The same ad should be treated as one creative asset across Google, Meta and TikTok, even when each upload has a different name or sits inside a different campaign structure. Score it against a shared business objective, keep platform conditions visible, and separate the strength of the creative from the way each platform delivered it.
That is the foundation of useful cross platform creative performance analysis. Without it, one video can look like several unrelated ads, and the media buyer loses the pattern that should guide the next round of production.
Give the creative one identity before comparing it
The first problem is not the score. It is identity.
A video might be called Testimonial_Final_2 in the source folder, renamed for a Meta ad, uploaded to TikTok under a campaign convention, and added to Google under another label. The names describe the uploads, not the underlying asset.
Create a stable creative ID before trying to compare performance. That ID follows the concept everywhere it runs.
A practical record might include:
| Field | What it identifies | Example |
|---|---|---|
| Creative ID | The shared source concept | TESTIMONIAL_A |
| Platform | Where this version runs | Meta |
| Format | The delivered asset type | Vertical video |
| Version | A meaningful edit | Short opening |
| Offer | The commercial message | Consultation |
| Batch | The production round | October batch |
| Launch date | When this version began running | October 1 |
| Destination | The landing page or offer | Consultation page |
The creative ID remains fixed. Platform, format and version can change.
This distinction matters when the opening changes but the rest of the video does not. A new opening is a version of the same concept if the claim, speaker and offer remain the same. A different speaker making a different argument is usually a new concept.
Write that rule down. Otherwise, the classification will drift as more ads enter the account.
Separate concepts from executions
A concept is the central idea of the ad. An execution is how that idea appears in a particular file or placement.
For example, the concept might be a customer explaining the problem that led them to buy. Its executions could include a vertical cut, a shorter cut and a version with a different first frame.
Grouping those executions under one creative ID lets you answer two different questions:
- Is the concept working?
- Which execution gives that concept its best chance on each platform?
Do not collapse every version into one row. Keep the individual executions available beneath the shared asset. The rollup reveals the larger pattern, while the version data explains it.
A naming system is useful, but names alone are fragile. Media buyers inherit accounts, campaigns get duplicated, and files are uploaded again. Keep a creative register outside the platform names, then map every live ad back to it.
Build a score around the buying objective
Cross platform creative performance is not a contest between the native dashboards. Each platform reports within its own delivery system, attribution settings and campaign structure. A fair comparison starts with the buying objective, not the most flattering metric available in each interface.
Begin with the outcome the campaign is meant to produce. That might be a qualified lead, a purchase or a visit to a key page. Then place supporting signals underneath it.
| Layer | Question | Useful signals |
|---|---|---|
| Business outcome | Did the creative contribute to the intended result? | Conversions, conversion value, CPA or ROAS |
| Traffic quality | Did it produce useful visits? | Landing page visits, cost per visit, on-site actions |
| Response | Did people react to the message? | Clicks, click rate, cost per click |
| Delivery | What conditions shaped the result? | Spend, impressions, reach, frequency, CPM |
| Fatigue | Is the asset weakening as exposure grows? | Frequency rising alongside CPM, CPC or CPA deterioration |
The primary score should follow the campaign target. Supporting signals explain why the result looks the way it does.
If the account is managed to CPA, a cheap click is not enough to declare a winner. If it is managed to ROAS, conversion count without conversion value is incomplete. If an asset has barely spent, its apparent lead should be treated cautiously.
Keep unlike conditions visible
A platform comparison becomes misleading when it hides important differences.
Check these conditions before ranking assets:
- Was the objective equivalent?
- Was the offer the same?
- Did every version send traffic to the same landing page?
- Were the audiences at a similar stage of awareness?
- Did the campaigns run during overlapping dates?
- Was one asset limited by budget or delivery?
- Did one version receive enough activity to support a decision?
You do not need identical setups. You need to know where the setups differ.
A creative can be strong on one platform and weak on another without either result being wrong. The audience context may differ. The format may fit one placement better. The platform may also have delivered the asset to a different mix of people.
Preserve those differences in the analysis instead of forcing one universal verdict.
Use a scorecard that a buyer can explain
A useful scorecard should make the decision clearer. It should not bury the buyer under a single mysterious number.
Start by grading each asset on three questions:
- Did it produce the target outcome?
- Did it produce that outcome efficiently enough for the account?
- Is its recent direction stable, improving or weakening?
Then add a confidence label based on how much evidence is available. Plain labels work well:
- Strong signal
- Directional
- Inconclusive
“Inconclusive” is a valid result. It means the asset has not produced enough evidence for a clean decision, or the conditions are too different for a fair comparison.
