Dynamic Product Ads (DPA)

DPAs pull product images, prices, and descriptions from a product feed and dynamically serve the right products to each user based on browsing behavior. Production concepts decide whether creative can be made at catalog scale or only for hero moments. Anything that requires a person to open a canvas per SKU stops working around the fortieth product. The brands that ship the most variants treat production as a system of templates, rules and specs rather than a series of one-off files. Read it next to Product Feed, catalog-ads, Retargeting.

Define it once, then apply it everywhere. Lock type ramp, palette, logo placement, offer treatment and legal lines in a shared brand system, and build templates that accept the product feed as input. Every export should already respect the placement's safe zones so nothing needs a manual crop after the fact.

Quality control is checkable, not subjective: does the asset meet the platform's ratio and file requirements, does key content sit inside the safe area, is text legible at thumbnail size, and does the product read in the first frame? Run that checklist before spend, because rejections and silent crops cost more than the render did. Why this matters commercially: DPAs power the highest-ROAS placements in ecommerce — retargeting cart abandoners with the exact product they viewed.

In Xeli this is a rule an agent enforces, not a note in a brief. Templates read the catalog, respect the brand system, and output every required ratio for every SKU at ad-ready resolution — so a 400-product drop becomes a queue to approve rather than a month of files. In the context of creative & production, that means the concept stops being something a person re-applies by hand every campaign and becomes a rule the system enforces on every asset it produces.

The recurring problems are predictable: running dpas on a broken feed; not overlaying brand elements on feed images; using dpa for cold traffic without a broad-audience variant. Each of these is a process gap rather than a knowledge gap — which is why the fix is usually a checklist, a template or an automated rule instead of more training.