Meta Ads video CTR benchmarks: the 3 ratios to measure

In 2026, the median CTR on Meta sits between 1.93% and 2.39% depending on the source: 1.93% at WordStream/LocaliQ (Sept. 2026), 2.39% at Triple Whale across more than 40,000 accounts. Neither figure isolates video. For a video creative, three ratios can be benchmarked: the hook rate (measured median 20.9%), the hold rate (median 4.6%) and CTR, read last. And your first reference remains your own last thirty days: public benchmarks are American, in dollars, and aggregate every format. On the French side, Médiamétrie’s l'Année Internet 2025, relayed by CB News in February 2026, shows that 68% of French people check their social networks and messaging apps every day and spend an average of 1h13 a day on them, or 39% of their total time online.

What is the average CTR of a video ad on Meta in 2026?

Nobody publishes it. The four most-cited Meta Ads CTR benchmarks in 2026 give a range of 1.93% to 2.39%, but none separates video from image, carousel or dynamic catalogue. So you are comparing your vertical Reel with an average that includes static retargeting ads in the Feed. That mix carries real weight: according to The Media Leader, drawing on the 2025 review from the Observatoire de l'e-pub, video accounts for nearly 60% of social advertising revenue in France, growing 22% against 8% for traditional formats.

The figures, side by side. Triple Whale, on a panel of more than 40,000 accounts and 17 sectors covering August 2025 to July 2026, reports a median CTR of 2.39%, up 15.97% from the previous year’s 2.06%. WordStream/LocaliQ, in its update of 14 September 2026, gives 1.93% on US traffic campaigns, against 1.71% in 2025. Flighted publishes 2.19%. MHI Growth Engine, on an e-commerce scope, drops back to a 1.93% median, with a top 25% at 2.78% and a top 10% at 3.71%.

Four sources, four figures, a 24% gap between the lowest and the highest. That is not statistical noise. It is four different definitions of the same word.

The only figure in the corpus that claims to speak about video comes from Prooflytics, which puts the CTR of a Meta video ad between 1.0% and 2.5%, with a 3-second view rate of 25 to 40% and a ThruPlay of 10 to 20%. The range is so wide that it contains pretty much the whole market. Useful for knowing you are not at 0.2%. Useless for choosing between two creatives.

One detail in the Triple Whale report deserves a second read: the median CPM rose from $13.30 to $15.06 (+13.24%) while WordStream’s median CPC fell from $0.70 to $0.60. Impressions cost more, clicks cost less. Mechanically, that means the market’s average CTR is going up. Meta confirms the pressure on media buying: the company reported an average price per ad up 12% in its Q2 2026.

Why do two reports give different CTRs for the same year?

Because “CTR” covers at least three distinct metrics in Ads Manager, and public reports almost never say which one they exported. Meta exposes CTR (all), CTR (link clicks) and outbound CTR. On a single video creative, the gap between the first and the third is commonly double, sometimes triple.

CTR (all) counts everything: link clicks, but also likes, comments, shares, clicks on the page name, caption expansions, and clicks on the video to open it full screen. On a chatty Reel with lots of comments, this figure swells without a single visitor landing on the product page. CTR (link clicks) counts only clicks to a destination, including destinations inside Meta. Outbound CTR keeps only exits from the platform. It is the most honest, the lowest, and the one almost nobody publishes.

Four other causes explain the rest of the gap.

• The campaign objective. At MHI Growth Engine, traffic campaigns show a 2.19% CTR at a $10 to $13 CPM, conversion campaigns 1.93% at a $16.80 CPM. The algorithm is not looking for the same people.
• Audience temperature. At MHI, retargeted audiences show a higher CTR than cold traffic, but at a CPM of $18 to $24: check this in your own columns before generalising.
• Placement. A full-screen Reel and a desktop right-hand column have nothing in common. The panels aggregate both.
• Sample composition. Triple Whale mostly measures high-volume American Shopify merchants. WordStream measures SMEs across all sectors, including insurance and legal services. These are not the same advertisers.

Before comparing your CTR with a benchmark, check one thing: the column you are looking at. If the report does not specify whether it is talking about CTR (all) or CTR (link clicks), the figure is not comparable, and no media decision should depend on it.

