By Lotte
842 products. 20+ brands.
One gallery that navigated them all on visual terms.
How By Lotte gave shoppers a rec experience that surfaced the most visually precise match from across the full edit - regardless of brand.
→ How cross-brand visual similarity gave shoppers access to the full By Lotte edit - not just the brand they landed on
RecoMelon installs in under a day. First influenced orders in 60min.
No code, no engineering, no catalogue prep.
Message our team on WhatsApp to see how RecoMelon fits your store. Or, run a free PDP audit - see your rec relevance score.
- The situation
842 products. 20+ brands. A rec module that showed more of the same brand when the best match might be from a different one.
By Lotte is not a single-brand store. It is a curated concept boutique - Forte_Forte alongside Tibi, Röhe alongside Nili Lotan, BYLOTTE Studios alongside Wandler. Every product on the site was chosen by a buyer who understood how it related to everything else in the edit. The selection is the product. The curation is what the customer is paying for.
The problem with a standard recommendation module in this context is specific: it respects brand boundaries that the buyer didn't intend. A shopper on a Forte_Forte PDP gets shown more Forte_Forte. A shopper on a Tibi piece gets shown more Tibi. This might feel logical - same brand, same aesthetic - but in a concept store it is often exactly wrong. The buyer stocked Röhe and Forte_Forte because they share a visual world. The rec module had no concept of that relationship.
What By Lotte needed was a rec gallery that understood visual similarity across the full 20+ brand catalogue - showing a shopper the piece that looks most like what she is considering, regardless of which brand made it. Not a different aesthetic. Not a complementary piece. The same visual register, better surfaced.
"The best match for a Forte_Forte piece might be from Röhe. The rec module only ever showed her more Forte_Forte."
- The category dynamics
Fitted. Crew-neck. Long-sleeves. By Lotte shoppers navigate by construction - across every brand in the edit.
Most fashion recommendation problems are about showing the wrong aesthetic. By Lotte's problem is structural: a rec module that treats brand as the primary axis of similarity in a store where the buyer's logic is visual and cross-brand.
The Dealbreaker™ data confirms how By Lotte shoppers actually navigate. The top filter terms - Fitted, Crew-neck, Buttons, Long-Sleeves, Cargo - are garment construction attributes. Not brand names. Not price filters. Silhouette descriptors. The shopper who filters for Fitted is not looking for a specific brand's fitted piece. She is looking for the fitted piece in the whole By Lotte edit that most precisely matches what she has in mind - regardless of who made it.
This is the logic of discovery in a curated multi-brand context. The shopper trusts the buyer's eye to have assembled a coherent visual world. RecoMelon's job is to navigate that world on visual terms - surfacing the closest visual match from across the full 1,560-product selection, not just from the brand on the current PDP.
- The intervention
Visual similarity across all 20+ brands. The full edit, navigated on visual terms.
RecoMelon mapped the entire By Lotte catalogue visually - across all brands and price points. Not by brand proximity, not by purchase history, not by category rules. By what the products actually look like: silhouette, construction type, drape, colour register and tonal weight.
A shopper on a Forte_Forte blouse PDP sees the pieces from across the full edit that share its visual DNA - including Röhe, Tibi or Nili Lotan pieces in the same register. The rec gallery surfaces the closest visual matches from the entire buyer's selection, not just the brand she happened to land on.
Dealbreaker™ filters let shoppers refine by the construction attributes that actually govern their navigation: Fitted or relaxed. Crew-neck or open. Long-sleeve or sleeveless. One interaction narrows 1,560 products to the silhouette type the shopper is working with - from any brand in the edit.
The result: a rec gallery that navigates the By Lotte catalogue the way the buyer assembled it - as a visual edit, not a set of separate brand inventories.
- The outcome
+33.1% AOV. 104 influenced orders. And 34 recs per session - the full edit, finally explored.
These are the signals that matter - the commercial proof that cross-brand visual similarity changes how shoppers navigate a curated multi-brand edit.
Executive interpretation
The +33.1% AOV delta - €375.92 to €500.41, a €124.49 lift - is the largest in the RecoMelon portfolio. Shoppers who engaged with rec-influenced sessions were spending more - not because they added a different category of product, but because the gallery surfaced the precise visual match that warranted the spend. The 34 recs per session (up from 11) is the behavioural proof: shoppers are moving through the full visual selection systematically, exploring what the buyer assembled across all 20+ brands before making a decision.
- Dealbreaker™ intelligence
Fitted. Cargo. Crew-neck. The filters show how By Lotte shoppers navigate construction - across the whole edit.
49 Dealbreaker™ filter interactions in the reporting period. The top terms - Fitted, Crew-neck, Buttons, Long-Sleeves, Cargo - are garment construction attributes. Not brand names, not price filters, not colour. Silhouette and construction type. The shopper who filters for Fitted is narrowing the entire 1,560-product selection to one silhouette register - and the result will include the most visually similar pieces from every brand in the edit that matches that description.
This is the precise mechanism that makes visual similarity useful in a multi-brand context. Dealbreaker™ doesn't just filter - it translates the shopper's construction intent into a cross-brand visual search. The result is the By Lotte buyer's selection, navigated exactly as she assembled it: by how things look, not who made them.
The Colour Crush data reinforces the visual coherence of By Lotte's edit. The warm, dark palette - burgundy, brown, amber, black - runs across the entire brand selection. When RecoMelon surfaces visually similar pieces across brands, it is working within a colour world that the buyer deliberately assembled. The signal is consistent and the gallery can navigate it precisely.
- The 34-rec session
34 recommendations per session. The full edit, explored before deciding.
The AVG recs per session metric - 34, up from 11 in the prior period, a +208% increase - is the most distinctive number in the By Lotte data set. It tells a specific story about what happens when a rec gallery stops being brand-siloed and starts reflecting the full visual scope of a curated concept store.
A shopper exploring 34 recommendations before making a decision is not confused. She is thorough. She arrived at By Lotte precisely because it stocks multiple brands in the same visual world - and for the first time, the rec gallery was making that breadth available to her in a systematic, visually coherent way. The 34-rec session is the evidence that shoppers were using the gallery exactly as intended: to find the most visually precise match from the full buyer's selection, not just the brand on the current PDP.
- Strategic takeaway
The curation was already there. The rec module just needed to read it on visual terms.
By Lotte's commercial opportunity was not hidden. It was encoded in the buying decisions the team has made for years - which brands to stock, how they relate to each other visually, what the whole edit looks like as a coherent world. The buyer's logic is that Forte_Forte and Röhe and Nili Lotan share visual DNA. The rec module's logic had been that they were entirely separate inventories.
The +33.1% AOV is what happens when those two logics are finally aligned. A shopper who trusted By Lotte's buyer to have assembled a coherent visual world found, for the first time, a rec gallery that was actually navigating it - surfacing the most visually precise alternatives from the entire selection, not just the brand she happened to land on.
The question for any multi-brand boutique or concept store is not whether your buyer has assembled a coherent visual world. They have. The question is: is your rec module navigating it - or is it treating your curated edit as a set of separate brand inventories that happen to share a domain?
- Implementation
No tagging. No brand rules. The visual relationships were already in the catalogue.
The cross-brand visual mapping, the construction-attribute Dealbreaker™ filters, the colour register intelligence - none of it required the By Lotte team to tag products, write brand-pairing rules or manually curate galleries. RecoMelon read the visual relationships across the full edit from product imagery alone. The curation intelligence was already there. RecoMelon surfaced it.
View more case studies
Shopify brands use RecoMelon to fix bounce, grow AOV, and turn PDPs into high-converting discovery tools. Here’s how.
