Hook Rate & Hold Rate: Read Your Video Ads

Most video ads lose their audience before the actual message even starts. Yet many teams judge creatives only by CTR, CPL or ROAS, then guess whether the video, the offer or the landing page is to blame. Hook rate and hold rate show you exactly where your video loses attention: in the first 3 seconds, and after them.

Hook rate is the share of impressions that turn into a video play of at least 3 seconds (3-second video plays ÷ impressions). Hold rate measures how many of those viewers stay, usually as ThruPlays or 15-second plays ÷ 3-second plays. Neither is an official Meta metric: both are ratios you build from standard metrics and set up as custom metrics in Ads Manager.

What exactly are hook rate and hold rate?

Both terms come from media buyers' day-to-day practice, not from Meta's metrics catalogue. Neither "hook rate" nor "hold rate" appears in Meta's field reference for ad insights. That has an important consequence: every agency, tool and blog can calculate them slightly differently. Before you compare numbers, compare formulas.

Hook rate answers the question: does the opening stop the scroll? The common formula is:

Hook rate = 3-second video plays ÷ impressions × 100

In its developer documentation, Meta defines the underlying event as a play in which the video ran "for an aggregate of at least 3 seconds", so three seconds in total, not necessarily in one go. If you want a stricter version, use 2-second continuous video views, which Meta reports as a separate field.

Hold rate answers the question: does the story carry after the opening? There are three common variants:

ThruPlays ÷ 3-second plays (the most widespread)
15-second plays ÷ 3-second plays (Meta tracks an event for at least 15 seconds of aggregate playback)
plays to 25% or 50% ÷ 3-second plays (useful when your videos vary a lot in length)

Pick one variant and write it into the metric's description. Otherwise, three months from now your team will be arguing about numbers that don't measure the same thing.

Which video metrics does Meta officially provide?

You don't need extra tools for the diagnosis. Meta's Insights API and Ads Manager deliver the raw data you combine:

Video plays: how often your video starts playing, counted per impression and excluding replays. Meta flags this metric as "in development" in its API reference.
3-second video plays: at least 3 seconds of aggregate playback, the basis of hook rate.
ThruPlays: the event Meta optimises video campaigns for and can bill on. When Meta made ThruPlay the default optimisation in 2019, Adweek described it as charging for videos watched to completion or for at least 15 seconds.
Video plays at 25%, 50%, 75%, 95% and 100%: the number of plays that reached that share of the video's length, which according to Meta includes plays that skipped to that point.
Average play time: average playback time per impression, including replays.

The percentage markers are your retention profile. Hook rate and hold rate compress it into two numbers; the percentage markers show where the drop-off happens.

Are there equivalents on YouTube and TikTok?

TikTok comes closest to the concept. Its Ads Manager reports 2-second and 6-second video views per impression session, plus views at 25, 50, 75 and 100 percent of length and the average play time per video view. Advertisers using focused view products also get 6-second and 15-second focused views, which also count interactions in the first seconds. So a clean TikTok hook rate is 2-second views ÷ impressions, and a hold rate is 6-second views ÷ 2-second views.

YouTube and Google Ads measure differently. For skippable in-stream ads, a view only counts after 30 seconds or at the end of the video; for in-feed and Shorts ads after 10 seconds, or at the end for shorter videos. View rate is views ÷ impressions. On top of that you get "video played to 25/50/75/100%", calculated on the number of players that started. There is no 3-second threshold. So don't carry your Meta numbers over to YouTube one to one; read the quartiles there as your hold-rate proxy.

How do you read hook rate and hold rate as a diagnosis?

The value of these two metrics isn't in any single number but in how they combine with what comes after: link CTR, landing-page conversion rate, and CPL or CPA. That is how you cleanly separate three sources of failure.

Decision tree for a video ad with weak results:

Hook rate low compared with your other creatives? The opening is failing. First frame, first line, text overlay, pace. Story and offer haven't been tested yet, because almost nobody sees them.
Hook rate good, hold rate low? The opening promises something the middle doesn't deliver. Classic case: a clickbait hook followed by a generic product video. Work on structure, editing and the transition after second 3.
Hook and hold good, CTR low? People watch but don't want to act. That usually comes down to the offer, the call to action or the wrong audience for this promise.
CTR good, conversion rate low? The video works. The problem sits on the landing page or in the form. Here, conversion optimisation will do more than the next creative.

One important rule: always compare relatively, within your own account, the same placement and similar video lengths. A 9-second Reel and a 60-second explainer have structurally different hold rates.

