Creative Performance by Platform: Meta, TikTok and LinkedIn
Meta, TikTok and LinkedIn each publish their own guidance on what makes an ad work, and the three sets of guidance do not match. TikTok asks for a hook in the first six seconds. LinkedIn reports its highest completion rates on videos of 7 to 15 seconds. Meta assembles combinations of your assets and reports the average across them.
This post sets out what each platform recommends in its own words, what its reporting does and does not tell you, and the steps that make a comparison across all three possible.
What the research says about creative
Nielsen published an article in October 2017 citing Project Apollo research from 2006, which found that 65% of a brand's sales lift from advertising came from the creative. The same article reported that the effect of media on sales had risen to 36% from 15% over the previous 11 years. The research is old, and the finding is the reason to measure creative carefully: the creative accounts for more of the result than the placement does.
Each platform puts that creative in front of people under different conditions. Placements differ, attention patterns differ, sound defaults differ, and the amount of automation between the file you upload and the ad someone sees differs. Measuring creative one platform at a time gives you three answers that do not line up.
Meta: one ad can run many combinations
Meta's developer documentation recommends wide distribution. On ad creative it says, "We encourage you to run ads across the full range of available placements," and that the easiest way to take advantage of that optimization is to leave the placement field blank.
Meta also assembles creative on your behalf. With Dynamic Creative, the documentation says, "we automatically deliver different combinations to different users." Setting ad_formats to AUTOMATIC_FORMAT lets a single ad mix images and videos across placements. Format automation, in Meta's words, lets you "create a single ad that automatically delivers multiple optimized ad formats."
That changes what the reporting means. You upload a pool of headlines, images and videos, Meta assembles combinations from that pool, and performance reports at the ad level. The number you read is an average across combinations you did not brief individually. To learn which part of the creative drove the result, you need a record of which assets went into each ad, and performance read at the element level.
TikTok: published specs and several creatives per ad group
TikTok's creative best practices for performance ads give specific recommendations:
Vertical video at 9:16, at a minimum of 720p resolution
The content proposition in the first 3 seconds
The hook in the first 6 seconds
On-screen text at 5 to 10 words per second
Sound or music
3 to 5 different creatives per ad group, and 3 to 5 diversified ad groups per campaign
The volume recommendation has a measurement consequence. Following it produces dozens of similar variants in a month, each differing from the others by a hook, a caption, a first frame or a voiceover. TikTok's reporting tells you which variant performed. It does not tell you which of those differences accounted for the performance.
LinkedIn: short videos, fast rotation, character limits
LinkedIn's best practices for video ads recommend keeping videos short. The page cites LinkedIn data from 2025 in which videos of 7 to 15 seconds saw up to a 300% lift in completion rates. The same study found that longer videos drove as many clicks and views as shorter ones, so LinkedIn also recommends testing longer cuts for demand generation, at three lengths: 5 to 15 seconds, around 30 seconds, and a minute or more.
The page adds two points about attention. Show what you want the audience to see in the first 10 seconds, because attention drops after that. Burn in subtitles, because a large share of members watch with the sound off.
LinkedIn's best practices for Sponsored Content cover the rest: 4 to 5 ads in each campaign, headlines under 150 characters, descriptive copy under 70 characters because anything over 100 can truncate on desktop, and images at 1200 x 627 pixels. For rotation, it recommends pausing the ad with the lowest engagement every 1 to 2 weeks and replacing it with new creative.
A 1 to 2 week rotation gives each asset a short run, so what it taught you has to be recorded while it is still live.
The three platforms measure different things
The guidance points at three different measures. TikTok's points at the first 3 and 6 seconds. LinkedIn's points at completion rate and the first 10 seconds. Meta's reporting aggregates combinations it built. Each measure works inside its own platform, and none of them can be compared with the others.
A comparison across platforms needs three conditions:
Metrics defined the same way in every account. Spend, impressions, clicks and the conversion you care about qualify. Platform-specific engagement rates do not.
A record of what was in each asset: length, aspect ratio, opening frame, whether a person speaks, whether there is text on screen, the offer and the claim. Without it you can rank ads, but you cannot say what to repeat in the next brief.
The same date range on every platform.
How to compare creative across the three platforms
Choose the metrics that mean the same thing on all three platforms and build the comparison on those.
Record what is in each asset. Ad names like Q4_V3_final do not describe the creative.
Use the same date range everywhere, every time.
Run the same creative element on more than one platform, so you can separate how much of the result came from the element and how much came from the platform.
Write the next brief against the elements that performed, and name them.
This is record keeping more than analysis. The work is attaching a description of what you made to the performance it earned, and keeping both after the campaign ends.
Where mktg.ai fits
mktg.ai reads the creative you have live and connects performance to the elements inside each asset, so the next brief gets written from evidence rather than instinct. It covers TikTok, Meta, Google, LinkedIn and The Trade Desk, so the comparison described above happens in one place.
For the fundamentals, start with Creative Performance Measurement: The Complete Guide. For the category, read What Is Creative Intelligence? A Plain-English Guide. For how the tools compare, see The Best Creative Analytics Tools in 2026, Compared. For why the underlying data decides what you can learn, read The Data Substrate Matters. For the long history of wasted ad spend, read Half the Money I Spend on Advertising Is Wasted.
FAQ
Does the same creative work on Meta, TikTok and LinkedIn?
The idea usually carries across all three. The execution has to change, because each platform publishes different guidance on length, framing, sound and text, and each applies a different amount of automation before the ad is served. Test the idea on each platform and build the execution for the platform it runs on.
How long should a video ad be on each platform?
LinkedIn recommends keeping videos short and reports that videos of 7 to 15 seconds saw up to a 300% lift in completion rates in its 2025 data, while also recommending tests at around 30 seconds and at a minute or more for demand generation. TikTok's creative best practices for performance ads do not name a single length; they ask for the content proposition in the first 3 seconds and the hook in the first 6. Meta sets its specs by placement and publishes them in its Ads Guide rather than as one recommended length.
How many creative variants should I run?
TikTok recommends 3 to 5 different creatives per ad group and 3 to 5 diversified ad groups per campaign. LinkedIn recommends 4 to 5 ads in each campaign. Both numbers assume you can tell afterwards which variant won and what made it different.
How often should creative be refreshed?
LinkedIn recommends every 1 to 2 weeks: pause the ad with the lowest engagement and replace it with new creative. Treat that as the shortest cycle in your plan, and record what each asset taught you before it is paused.
Why do the platforms report different numbers for the same creative?
Each platform counts a different event, applies its own attribution window, and applies its own automation. A view on one platform is not the same event as a view on another. A comparison across platforms has to be built on the metrics that are defined the same way everywhere.
What should I fix first if I cannot compare creative across platforms?
The record of what is inside each asset. Most teams already have spend and impressions in a usable form. Few have a description of the creative itself, and without that description every comparison stops at which ad won.
Book a demo
If you want to see how this works, book a demo.
Comments