen-second rule, the second uses a time threshold, and the third uses a share of whatever duration you made.^[8][1]

For a Meta dashboard comparing same-length ads, we recommend the ThruPlay formula only if your question is progression from the early gate to Meta's defined result. Write “ThruPlays / three-second plays” in the chart header and keep the video lengths next to it. If you want to know whether people get to a particular spoken claim, use the event closest to that moment; if you want relative progress across lengths, a quartile can help, with the skip caveat. Never splice one definition into an earlier chart because it gives a prettier result.

Suppose a fictional brief cut is shorter than fifteen seconds and most viewers nearly finish it. Meta can record a ThruPlay at the 97% short-video condition; that does not mean viewers watched fifteen seconds.^[8] Calling it a “fifteen-second hold” would invent an event the cut could not contain. Likewise, a high 75% milestone on a very short clip does not guarantee the viewer heard an ask that begins after that position. Tie the denominator and endpoint back to the actual edit.

For TikTok, use its native two- and six-second columns when you want an early/middle comparison, but describe it explicitly as six-second views ÷ two-second views; remember that the six-second count admits an early engagement.^[3] That is an analyst-created ratio, not TikTok's official hold rate. If you are measuring a longer video, its quartiles can provide a complementary location-based read. Do not label a TikTok ratio “Meta hold” and place it in the same series.

A fictional funnel, calculated without shortcuts#

Here is an invented Meta example, chosen for easy arithmetic rather than offered as an industry benchmark: 10,000 impressions → 2,800 three-second plays → 700 ThruPlays → 84 link clicks. We deliberately put events in a planning sequence. The chart does not claim that all 84 clicks necessarily came from those 700 ThruPlay events; platform columns are aggregate counts, and someone can click before finishing.^[17][8]

Fig. 01

1. Impressions10,000
2. 3-second plays2,800
3. ThruPlays700
4. Link clicks84

• Hook: 28% · 3-second plays / impressions
• Hold: 25% · ThruPlays / 3-second plays
• Clicks / ThruPlays: 12% · link clicks / ThruPlays

Illustrative, not data[meta-video][meta-link-clicks][motion-hook][motion-hold]

The first ratio is 2,800 / 10,000 = 28.0%, the named practitioner hook formula. The second is 700 / 2,800 = 25.0%, the named practitioner ThruPlay hold formula.^[5][6] The third is 84 / 700 = 12.0%, a useful aggregate comparison only if labeled exactly that way; it is not a platform CTR and not a conditional conversion probability. Meta's defined link CTR is 84 / 10,000 = 0.84% in this fictional report.^[2] If you care about qualified site visits, check outbound or destination behavior and the landing page rather than interpreting every link click as a purchase.

Now take a separate, fictional TikTok-shaped set: 10,000 impressions → 3,500 two-second views → 1,400 six-second views → 70 destination clicks. Its adjacent invented ratios are 35.0%, 40.0% and 5.0%; the platform-defined click-through share of impressions would be 0.7% under TikTok's destination-click definition. The six-second count could contain early engagements, so even this neat descending diagram is not a literal replay of each person's watch session.^[3][4][18]

Fig. 02

1. Impressions10,000
2. 2-second views3,500
3. 6-second views1,400
4. Destination clicks70

• Early gate: 35% · 2-second views / impressions
• Middle gate: 40% · 6-second views / 2-second views
• Clicks / 6s: 5% · destination clicks / 6-second views

Illustrative, not data[tt-video][tt-clicks][tt-ctr]

Figures help locate a weak stage, but the paths between aggregate counts are hypotheses. If a destination click happened immediately, its counter need not wait for the six-second gate. For the stronger causal question—whether one opening actually improved outcomes—compare variants with matched objectives, attribution and delivery conditions. The small-budget testing guide explains why early apparent wins can be noise.

Ask the next question, not just for a larger number#

Start with the most specific failure you can observe. Are people even seeing the opening? Are they reaching the proof? Are they acting after they understand it? Does the action pay for the spend? Each question needs its own measure and a look at the actual cut. A single composite score would hide which production choice deserves revision.

