Customer Success Story

5+
uses, 1 tag set
90K
outfit combinations regenerated as the catalog turns
3-5
complete outfits for different occasions, each passing a styling check
0 hrs
manual styling where every look was previously built by hand
AZA Group runs Renee, Born2be and Vices out of Warsaw. It sells about 115,000 SKUs, mainly in Poland and Romania, and handled 10 million orders. AZA builds most of its own technology and runs its own photo studios, so it doesn't outsource lightly. But two problems kept growing with the catalog.
First, generic fashion tags didn't describe AZA's products the way its teams sold them. Situation, season and occasion were missing, and search, filters and campaigns all suffered.
Second, outfits were built by hand. With 115,000 SKUs and new arrivals every week, only a handful of products ever got a hand-curated outfit, and almost none had Complete the Look recommendations.
AZA started with tagging. Four to six months later it came back with the harder problem: outfits for every product, tailored to the taste of their fashion-forward shoppers, not just matched by category or color.
90,000+ outfits generated and regenerated automatically as stock changes.
Multiple looks per product, each checked for vibe, occasion and silhouette, ready for product pages and personalized marketing.
Custom tagging built for many uses: occasion, season and category taxonomies tailored to Renee and Born2be now power search, filters, collections, outfits and campaign targeting.
Standard tags describe what a product is. AZA also needed to know when and why shoppers buy it. YesPlz customized the tagging like occasion, season and category taxonomies for Renee and Born2be, so one set of tags serves every team: search and filters for shoppers, collections for merchandisers, targeting for campaigns, and outfits for Complete the Look.
Here, the AI reads a dress as glam and made for a night out, and those tags shape every look built around it.
Project 2: AI outfit recommendations
After tagging went live, AZA came back for outfits. Built on AI Complete the Look, YesPlz now generates multiple ready-made outfits for every one of 30,000 products, each styled for Renee and Born2be shoppers.

A good outfit is more than matching categories and colors. It has to fit the persona, lifestyle and taste of Renee and Born2be shoppers. YesPlz AI scores every look for vibe, occasion and silhouette, then works hand in hand with AZA's merchandising team: they review, fix and rate looks, and each fix teaches the AI what their customers actually wear.

AI evaluates: each outfit is checked for vibe, occasion and silhouette match before it ships.
Merchandiser moderates: swap or add any piece, from AI-suggested alternatives or catalog search.
Feedback trains the AI: a rating, what needed fixing (here, material and occasion), and a one-line reason the AI learns from.


How it actually works: four stagesRepresent: visual embeddings capture how a garment actually looks, alongside the custom tags from project1. Tags narrow the field; they don’t decide what goes with what.
Generate: outfits are built across thousands of situations, not a handful of templates.
Evaluate: every look is scored for vibe, occasion and silhouette before it can ship. This is the step a stylist does last and a rule engine never does at all.
Personalize: Renee and Born2be’s merchandisers review and rate looks, and shopper behavior tunes them toward the personas and lifestyles in AZA’s traffic.
Custom Product Tagging: tags built for AZA's catalog.
Before: Generic attributes: basic tags like category, color and length, with no sense of when or where to wear it.
After: AZA's own taxonomy: the same product tagged with Renee and Born2be's own occasion, season and category taxonomy, ready for search, filters, collections and campaigns.
AI-Powered Curation: Complete the Look
Before: Almost no outfit recommendations: only a handful of products had a look, each curated by hand.
After: Multiple ready-made looks on every product: 30K products, 90K+ outfits, refreshed as the catalog changes.
Merchandising Workflow: styling tuned to their shoppers' taste
Before: Every look built by hand: so only a few products ever got one.
After: AI builds, merchandisers refine: Renee and Born2be's merchandisers review looks against their shoppers' persona, lifestyle and taste, and every fix trains the AI.
Get the attributes right first. Custom tagging made the outfit project possible, and the same tags now feed search, filters and campaigns. Get the data layer right and everything built on it gets easier.
Outfits need taste, not just tags. Matching attributes alone gives safe, repetitive looks. Merchandiser feedback makes them right for Renee and Born2be shoppers specifically.
Fashion AI (tagging, visual embeddings, outfit styling) is specialist work. YesPlz plugs into your stack by API, so your team keeps control of the experience.
Yes. For AZA, we customized the occasion, season and category taxonomies for Renee and Born2be, so the tags fit how their teams search, filter, build collections and target campaigns.
Visual embeddings and tags generate looks across thousands of situations. Before shoppers see a look, the AI cross-checks it on every styling dimension a stylist would: vibe, occasion, season, silhouette, color, pattern, material and product mix. Merchandisers then review and rate looks, and their feedback sharpens those checks over time.
Outfits regenerate automatically as products are added or sell out.
Yes. The same attributes feed search, filters, campaign targeting and personalization.