We have applied this rule internally since a client reproached us for a CTR “below market”: he was reading the outbound CTR in his dashboard, we were reading the CTR (link clicks) in Ads Manager. Same account, same week, 1.4% versus 2.6%. Nobody was wrong. Nobody was talking about the same thing.

Is CTR the right metric for judging a video creative?

CTR is the last link in a three-stage chain, and therefore the least diagnostic. A video that underperforms on CTR can fail in three very different places: it doesn’t stop the scroll, it stops the scroll but loses attention before the argument, or it holds attention but gives no reason to click. CTR alone doesn’t tell you which of the three.

The three ratios, with the exact formulas as buyers actually calculate them:

• Hook rate (also called thumbstop ratio or scroll-stop rate) = 3-second views ÷ impressions. Measures how well the first second stops the thumb.
• Hold rate = 15-second views ÷ 3-second views, or 75% views ÷ 3-second views depending on the school. Two different denominators, and that is where everything is decided.
• CTR = link clicks ÷ impressions. What is left once the first two have done their job.

If you need a clean definition of what a hook is, how it is built and what it really costs to produce, the dedicated article covers it. Here, we only talk about the ratio and what it is worth as a decision tool.

An important point, rarely stated: Meta publishes no official hook rate benchmark. The platform doesn’t even name the metric. The hook rate is a construct of media buyers, calculated from two native columns. Every figure you will read on the subject comes from private panels or practitioners’ statements. No central authority, no canonical definition. That is why the ranges contradict each other so much.

What are the real hook rate and hold rate benchmarks in 2026?

The measured median hook rate is 20.9%, not 30%. The Biddyco study, covering 479 Meta videos and $10.1 million in ad spend, gives a median of 20.9%, a first quartile at 14%, a third quartile at 30% and a top 10% above 45%. On the hold rate side, the median stands at 4.6%, with quartiles at 3% and 7%.

These figures clash with what is circulating. Adsights, in an update of 27 July 2026, cites a practitioner range of 25 to 35% with an anchor around 30%, distinguishing Feed cold (18–28%) and Reels (24–36%). Eonik, on 6 June 2026, sets thresholds: hook rate above 25% acceptable, 30–35% excellent, below 15% structural failure. And for hold rate: 20–25% strong, below 12% failure.

A median hold rate of 4.6% against a failure threshold of 12%. If both were talking about the same thing, more than 90% of the market would be in structural failure. Obviously that is not the case.

The explanation comes down to a denominator

Meta natively exposes views at 25%, 50%, 75% and 100%, plus 3-second views and ThruPlay. Depending on which column you take as the numerator, you get two universes of figures with no relation to each other. A hold rate calculated as 75% views ÷ 3-second views runs around 4 to 7% on short video. The same asset, calculated as 15-second views ÷ 3-second views, can show 20%. It is not the creative that changes. It is the division.

The operational conclusion, and it is a dry one: a hold rate benchmark without an explicit formula is worthless. Before adopting a threshold you found on LinkedIn, ask for the formula. In nine cases out of ten, nobody can answer.

A methodological note on the gap between Biddyco and practitioners. The round ranges, 25–35% and a 30% anchor, are declared figures, often drawn from the best-managed accounts of whoever is speaking. The 20.9% median is a measurement across the full distribution, losers included. Both are true in their context. But if you set 30% as your floor, you will kill creatives that are above the market median.

Does the hook rate really predict winning creatives?

No, and it is the most uncomfortable lesson in the 2026 corpus. The Biddyco analysis of 479 Meta videos and $10.1M of spend concludes that the hook rate cannot identify winning creatives. In other words: the metric the ecosystem has been selling for three years as “the” video metric is not a reliable selection criterion.

The mechanism is easy to understand once you’ve seen it. Two videos can both show a 30% hook rate and have a ROAS gap of one to six. The first stops the scroll with a promise that matches the product; the second stops the scroll with a shock shot, a screaming figure or a visual that has nothing to do with what is being sold. In both cases, the thumb stops. In only one case does the person buy.