Fig. 03

1. 01 → Did they stop?Named early-play rate
2. 02 → Did they stay?Named hold formula
3. 03 → Did they act?Link CTR / destination
4. 04 → Did it pay?Outcome against spend

Illustrative, not data[meta-video][meta-link-ctr][meta-cpm][tt-video]

If the early rate is weak, inspect the first visible frame and the first line. Does it name a real problem, show the product and make a promise the middle can deliver? The second-by-second anatomy guide gives a way to rewrite that opening and preserve the evidence. If early attention is strong but later progression weak, inspect the first cut after the hook: you may have teased a result and then wandered into background information. If both are strong but link CTR is weak, look at the offer, CTA, price and destination—not just at faster edits.

If clicks are strong but attributable outcomes are weak, the weak stage may be the page or the offer. Check whether the landing page matches what the ad demonstrated and whether the conversion measurement is configured as intended. You cannot repair a mispriced item or a slow checkout by adding a louder first frame. Separate creative diagnosis from business diagnosis before spending a new test budget.

CPM is spend / impressions × 1,000 in Meta's definition, and CPC (link) is spend / link clicks; they tell you how much delivery or a click cost, not whether the demonstration was persuasive.^[19][20] A cheaper CPM with poor progression can be less useful than a higher CPM that reaches a relevant audience. Always place the spend, objective, audience, placement and observation window next to a funnel comparison.

Worked diagnosis: the same CTR, different failures#

Consider three fictional Meta ads with the same 10,000 impressions and 100 link clicks each, so each has the same 1.0% link CTR under Meta's definition.^[2] These invented counts are deliberately simple. We also assume the same length, placement mix, audience and reporting window for the exercise only; real delivery rarely holds those constant by accident.

The equal CTRs do not mean the ads are equally effective or that the clickers followed the same path. A may be reaching a small, motivated subset; B may attract curiosity without proving the claim; C may persuade viewers who still dislike the price. The table alone cannot establish those motives. Watch each cut, inspect the placement breakdown, and propose one change tied to a visible defect. Then decide the sample and outcome before running a comparison, as described in the creative-testing guide.

The comparison also shows why “fix the lowest percentage” is a weak rule. A's later progression looks relatively high partly because its early gate is small. B's early rate looks excellent while its proof may disappoint. For each variant, write the actual viewer question the edit should answer, and choose a next cut that could falsify your diagnosis. A different thumbnail alone is not a fix if the problem is an unsupported promise.

Metrics that mislead, and common mistakes#

An unlabeled view count collapses starts, timed views, engaged views and public counts into one word. A video can accumulate starts without meaningful watching, or a Shorts ad can gain public views immediately while its ad-view threshold is different.^[7][11] Report the exact column and platform beside the number. Do not quote a Meta “video play” as if it meant a ThruPlay.

“CTR” also needs a qualifier. Meta's CTR (link click-through rate) uses link clicks and impressions; TikTok separates “Clicks (all)” including social interactions from destination clicks.^[2][4] A dashboard's all-click CTR can rise when people interact without visiting the page. If your decision concerns the site, choose the destination-oriented column and check the actual landing-page result.

Mixed placements make averages slippery. A Reels creative may appear beside other formats, and different cohorts can receive different frequency or watch under different interface controls. A healthy total can conceal a weak placement; a weak total can reflect a shift in where delivery went. Break out placement, campaign objective and cut length before attributing a change to the edit. Average play time includes replays in both Meta and TikTok definitions, so a mean also cannot substitute for the distribution of where viewers leave.^[9][3]

Common mistakes#

• Calling hook and hold official metrics. Their component counts are defined by platforms; their chosen ratios are practitioner conventions.^[5][6]
• Changing denominators halfway through a report. Reach-based thumb-stop and impression-based hook can differ without a single frame changing.^[13]
• Reading ThruPlay as exactly fifteen seconds for every video. Meta includes near-completion of shorter videos.^[8]
• Treating a six-second TikTok view as guaranteed six-second attention. An early engagement also qualifies.^[3]
• Assuming funnel counts identify a person's path. An aggregate click may arrive before the later play threshold; use the ratios to form hypotheses, not to reconstruct a session.
• Judging creative only by cheap reach or total CTR. Check the stage and the business outcome with the same placement and objective before commissioning another cut.^[19][2]

