Personalizing Product Imagery Across Thousands of SKUs

Retailers personalize product imagery across thousands of SKUs by generating each image from the live catalogue and the shopper’s own data, rather than hand-building variants. Whilter’s Snype did this for bigbasket, matching imagery to what each cohort actually buys: funnel completion rose 42%, repeat orders 31% and basket size 26%.

The arithmetic is what breaks hand-built personalization. A few thousand products, a dozen customer cohorts, five channels and three formats add up to hundreds of thousands of possible images. No studio makes that many, so most retailers send everyone the same banner and call it a campaign.

What tools help retailers personalize product imagery across thousands of SKUs?

Tools that generate imagery from data instead of editing it by hand. In Whilter’s stack, that work sits in two engines today and a third that is on the way.

• Snype generates a 1:1 personalized image and video for every recipient from first-party data, with product recommendations, offers and calls to action merged in. Whilter’s e-commerce page describes it as high-SKU personalized imagery: creative adapted to the individual, so a large catalogue stays personal without hand-built assets.
• ElevateOS generates 40–60 brand-compliant video and static ad variants per campaign, with multiple hooks, formats and angles per product, for Meta, Google and TikTok.
• Stud.io, Whilter’s creative acceleration platform for brand-safe variants at scale, is on the roadmap for H2 2026 and isn’t available yet.

Whatever you shortlist, ask whether the tool reads your live catalogue or a one-off export. An export goes stale the day prices change.

Why does hand-built imagery break at catalogue scale?

Because the work grows with every product, cohort and channel you add, and the budget doesn’t. bigbasket’s case describes the pattern. As a high-SKU grocery marketplace, it ran generic broadcasts because manual production couldn’t personalize at catalogue scale. Each shopper’s cohort behavior existed in the data but never reached the imagery or the message they saw.

W for Woman hit the same wall from the other side. A personal try-on visual for every shopper, across a high-SKU fashion catalogue, is far beyond what any studio can shoot.

Generation changes which part of the work grows. The template, the brand rules and the data connections are built once. After that, a new product or a new cohort is a new row of data, not a new brief.

What data does per-shopper imagery need?

Two kinds: the catalogue, and the shopper. From the catalogue you need the product images, names, live prices and stock, so every image resolves to something the shopper can actually buy. Whilter’s partners page puts live catalogue, pricing and stock in the data layer for exactly that reason. From the shopper you need first-party behavior, such as what they buy, how often and in which categories, grouped into cohorts you can act on.

Snype connects to a CRM, CDP, spreadsheet or product database with merge fields and segments, and you choose which fields drive the personalization. No third-party or PII data is required. Start with one category where the catalogue data is clean, so the first test measures personalization rather than data quality.

How do you keep every image true to the product?

Personalize around the product, never the product itself. The product’s color, packaging, label and claims sit in a fixed layer that generation does not touch. The variable layer is everything around it: the selection of products, the scene, the headline, the offer and the language.

That split matters most where the shopper sees the product on themselves. Snype’s virtual try-on and face replacement work from a single source asset, so the product image the shopper sees starts from the one your team approved. Before launch, check a sample from every category against the source images, because a generated image that misrepresents a product is a returns problem, not a creative one.

Where should personalized product imagery run?

Wherever the shopper already looks. Snype delivers on WhatsApp, Instagram, RCS, mobile app and email, and ElevateOS takes ad variants to Meta, Google and TikTok.

Chat is the channel retailers most often skip. W for Woman ran virtual try-on entirely inside WhatsApp, with no new app and no microsite. The shopper picks a piece and Snype returns an on-brand image of them wearing it, generated on demand, with an 86% image-generation success rate. The result was a 4× engagement uplift and 65% WhatsApp engagement. We covered how that flow works in Virtual Try-On on WhatsApp for Fashion Retail.

What results has personalized product imagery delivered?

The clearest result is bigbasket’s. Snype split the shopper base into cohorts, drew imagery for each cohort from the catalogue to match what that group buys, and generated the message and offer per shopper across channels in a single run. Funnel completion rose 42%, repeat orders 31% and basket size 26%.

The case page reads the three numbers separately. When the imagery matched what a cohort buys, shoppers hesitated less mid-funnel. Retention timed to each cohort’s rhythm brought them back. Right-fit offers grew the basket. Measure the three separately in your own test too, against a control group that keeps the generic imagery, because one can move without the others.

To see your own catalogue mapped onto per-shopper imagery, book an e-commerce growth assessment.