Bannerflow has given a name to the AI features it has been building inside its Stockholm creative platform: Creative Intelligence. The October 12 launch groups image generation, predictive attention scoring, element-level performance analysis and feed-driven dynamic ads into four stages: Create, Evaluate, Understand, and Act and scale.
The company counts more than 2,000 brand customers and over EUR 25 million in annual recurring revenue. The pitch is simple: judge an ad’s likely attention before any media budget goes live. The evidence is thinner. Bannerflow disclosed no availability dates, no pricing and no validation data for the predictive models doing that judging.
What sits inside the suite
- Create: generate, edit, expand and upscale images, plus describe a product feed in plain language instead of configuring fields.
- Evaluate: attention heatmaps, focus analysis and clarity scoring to flag distracting elements before launch.
- Understand: Creative Insights links elements such as calls to action, colour and animation timing to campaign performance.
- Act and scale: feed-driven dynamic ads keep pricing, stock and availability synced across products, formats and markets.
Where the proof is missing
Bannerflow does not explain how the attention model was trained, what eye-tracking or panel data sits behind the heatmaps, or how closely its predictions have matched real outcomes. That gap matters because attention metrics have become their own currency in programmatic buying, and vendors selling them typically publish at least some validation against in-market results.
“Marketing teams have more performance data than ever, but too little of it inspires the next piece of creative.”
That is Chief Product Officer Moses Velasco, and the diagnosis is fair. The question is whether a model with no published accuracy can fix it.
Competition is already scoring creative
Meta added Creative breakdown in Ads Manager, Google put one-click creative testing into Asset Studio, and Innovid launched auto-labelling features. A Kroger, Vidmob and MMA study reported 81% prediction accuracy across 1,934 ads. Bannerflow’s announcement has no comparable figure.
There is also a distinction worth keeping: Google’s move to put Adelaide attention scores into DV360 bidding measures attention in flight. Bannerflow’s heatmaps operate on the creative file before it runs. They are complementary in principle, but a pre-launch prediction is not the same as observed in-market attention.
What to do about it
The workflow pain is real. Separate EMARKETER and Perion research covered by PPC Land in April 2026 found 89.2% of marketers called creative important, but only 3.6% said it was well understood and actively optimised. Feedback often arrives two to four weeks after launch.
A pre-flight score could save budget if it is validated. Until then, ask these questions before buying:
- Which verticals and outcomes was the model tested against?
- Can creative insights be exported, or do they stay inside Bannerflow?
- Will you run a controlled test before making scoring a default workflow step?
Bannerflow’s own platform data shows iGaming made up 76.1% of impressions in 2026. A model shaped by that history may not transfer to every brand. Treat attention heatmaps as a directional signal, not a budget decision.
Source: PPC Land