Checklist#

• Name the platform, campaign product, placement, cut length and date range for every comparison.
• Record the exact column definition and denominator beside each custom hook or hold rate.
• On Meta, distinguish plays, three-second plays, ThruPlays, quartiles and average play time.
• On TikTok, note the engagement alternative in six-second views and use destination clicks for site traffic.
• On Shorts, separate impression, public view, engaged-view and TrueView view before comparing numbers.
• Calculate early attention, later progression and link or destination CTR separately.
• Watch the cut at the weakest stage and write one testable change before making a variant.
• Compare the landing-page result with spend; do not let a cheap impression masquerade as a sale.

Sources

1. Meta Business Help Center, “About video ad metrics,” including three-second, two-second continuous, ThruPlay and milestone definitions, www.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3] ↩^[4] ↩^[5] ↩^[6] ↩^[7] ↩^[8] ↩^[9]

2. Meta Business Help Centre, “CTR (link click-through rate),” link clicks divided by impressions, www.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3] ↩^[4] ↩^[5]

3. TikTok Ads Manager Help, “Video play metrics,” including starts, timed/engaged views, quartiles and replay treatment, ads.tiktok.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3] ↩^[4] ↩^[5] ↩^[6] ↩^[7] ↩^[8] ↩^[9] ↩^[10] ↩^[11] ↩^[12] ↩^[13] ↩^[14] ↩^[15]

4. TikTok Ads Manager Help, “About basic metrics and definitions,” all versus destination clicks, ads.tiktok.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3] ↩^[4]

5. Motion, “What is Thumbstop rate / hook rate?” SECONDARY practitioner formula and non-platform benchmark discussion, motionapp.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3] ↩^[4] ↩^[5]

6. Motion, “What is Hold rate?” SECONDARY practitioner ThruPlay/three-second-play formula, motionapp.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3] ↩^[4] ↩^[5]

7. Meta Business Help Center, “Video plays,” www.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3]

8. Meta Business Help Center, “15-second ThruPlays,” including the 97% short-video condition, www.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3] ↩^[4] ↩^[5] ↩^[6]

9. Meta Business Help Center, “About video ad metrics calculation,” including replay time in average play time, www.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3]

10. Meta Business Help Center, “About video ad metrics calculation,” unique seconds for timed thresholds, www.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩

11. Google Ads Help, “About YouTube ads and view metrics,” including Shorts impression, TrueView and engaged-view definitions and the October 2025 rename, support.google.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3] ↩^[4]

12. Google Ads Help, “Frequently asked questions” about YouTube Shorts ads, support.google.com (opens in a new tab) (accessed 2026-09-28). ↩

13. Sepia Lab, “Hook Rate Calculator,” and Tadka, “Hook Rate Formula 2026,” SECONDARY reach/impression naming disagreement, sepia-lab.com (opens in a new tab) and tadkai.io (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2]

14. Meta for Developers, “Metrics for Bidding with Mediation,” describing impression counting for image and video ads; this page is in Audience Network documentation, developers.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩

15. Motion Help Center, “Tell stories and build hypotheses with metrics,” SECONDARY practitioner two-second TikTok versus three-second Meta convention, help.motionapp.com (opens in a new tab) (accessed 2026-09-28). ↩

16. adlibrary.com, “Hold Rate Benchmarks and Fixes for Meta Video Ads,” SECONDARY practitioner fifteen-second and 75%-milestone alternatives, adlibrary.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2] ↩^[3]

17. Meta Business Help Center, “About link clicks and outbound clicks,” www.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩

18. TikTok Ads Manager Help, “About ads metrics in Seller Center,” destination CTR definition; cited from Seller Center rather than core Ads Manager help, ads.tiktok.com (opens in a new tab) (accessed 2026-09-28). ↩

19. Meta Business Help Centre, “CPM (cost per 1,000 impressions),” www.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩ ↩^[2]

20. Meta Business Help Centre, “CPC (cost per link click),” www.facebook.com (opens in a new tab) (accessed 2026-09-28). ↩