Why benchmarks do more harm than good here

There are plenty of "good hook rates" floating around online. They have three problems. First, the formula isn't standardised. Second, the values depend heavily on placement, industry, video length and audience. Third, the sources are rarely traceable. Even Google cites roughly 10–15% as an average view rate for in-stream ads and stresses in the same sentence that results vary widely.

The better reference is you: take the median of your creatives over the last 30 to 90 days, per placement, as your baseline. A new video is then neither "good" nor "bad", it sits above or below your own standard. That is also the logic behind the idea that creative is the new targeting: you optimise against your own distribution, not someone else's numbers.

Testing logic: let AI variants compete properly

AI tools make it cheap to turn one video into ten variants. That is exactly what tempts teams into bad tests, where hook, music, voice-over, length and offer all change at once. In the end you know which video won, but not why.

A modular approach is cleaner. Keep the body of the video identical from second 3 onwards and swap only the opening seconds: a different first frame, a different opening question, a different text overlay. Hook rate is then the test metric, and hold rate is the check on whether the new hook still fits the rest. Only once you have a winning hook do you vary the middle section, and then you measure hold rate first.

Three rules help:

One variable per test round. Hook variants against hook variants, story variants against story variants.
Same length, same placement. Otherwise you are measuring format differences, not creative quality.
Enough volume before the verdict. Don't decide after a few hundred impressions. Set the impression threshold for your analysis in advance.

If you want to produce AI variants at scale without diluting the brand, a fixed production system for AI content creation with clear rules for hooks, formats and naming pays off.

In practice: set up hook rate and hold rate in Ads Manager

Here is how to build both as custom metrics that are then available in every view:

In Ads Manager, open the Columns dropdown and choose Customise columns.
Next to the search bar, click the link to create a custom metric.
Enter the hook rate formula: 3-second video plays divided by impressions. Autocomplete suggests the metric names.
Give the metric a clear name, such as "Hook rate (3s/impr.)", and set the format to Percentage. Don't skip this step, or you'll see decimals.
Write the formula into the optional description and decide whether everyone in the business account can use the metric.
Repeat for hold rate, for example "Hold rate (ThruPlay/3s)" with ThruPlays divided by 3-second video plays.
Save the columns as a preset, adding impressions, link CTR, cost per result and video plays at 25/50/75/95%.

A formula needs at least two metrics and an operator, and division by zero isn't possible, so the field stays empty for creatives without any plays.

Worked example (hypothetical figures)

Worked example for illustration, not real campaign data. A SaaS company tests two hook variants with an identical video from second 3 onwards:

	Variant A	Variant B
Impressions	40,000	40,000
3-second plays	10,000	5,200
ThruPlays	2,000	1,560
Hook rate	25.0%	13.0%
Hold rate	20.0%	30.0%

Variant B holds its smaller audience better but loses almost half the viewers at the opening. In absolute terms, A still delivers 440 more ThruPlays. The reading: A's hook works, but it also pulls in people who don't care about the topic. Next test: combine hook A with a sharper qualifier borrowed from B and check whether hold rate rises without hook rate collapsing. Whether it's worth it in the end isn't decided by hook rate, but by the CPL or CPA sitting right next to it in the same report.

Hook rate and hold rate in the bigger system

Both metrics are leading indicators. They tell you quickly whether a creative earns attention, long before you have enough conversions for a statistically sound verdict. But they never replace the business metric. A video with a record hook rate that attracts the wrong people is more expensive than one with an average hook and a well-qualified audience. So anchor both metrics in your performance marketing reporting as a diagnostic tool, not a goal.

FAQ

Is hook rate an official Meta metric?

No. Meta provides the components, impressions and 3-second video plays, and you build hook rate yourself as a custom metric. That is why values differ between tools that use different formulas.

Which hold rate formula is the right one?

There isn't a single right one. ThruPlays ÷ 3-second plays is the most widespread. If your video lengths vary a lot, percentage markers such as 25% or 50% ÷ 3-second plays are more meaningful. What matters is that you fix one formula and use it consistently.

Can I compare the numbers across Meta, TikTok and YouTube?

Only to a limited extent. TikTok works with 2- and 6-second thresholds, while YouTube only counts views after 10 or 30 seconds. Compare creatives within one platform and placement, not across platforms.

Sources

Want more of the attention you get to turn into customers? Let’s spend 15 minutes on your setup, free and without a pitch deck. Book your free call

Hook rate and hold rate: how to read the first 3 seconds of your video ads