Solutions / Creative analytics

Ad creative performance analyzer

Every creative you have run, paired with its own spend and metrics and described by a vision model, so the report says which elements drive CTR instead of which ad ID did.

Marketing Paid ads ML
Every creative
in one table, refreshed daily
3–5 hours
a week back from manual reporting
Element level
results, not just ad level
Who it is for

Performance teams running enough creative volume that the pattern is real — roughly thirty or more creatives a quarter across Meta, Google or TikTok.

Short answer

Spend, impressions, CTR and cost per result are pulled per creative from every platform you buy on, and each creative is described by a vision model — format, dominant colour, whether a face is present, how much text, what the hook is, whether the offer appears. Join the two and the weekly answer stops being “ad 4718 won” and becomes “short hooks with the price on frame one beat product shots by 30% on cost per lead, across 84 creatives”.

The problem

You know which ads won. You do not know what made them win, so the next batch is a guess.

01

Reporting is manual and stale

Three platform exports, a pivot table, a screenshot in the deck. It takes half a day a week, and by the time it is read the campaign has moved on.

02

Creatives are not described anywhere

The data has ad IDs and thumbnails. Nothing records that this one had a face, a price on frame one and a two-second hook, so no analysis can ever ask which of those mattered.

03

Learnings do not survive the quarter

What worked lives in a designer’s head and leaves with them. Every new hire relearns the same lessons at the same cost.

How it works

From trigger to result, step by step.

01

Pull the numbers per creative

Spend, impressions, clicks, CTR, conversions and cost per result from Meta, Google, TikTok and wherever else you buy, into one table with one row per creative per day.

02

Describe the creative itself

A vision model tags each image or video frame set: format and ratio, dominant colours, faces, text density, hook type, whether the offer or price is shown, whether it is product, lifestyle or UGC.

03

Join and normalise

Attributes meet metrics, normalised by placement and audience so a comparison is not just a comparison of budgets. The joining is where most of the honest work is.

04

Report what holds up

Which attributes are associated with better CTR and cost per result, with the sample size next to each finding. Where the volume is too thin to say, we print that instead of a number, because a confident wrong answer here costs real money.

05

Turn it into a make-next list

A weekly digest: what to make more of, what to retire, and which combinations have never been tested. It lands in Slack or your BI tool, and it feeds the next creative brief.

Before / after

What changes on the ground.

Today, by hand
×Half a day a week rebuilding the same report
×Creatives described nowhere in the data
×Learnings live in one person’s head
×Next batch is briefed on instinct
With the automation running
›One daily table across every ad platform
›Every creative tagged by element and hook
›Findings reported with sample sizes attached
›A weekly list of what to make next
What you get

Delivered, not demoed.

›A creative-level data warehouse across your ad platforms
›Vision tagging of every creative into structured attributes
›Analysis of attribute against performance, with honest sample sizes
›A weekly digest and a dashboard your team can filter
›Documentation and handover, so the pipeline is yours
Built with

We build in your stack rather than moving you onto ours. The list below is what this solution most often connects to — other systems are a scoping question, not a blocker.

Python pandas Meta Marketing API Google Ads API TikTok Ads API vision models BigQuery dbt Looker Studio
Time to production4–6 weeks
Build priceFixed quote
First stepFree mini-audit
Honest limits

When this is not the right solution.

·If you run three creatives a month, there is no pattern to find. We would be selling you a dashboard that reports noise with a straight face.
·If creative and audience always change together, nothing can separate their effects. We will show you what to vary first, and that costs nothing.
·It reports association, not proof. Treat it as a shortlist for what to test next — the test is still the thing that tells you.

Questions we get about this one

Thirty or more creatives with meaningful spend gives a first read; a hundred makes it solid. We check your volume before quoting, and we will say if it is too early.

Bring us the process that hurts.

The mini-audit is free: we take your version of this process apart and tell you plainly whether automating it pays. If it does, you get a scope and a fixed price.