Competitor Analysis
Weekly competitive intelligence on rivals' search visibility, price moves and campaign timing, delivered as a prioritized opportunity list.
Explore
AI-assisted analysis
Ad Creative Analysis: AI-Tagged Creative Performance and Data-Driven Briefs
A single metric rarely tells you why an ad works. In one video, the opening scene, the second the product appears, subtitles, music, the offer and the length all play a part at once. Our ad creative analysis breaks the creatives in your Meta, TikTok and Google Ads accounts into these elements, tags them with AI models, matches them to performance data and shows which element is actually driving results.
Purpose and appearance
What do we use this service for?
We expect creative analysis to show which hook, message, format and offer drive sales, so that new creatives are produced on that evidence. The goal is to catch creative fatigue early, stop wasting tests and write briefs that bring down cost per conversion. The findings set the order of new tests on Meta, TikTok and Google Ads.
Funnel stages it serves
- Awareness
- Consideration
- Conversion
- Loyalty
Which creative elements do we tag?
The models read the image, video frames and copy and apply the same tag set to every ad. The tags contain no subjective labels such as 'beautiful' or 'eye-catching'; each tag is an attribute that can be marked as present or absent. That way, two different ads can be compared on the same scale, and findings do not depend on anyone's personal taste.
- Hook: is there a product, a human face, on-screen text or a question in the first 3 seconds
- The second the product first appears and how long it stays on screen
- Audio: voiceover, music, subtitles suited to sound-off viewing
- Format: 9:16, 4:5 or 1:1 aspect ratio, video length, number of carousel cards
- Production type: studio shoot, user-generated content, creator content, product animation
- Offer and call to action: discount, installments, free shipping; buy now, message on WhatsApp
Analyzing creative performance without false conclusions
This is the stage where wrong conclusions are most often drawn. Automated setups such as Advantage+ sales campaigns and TikTok Smart+ concentrate budget on a few creatives; an ad with little spend can look weak not because it is bad, but because the system never tested it properly. That is why ads below the impression threshold are left out of the results table.
We compare creatives only within the same campaign objective, the same audience type and similar placements; a video served to remarketing audiences is never judged on the same scale as one served to cold audiences. For videos, we divide 3-second views by impressions to read the strength of the hook, and use completion rate to read how well the content holds attention. Our team manually checks a sample of the tags every month; models can make mistakes, especially with Turkish on-screen text and fast-cut videos.
Questions the creative tag table answers
For each question, the model returns an answer along with the number of ads and the spend behind it. A finding based on only a few ads is not written up as a rule; it is flagged as an idea to test. We stress-test findings by asking the same question to two models separately; if the answers diverge, the decision is made from the raw performance table.
- Is there a difference in cost per conversion between videos that show the product in the first second and those that show it later?
- Do creator videos work better with cold audiences or in remarketing?
- In which price range does highlighting installments increase click-through rate?
- Which creative type loses performance fastest as frequency rises?
From creative insight to brief, from brief to test
Approved findings become a short brief for the creative team: what the hook should look like, the second the product should appear, which aspect ratio and length, which offer. New creatives are slotted into the test calendar of the 90-day plan: A/B tests on Meta, creator variations with Spark Ads on TikTok, and asset tests in Demand Gen on Google. Test results feed back into the tag table.
We do not recommend this analysis for accounts with few ads in total that launch only a handful of new creatives a month; the tags cannot find enough samples to show a meaningful difference. When data is limited, we start with simple A/B tests. We review competitors' creatives as a separate piece of work in our competitor ad strategy analysis.
How we manage Ad Creative Analysis
- Creative inventoryWe pull ads published over the last 6-12 months into BigQuery, along with their spend and results.
- Fixing the tag setWe define the tag set by category and refine the tagging instructions using sample ads.
- Creative tagging and manual verificationThe models extract hook, audio and format tags from frames and copy; our team manually verifies a sample and adjusts the instructions accordingly.
- Thresholds and comparison groupsWe set the impression threshold and group ads by objective, audience type and placement.
- Briefs and test calendarWe turn approved findings into briefs and write into the plan how new creatives will be tested on each platform.
Ad Creative Analysis: metrics and requirements
Metrics we track
- 3-second view rate
- Video completion rate
- Click-through rate
- Cost per conversion
- ROAS
- Creative fatigue curve
What you need to get started
- Read access to Meta, TikTok and Google Ads accounts
- Access to creative files or ad previews
- Conversion tracking that works correctly
- A creative team or creators to execute the briefs
Official sources
The Brand Move Data and AI Team
Contact us nowTo learn more about Ad Creative Analysis or other AI-assisted analysis services, contact The Brand Move Data and AI team today.
Explore further
AI-Assisted Strategy overview
Other AI-Assisted Strategy services
Seasonal Behavior Analysis
A brand-specific season curve from past demand, traffic and sales data. We place budget before demand rises.
ExploreCompetitor Ad Strategy
Competitor ad research from public ad libraries: we map rival messages, offers and formats and find the positioning space left open.
ExploreFrequently asked questions
Ad Creative Analysis: frequently asked questions
What is ad creative analysis?
It is the process of breaking ads into comparable elements, such as the hook, product timing, audio, format and offer, and linking each element to performance. The output is not a ranking of individual ads but evidence on which elements work, which feeds the next brief.
Does the AI produce the creatives itself?
No. This work is analysis and does not cover production. The models tag existing ads and match them to performance; our team writes and approves the brief. Creatives are then produced to the brief by your design team or creators.
How many ads do we need for the analysis to be meaningful?
There is no single number. What matters is that every group being compared contains ads with enough impressions and conversions. If the number of ads is low, we flag findings as test ideas and deepen the analysis as data builds up. We set the scope after looking at your account data.
How do you spot creative fatigue?
We track each creative type's performance as frequency rises and plot a fatigue curve. This shows which formats lose efficiency fastest, so replacements can be briefed before results drop.
How is this different from competitor ad analysis?
Creative analysis works on your own ads and their performance data. Reviewing competitors' creatives is a separate piece of work, covered in our competitor ad strategy analysis.
Is personal data shared when creatives are sent to the model?
No. The models receive ad images, copy and aggregated performance figures; no user-level data such as names, emails or phone numbers is sent, and raw data stays in your BigQuery project. Data sent through the enterprise API is not used for model training by default.
Let's run a first Ad Creative Analysis review together.
In a free call we look at your competitors and seasonal data together and show you, with a concrete example, what AI-assisted analysis can deliver first.
Book a growth call
No sales pressure. Just findings.