Build a Creative Performance System That Actually Works

One question that never gets a good answer: "How is the new creative performing?"

It's usually answered with a deck of cherry-picked metrics that can't be compared to anything and won't exist next quarter.

Here's a better way, with an AI assist.

Pull weekly creative-level data from your media platforms — going back as far as you can. Use conversion rates, not CPA or CTR. Index everything against a baseline of the total of the time period you pulled (X years). Bucket by brand vs. performance, campaign, time period, whatever matters to you. Feed it to AI and start asking questions.

Once it's set up, you can finally answer:

• Is creative wearing out — or not?

• Does holiday creative actually perform better than you think?

• Is your brand creative moving upper-funnel metrics or just burning money?

• How does this campaign compare to the one three years ago?

The data foundation takes some work upfront. It's worth it.

THE FULL SETUP

Why conversion rates, not CPA or CTR

CPA mixes creative impact with media efficiency — a great ad in a bad placement looks terrible on CPA. CTR omits too much, and on social especially, the people who click rarely look like the people who buy. Conversion rates isolate creative impact better than either.

Use two metrics for every channel:

• Upper funnel: Traffic conversion rate = Impressions ÷ Visits

• Lower funnel: Your established KPI conversion rate = Visits ÷ Purchases / MQL / form complete / whatever you use

Run both for Brand and Performance spend separately — even when a team isn't focused on both. You'll learn things.

Step 1: Pull the data

Pull from every media platform you run. Go back as far as the platform allows — be greedy, there's no cost to more history.

Cut by week, not month. Campaigns and creative launches rarely align to month boundaries. Weekly data lets you set precise before/after dates, or just group to months when that makes more sense.

Make sure you're at the individual creative unit level. The naming varies by platform — your media team will know what level that is.

For each platform download, you need at minimum:

• Week/month/year

• Channel

• Creative unit name/ID

• Impressions

• Visits and or Sessions

• Your lower-funnel KPI volume

Put everything in one Google Sheet or Excel file. One tab per channel. Don't clean it yet — just get it all in one place.

Step 2: Bucket the creative

You need a taxonomy to compare anything meaningfully. It doesn't have to be perfect — 80% coverage is fine.

Start with two buckets: Brand vs. Performance. This is usually easy — assume everything that ran in the brand budget is brand creative and vice versa. Skip boosted organic or anything else for now.

Add campaign bucketing next. Campaigns can require judgment calls (is just using the campaign tagline in an email footer good enough to be counted?), but bucketing by flight date gets you most of the way there (how was performance during the period of the campaign vs. before and after).

Upload the file to your AI of choice — full model, not a lite version. Tell it your bucketing logic and ask it to label every creative unit. Iterate. Check the output in a different AI with the same prompt. You don't need 100% accuracy here - missing a handful of creative won't skew a big data set.

Step 3: Index everything

Indexing is what makes the data comparable across time periods, channels, and budgets of wildly different sizes.

Ask your AI to do three rounds of indexing:

• Round 1 — Channel-level index: For each channel separately, index each creative unit's conversion rates against that channel's average across the entire time period. 100 = average. Above 120 = better than average. Below 80 = worse. In the middle is average.

• Round 2 — Bucket-level index: Within each channel, index Brand creative against Brand average and Performance creative against Performance average. Now you can see whether Brand is getting better or worse at driving traffic or conversion over time.

• Round 3 — Total index: Combine all channels. Index total upper-funnel and lower-funnel conversion rates across all channels for the full time period. This is your master trend line.

Step 4: Start asking questions

Once indexed, upload to AI and ask all the questions:

• What's the trend in upper vs. lower funnel conversion rates over the past 3 years — are we getting better or worse?

• Which campaigns over-indexed on upper funnel but under-indexed on lower — and vice versa?

• Is Brand creative improving or declining at driving visits? What about conversions?

• At what point does creative typically wear out — and does it differ by channel or bucket? This would in the form of "how many weeks does it take for the index to decline?"

• How does holiday creative perform vs. evergreen across both funnel metrics?

• Which creative units from more than 2 years ago would still be above average today?

The more history you have, the better the answers. Three years of weekly data across four channels is tons to see meaningful patterns.

Have your ops team or channel team set up an API feeding the platform data into the master sheet so you can keep it updated. A key aspect of this is to be clear with the AI or team doing the work what you want to compare the creative against: 1. All creative, or just brand or performance? 2. Just the videos currently live? 3. Just when the campaign was live, or also before and after? 4. Just this year or quarter, or the full time period or some specific timing?