Solutions · Marketing and PR
Internal trial: ad concepts, revised after feedback.
INTERNAL TRIAL, 19 October 2026. We used Mixa to make ad concepts for our own company, reviewed them against our founder’s positioning notes, then revised the audience and the creative.
The pain
Our first ads spoke to “busy small-business owners”. Our founder’s own notes put the company somewhere else: personalised apps for each employee, for operations-heavy businesses. A generic ad would have wasted the first test budget.
What we tested
Round 1: two square ad concepts for a broad small-business audience, made with Mixa’s built-in image generation. Feedback: the founder’s recorded positioning decisions, with no ad platform data. Round 2: two concepts for operations managers at multi-site logistics and hospitality businesses, built around one real task, a dock shift handover.




What we measured
Copy accuracy: 4 of 4 images reproduced the headline and subline exactly. Brand rules: 3 of 4 broke our “square corners only” rule, and 1 of 4 added stray bold text. Round 2 names one audience and one task; round 1 names neither. Each round took about a minute to generate.
There is no reach, click or conversion data. The audience change is a judgment from written feedback, not a measured uplift.
Verdict: works for drafts, needs setup to learn
Making and revising creative works today. Learning which audience is better does not: Mixa has no ad-account connection, so the feedback loop runs on human notes only.
Missing piece and next step
Missing: a read-only connection to an ad account’s results, and brand rules the image step enforces itself. Next step: Engineer scopes a read-only ad-results connector. Product Lead writes the brand rules into the generation prompt template.
Limitations
Internal trial on our own company, not a client result. One run, one person reviewing, no control group. No ads were published, targeted or paid for. The “feedback” is our founder’s written positioning, not audience response.
See it on your own work
These are trials on our own company. We would run the same kind of trial with you, on your tools and your rules.
See a Demo