We have produced an unreasonable number of hooks that performed on the ratio and died at checkout. The worst example remains a “POV: you just received your order” hook edited over a generic unboxing. Hook rate around 38%, far above the Biddyco third quartile. Decent CTR. Zero sales beyond day one. The hook was good. It said nothing about the product.

What this changes in an iteration process:

• The hook rate becomes a low-end elimination filter, not a promotion criterion. Below 14% (Biddyco first quartile), we look at the thumbnail and the first second. Above that, we conclude nothing.
• The kill threshold is set on a business metric, CPA, ROAS, cost per add-to-cart, with an impressions floor, never on an attention ratio alone.
• Hook rate comparison only makes sense between variants of the same angle. Comparing the hook rate of a product demo with that of a customer testimonial tells you nothing usable.

And when your CTR falls week after week, don’t jump to the easy conclusion: a falling CTR is not necessarily ad fatigue. An audience change, seasonality, a widening of the learning phase all produce exactly the same curve.

Can these American benchmarks be transposed to the French market?

Rates transpose roughly, costs not at all. A hook rate, a hold rate, a CTR are unitless ratios: they depend on the creative, the placement and the audience, not the currency. A CPM, a CPC, a CPA are amounts, and they are structurally different between the American and French markets. The French market also follows its own dynamic: according to the 2025 review of the Observatoire de l'e-pub by the SRI, Social grew there by 15% to reach €4.18 billion, or 34% of the digital advertising market, driven by the rise of video.

The order of magnitude of the gap. Triple Whale gives a global median CPM of $15.06. Lebesgue, in its e-commerce benchmarks by country, puts the Facebook CPM in France at around $6.95. It is tempting to conclude that France costs half as much. That would be exactly the mistake this article denounces: two panels, two periods, two sector scopes, two aggregation methods. The real gap exists, American ad density is higher, auctions are tighter there, but the factor of two is not demonstrated by these two figures placed side by side. We cite it as an illustration of the comparability problem, not as steering data.

The same reasoning applies from one platform to another: the equivalent benchmarks on the TikTok side show cost structures different from Meta’s on the French market, without making one “cheaper” than the other in absolute terms.

The Advantage+ black hole

There is a problem deeper than currency. According to MHI Growth Engine, Advantage+ Shopping accounts for 62% of e-commerce conversions in 2026, and 89% of e-commerce advertisers have deployed CAPI. When 62% of your conversions go through a campaign where you no longer choose the placement, benchmarking a “Reels CTR” becomes structurally impossible on the majority of spend. Your impressions are split between Feed, Reels, Stories, Audience Network and Marketplace according to a trade-off you no longer control and that reporting only partially details.

Today there is no way to reconstruct a clean video CTR by placement on an Advantage+-dominated account. Nobody knows how, and the tool vendors who claim otherwise are estimating. The operational answer we apply: isolate a test budget in a manual campaign with forced placement, 15 to 25% of spend, solely to produce readable figures on new creatives. The rest runs on Advantage+ and is judged on CPA.

What moves by sector

The sector gaps recorded by Triple Whale weigh more than country gaps on many accounts. Beauty shows a median CPM of $18.80 with a conversion rate of 1.79%, down 10.45% year on year. Health and wellness rise to a $21.80 CPM for 1.50% conversion, down 13.99%. Fashion and accessories stay at $13.25. On CTR for traffic campaigns, shopping and collectibles climb to 4.13%, travel to 2.76%, sport to 2.60%.

If you sell facial skincare and compare yourself with an all-sector benchmark, you are partly comparing yourself with sellers of trading cards. The word “benchmark” doesn’t excuse that.

How do you build your own video Meta Ads CTR benchmarks in 14 days?

Your useful reference is the distribution of your last thirty days, segmented on three axes and locked into a written definition. Fourteen days are enough to produce a first usable baseline if you have volume, provided you don’t segment more than your spend allows.

The minimum segmentation comes down to three dimensions: placement (Reels / Feed / the rest), temperature (cold prospecting / retargeting), format (video / static). Six to eight cells at most. Beyond that, each cell contains three creatives and the median no longer means anything.

