Ad variant engine
Same team, same hours, 3x the creative output, with every variant on brand and ready to A/B test
- Every ad variant built by hand, from scratch, by in-house designers
- Under deadline pressure, variants got cropped rather than composed
- Not enough ads to test
- Multiple on-brand variants generated from a single approved concept
- 1:1 and 9:16 outputs produced automatically for every campaign
- Designers freed for the creative direction that actually needs them
A large Dubai real estate brokerage was running paid ads at volume, and their in-house design team was the bottleneck.
Every campaign started with a developer brochure. Designers pulled the imagery, built the creative, then rebuilt it again for each placement ratio. When timelines were tight, they cropped instead of composing, and the work suffered.
The constraint wasn't talent, it was arithmetic. A/B testing needs variants, and variants cost designer hours the team didn't have.
SetMyClaw built an engine that takes the brochure and the team's approved concept, then generates on-brand variants and placement ratios automatically, so the same designers ship several times the testable output in the same working day.
They had the audience, the budget and a capable creative team. What they didn't have was a way to produce enough variants to learn from.
- Every creative built from scratch, with no reusable production layer
- Manual resizing to each placement ratio, repeated for every campaign
- Time pressure forced cropping over proper recomposition, hurting quality
- Copy variants limited to what a person could write and lay out by hand
- Brochure imagery extracted manually before any design work could begin
- Not enough variants in market to A/B test with any confidence
- Off-the-shelf generators ignored brand guidelines and produced unusable output
Phase 1: Source extraction
- Developer brochures parsed automatically, with imagery and key details pulled out
- Assets normalised into a structured library the engine draws from
- Brand guidelines encoded as constraints, not suggestions, so output arrives on brand
- Existing campaign creative used as reference, categorised by type and past A/B pairing
Phase 2: Variant generation
- Multiple creative variants produced from a single approved concept
- Copy variants written and composed onto the creative for headline testing
- Both in-painting on existing assets and fully generated alternatives, depending on the brief
- Model guided by paid social best practice, so variants are testable rather than merely different
Phase 3: Placement automation
- 1:1 and 9:16 outputs generated for feed and stories from one master concept
- Elements recomposed for each ratio rather than cropped to fit
- Day and night treatments produced from a single source image
- Variants delivered ready to load, requiring no manual prep before launch
Phase 4: Feedback loop
- Ad performance pulled back in so the engine learns which variants actually convert
- Automatic weighting toward the styles and copy angles that win
- Extension into listing photography, automating the exposure, lighting and shadow work the team documented from their photographers' own process
The developer brochure is parsed automatically, with imagery and key details extracted into a structured asset library.
The team builds the master concept exactly as they always have. The engine works from their creative judgement, never around it.
Multiple on-brand alternatives are produced from that single concept, including copy variants composed directly onto the creative.
1:1 and 9:16 versions are recomposed for each placement rather than cropped, along with day and night treatments where useful.
Every variant is reviewed before anything ships. The engine proposes, the team decides.
Approved variants are delivered ready to load, with enough volume in rotation to A/B test properly.
Let's identify the highest leverage opportunities in your business.
Book a strategy call and I'll show you exactly where AI can make the biggest impact.
- ✓Custom audit of your funnel
- ✓Prioritized automation opportunities
- ✓Clear next steps