Example: one video with three platform versions
This is an illustrative example.
Say a campaign has a $50 target CPA. The same testimonial video runs on Google, Meta and TikTok under different upload names.
The buyer maps all three uploads to TESTIMONIAL_A. The scorecard shows:
| Platform | Result against target | Delivery context | Reading |
|---|---|---|---|
| Near the $50 target CPA | Consistent delivery | Keep running and watch direction | |
| Meta | Below the $50 target CPA | Frequency is rising | Current winner, prepare a replacement |
| TikTok | Above the $50 target CPA | Limited conversion activity | Inconclusive, do not overreact |
The useful conclusion is not simply “Meta wins.”
The concept has produced a credible result across more than one platform. Meta currently has the strongest execution, but its rising frequency means the buyer should not wait for a clear decline before preparing another version. TikTok has not yet earned a positive or negative verdict.
The next action could be to preserve the testimonial concept, make a fresh opening for Meta, and test a platform-specific cut on TikTok. That decision comes from the combined pattern, not from choosing the lowest row in a report.
Avoid blending everything into one average
A combined number can hide the reason an asset is working.
Suppose most of the spend went to one platform. A portfolio average would mostly describe that platform, even though the label suggests a balanced cross-platform result. The same problem appears when one version has a much longer run than the others.
Show the combined result, but keep platform rows and execution rows underneath it. The buyer should be able to move from the rollup to the actual source of the result.
Turn the pattern into the next creative batch
The point of creative scoring is not to decorate a report. It is to make the next production request more specific.
Once the winners and weak spots are clear, break each asset into elements the creative team can act on:
- Opening frame or first line
- Core claim
- Proof or demonstration
- Speaker or visual subject
- Offer
- Call to action
- Length and format
Then identify which element appears to repeat across the strongest executions.
If several winning versions use the same problem statement, keep that statement and change the opening treatment. If the concept works but one format trails, adapt the execution before discarding the concept. If every execution of the concept struggles under comparable conditions, the central idea may need to change.
Plan controlled variations
Each new batch should answer a question.
A practical brief might request:
- One close variation of the current winner
- One version with a different opening
- One new concept based on another customer concern
- One staged replacement for an asset showing fatigue
These are not promises that a particular change will improve performance. They are defined tests that make the result easier to interpret.
When too many elements change at once, the next result tells you that the new ad won or lost, but not why. Keep a record of the variable each version is meant to test.
For a broader production process, use the creative testing framework for agencies. When the main question is volume and delivery, the guide to testing new creative without starving your winners covers how to introduce new work while protecting current performers.
Watch for fatigue as a pattern
Fatigue is not one metric crossing a universal line. Look for a combination of changes around the same asset.
Frequency may rise while CPM or CPC also rises. CPA may begin getting worse. The pattern becomes more useful when it appears after the asset has been delivering consistently, rather than during a short period of uneven delivery.
When that pattern appears, the decision is not always to abandon the concept. The buyer might rotate in a staged version with a new opening, visual treatment or edit while preserving the proven argument.
This is where cross platform creative performance becomes operational. The result is a production instruction tied to account evidence, not a general request for something fresh.
Frequently asked questions
How do you compare creative performance across platforms?
Give the asset one shared creative ID, group every placement where it runs, and compare results against the same business objective. Review both outcome metrics and delivery conditions so platform allocation does not get mistaken for creative quality.
Can the same ad work on Meta and TikTok?
Yes, the same core ad can work on both platforms, but its performance may change with the placement, audience and surrounding campaign. Track the source asset as one concept while keeping platform versions separate.
What is creative analytics?
Creative analytics is the practice of connecting ad assets to delivery, engagement and conversion data. It helps a media buyer see which concepts are working, where they work and when their performance is weakening.
How do you name creative so it can be tracked across platforms?
Assign each concept a stable creative ID and add separate fields for the platform, format, version and launch batch. Keep the shared ID unchanged when the same source asset is uploaded under a different platform name.
How do you choose the next batch of creative?
Start with the strongest repeatable pattern, then change one meaningful element in each new version. Include a controlled variation, a new angle and a replacement for any asset showing fatigue.
What Ads Arsenal does for this problem today
Ads Arsenal finds the same creative across campaigns and platforms, then scores it as one asset. It shows what is winning and what is getting tired, using signals such as frequency rising, CPM or CPC increasing, and CPA getting worse.
Those signals produce a recommendation to swap in creative you have already staged. The reasoning and account numbers appear with the recommendation, and nothing changes until a person approves it. Ads Arsenal is in a free design-partner beta for agencies and in-house teams running real spend on Google, Meta and TikTok.