Most performance marketing teams do not have a media buying problem. They have a creative volume problem. Ad platforms like Meta now do most of the targeting work for you, which means the creative itself has become the main lever. The team that can test more good ideas, faster, usually wins.
At Per4mance Guru we run paid social and search for brands in India, the UAE and other markets, and creative testing kept becoming the bottleneck. This post walks through the simple workflow we now use, and what we learned building it. No fancy stack required.
The problem: strategy is slow, production is slower
A typical creative cycle used to look like this:
- Someone pulls last month's ad data into a spreadsheet.
- A strategist reads through it and writes a brief.
- A designer or editor produces two or three variations.
- The ads go live, and the cycle starts again.
Each step depends on a person, and each hand-off loses context. By the time a new creative launched, the insight behind it was often a few weeks old.
Step 1: Pull the data in one place
The first fix was boring but important. We pull ad-level data (spend, impressions, click-through rate, cost per result, hook rate for video) into a single table every week. Even a Google Sheet works. The goal is that nobody has to open five dashboards to answer "what worked last week?"
A small script can do this through the platform APIs, but you can start with manual exports. Consistency matters more than automation at this stage.
Step 2: Let AI do the first read, not the final call
We give the weekly table, plus the ad copy and a short description of each creative, to a large language model with a fixed prompt. The prompt asks it to:
- Group ads by angle (for example price, social proof, problem and solution, founder story)
- Point out which angles are improving and which are tiring out
- Flag ads where the click-through rate is good but the conversion rate is poor, which usually means the landing page and the ad are telling different stories
The output is a first draft, not a decision. A strategist reads it, disagrees with parts of it, and adds what the data cannot show, such as seasonality or a competitor launch.
Step 3: Turn insights into a structured brief
The biggest time saver was a fixed brief format. Every brief has the same fields:
- Angle: the single idea the ad is built around
- Audience moment: where the viewer is in their decision
- Hook: the first line or first three seconds
- Proof: what makes the claim believable
- Format: static, carousel, UGC style video, and so on
Because the structure is fixed, AI can help draft several hook and copy variations for each brief, and humans pick and edit the best ones.
Step 4: Use AI for production where it is genuinely good
AI image and video tools are very good at some things and weak at others. In our experience they work well for:
- Background and scene variations of an existing product shot
- Quick concept mock-ups to align a team before a real shoot
- Resizing and adapting one approved creative into many placements
They are weaker at keeping a brand's exact look consistent, at product accuracy and at anything that needs a real human face to build trust. We keep a person in the loop for every final asset.
What we learned
- Speed without strategy just creates more bad ads. The pipeline only works because a human still owns the angle.
- Fixed formats beat clever prompts. A consistent brief template improved output more than any prompt tweak.
- Measure the right thing. Track how many new angles you test, not just how many ads you produce.
- Keep the data loop short. Insights that reach the next creative within a week are far more useful than a perfect monthly report.
Wrapping up
You do not need a big team or an expensive tool to start. A weekly data table, a fixed prompt for analysis, a strict brief template and an AI tool for variations will already shorten your creative cycle a lot.
If you are working on something similar, I would love to hear what your workflow looks like in the comments. You can also find more about how we approach performance marketing at per4mance.guru.
Top comments (0)