The impressions floor. We don’t read a CTR below 8,000 to 10,000 impressions per creative, and we don’t read a hold rate below 3,000 3-second views. Below that, the confidence interval is wider than the gap you are trying to measure. For the detail of normalisation, how to neutralise period, budget and seasonality effects, the 14-day protocol published on the TikTok side transposes entirely to Meta: only the exported columns change.

The dashboard fits in three lines per cell: median, third quartile, number of creatives. No average, one viral creative blows it up and you set yourself an unattainable target. No maximum either, for the same reason. The median says where you are, the Q3 says what your best can do, the count says whether the figure deserves a look.

Last rule, the one always forgotten: write the definition into the file. “CTR = link clicks ÷ impressions, CTR (link clicks) column, Ads Manager, account time zone.” Three lines of documentation that will spare you, six months later, from re-comparing two incomparable figures, as the whole ecosystem has been doing for three years.

How many creatives do you need for an in-house benchmark to mean anything?

Below ten distinct video creatives per segmentation cell, you don’t have a median, you have an anecdote. The Biddyco study rests on 479 videos. You will never have that many per month. But between 3 and 10, the jump in readability is clear.

The arithmetic is brutal. If you publish four videos a month and segment two as cold and two as retargeting, your “cold benchmark” rests on two assets. The median of two values is their average. A single bad hook moves it by twenty points.

The accounts where ratio-based steering really works run at a minimum of twelve creatives a month, a majority of them iterations on already-validated angles rather than new concepts. Throughput is not a production whim: it is the condition for any figure to become readable. On the mechanics between creative volume and time to readability, how many creatives before a figure is readable details the calculation.

And if you can’t sustain that throughput, own it: steer on CPA and ROAS, ignore attention ratios. A shaky in-house benchmark leads to worse decisions than no benchmark at all. We have seen teams kill their best creative of the year because it was at a 17% hook rate in the first week, on 4,000 impressions. It came back in retargeting three months later and lasted six weeks.

The market publishes Meta Ads video CTR benchmarks that aren’t: all-format averages, in dollars, with no definition of the click. The real question is not whether you are at 1.9% or 2.4%, but how many of your creatives exceed your own third quartile, and why those ones.

Frequently asked questions

What is a good CTR for a Meta video ad in 2026?

The all-format reference range runs from 1.93% (WordStream/LocaliQ, Sept. 2026) to 2.39% (Triple Whale, 40,000+ accounts). In e-commerce, MHI Growth Engine puts the top 10% at 3.71%. None of these figures isolates video: treat them as an order of magnitude and first compare yourself with the median of your last thirty days, segmented by placement and audience temperature.

What is the difference between CTR (all), CTR (link clicks) and outbound CTR?

CTR (all) includes likes, comments, shares and clicks on the page. CTR (link clicks) counts only clicks to a destination. Outbound CTR keeps only exits from Meta. On a single video, the gap between the first and the third is commonly double. No public benchmark specifies which one it measures.

How do you calculate the hook rate on Meta?

Hook rate = 3-second views ÷ impressions. Meta doesn’t name this metric and publishes no official benchmark for it: it is a construct of media buyers. The Biddyco study, on 479 videos and $10.1M of spend, measures a median of 20.9%, a first quartile at 14% and a top 10% above 45%.

Why is the median hold rate 4.6% when you often read 20%?

Because they are two different formulas. Biddyco calculates 75% views divided by 3-second views, which gives a 4.6% median. Most buyers calculate 15-second views divided by 3-second views, which produces values around 20%. Always demand the formula before adopting a threshold.

Can you use American CTR benchmarks for a French account?

Rates, yes, with caution: hook rate, hold rate and CTR are currency-free ratios. Costs, no. Triple Whale gives a global median CPM of $15.06 while Lebesgue puts the France CPM at around $6.95, but these two figures come from different panels and methods and prove no precise gap.

Does Advantage+ prevent you from benchmarking CTR by placement?

Largely, yes. Advantage+ Shopping concentrates 62% of e-commerce conversions according to MHI Growth Engine in 2026, and the advertiser no longer chooses the placement there. To keep readable figures, we isolate 15 to 25% of the budget in manual campaigns with forced placement, dedicated to testing new creatives, and judge the rest on CPA.