Which Creative Variant Won, and Why: Automated Categorisation for Cleaner PCA

The winner is in there somewhere. You just cannot prove it.

It is 4pm the day before the client review. You have four platforms open, each with its own naming convention, and you are trying to work out which creative variant actually carried the campaign. One DSP calls it "HeroVid_15s_v2". The social export says "Summer_A_final_FINAL". The display line items are a soup of ad server IDs. Somewhere in that mess is the answer the client will ask for in the first five minutes: which creative won, and why.

You know the campaign performed. You just cannot stitch the story together cleanly enough to defend it. So you eyeball it, make a judgement call, and hope nobody asks about the 15 second cut versus the 30.

That is not analysis. That is a reconciliation exercise dressed up as insight, and it is eating your PCA time alive.

Messy labels produce messy narratives

The root problem is not that the data is missing. It is that the same creative idea wears a different name on every platform. Without a consistent layer that says "these six line items are all the same hero video", you cannot aggregate performance by variant. And if you cannot aggregate by variant, you cannot say anything confident about creative at all.

This matters because creative is the question clients care about most. Media mix and targeting are your craft. Creative is theirs. When they ask which execution to put more budget behind next quarter, a vague answer undermines the whole retention conversation.

Consistent creative categorisation fixes this at the source. When every asset is tagged against a shared taxonomy, by format, duration, message, variant, and funnel stage, the numbers roll up into something you can actually interrogate. The 15 second cut drove efficient reach. The 30 second built consideration and held viewers past the first quartile. The product-led static outperformed the brand-led one on lower-funnel conversions. That is a narrative. That is defensible.

What good categorisation lets you say

Once your variants are labelled consistently across platforms, the PCA writes a lot of itself:

• A clear winner, with context. Not just "Variant A got the best CTR" but "Variant A delivered the strongest completion rate at the lowest cost per completed view, consistent across YouTube and the open exchange."
• Quartile and attention read-outs by creative, not by line item. You can finally compare how different cuts held attention through to the fourth quartile, rather than comparing platform artefacts.
• Message-level learnings that survive to the next brief. When message themes are tagged, you can tell the client that promotional framing beat emotional framing for this audience, and carry that into the next campaign.
• A like-for-like comparison across channels. Because the taxonomy is shared, you are comparing the same creative idea wherever it ran, not comparing apples with ad server IDs.

The difference in the room is obvious. A labelled PCA answers the client's questions before they ask them. An unlabelled one spends the meeting explaining why the numbers do not quite line up.

Why this has to be automated

You could, in theory, build a naming convention, send it round the team, and enforce it by hand. In practice you know how that goes. One trafficker is off sick, a late swap-in misses the convention, a platform truncates the name, and your clean taxonomy has holes in it by week two.

Manual tagging also does not scale across a client portfolio. One campaign is doable. Fifteen live campaigns across eight clients, each with in-flight optimisations, is not. The labelling has to happen automatically as the data comes in, applying your taxonomy consistently whether the asset came from a DSP, a social platform, or an ad server.

That is the point where this stops being a reporting chore and becomes a repeatable process. When categorisation is automated, your analysts spend their time on the "why", not on cleaning up the "what". The insight is the deliverable, not the data wrangling behind it.

Turn creative chaos into a clean story

Media Ridge pulls your cross-platform campaign data into one place and categorises creative automatically against a consistent taxonomy, so every variant rolls up correctly no matter where it ran. The result is a PCA that names the winning creative, explains why it won, and holds up when the client pushes back.

If your last three PCAs involved more spreadsheet reconciliation than actual analysis, it is worth a look. Explore what it does on our features page, or book a demo and bring a messy campaign with you. We will show you what clean looks like.