Lotte Department Store has put an internal merchandising system, Brand AI, into day-to-day use, folding data on more than 4,000 brands into one place for tenant performance, new-brand scouting and decisions on store entry or exit.
According to Sedaily, the tool is already live in merchandising work. A 3D network graph treats brands as linked nodes, including purchase correlation. Brand DNA lines a target brand up against as many as five peers on a radar-style view. An AI Report dashboard pulls customer and sales measures together; AI Insights combines several data points to flag patterns for merchants.
That mix of signals matches how department stores actually allocate scarce floors. Ranked sales are rarely enough on their own. Teams also weigh customer age, cross-shopping, category overlap, traffic and whether a name strengthens a floor’s identity. Linking those variables can show relationships that sit poorly in separate spreadsheets.
Lotte says staff are using the system in ordinary work. Within a month of launch it reported a 70% return-user rate, and internal surveys said productivity on comparable tasks more than doubled versus earlier methods. Those adoption and productivity numbers are the company’s own.
A conversational layer is next. Lotte plans a chatbot in the second half of the year so employees can ask for analyses in natural language—for example, brands ranked by the share of customers in their twenties. It also plans real-time web search via an API, combining outside trend information with the commercial data already in the system.
The more durable piece is the data model under that interface: how similarity, influence, customer overlap and performance are defined. Once teams trust the scores, those definitions can shape which brands get attention before a lease conversation starts.
Tenant choices still sit with people. Leases, strategic commitments, new concepts with little history and local knowledge will not all be in the data, and a model trained on what has worked can also prefer familiar patterns over genuinely new names. Brand AI is therefore better read as a shared evidence layer—faster assembly of the facts, visible assumptions, one picture of the portfolio—than as an automated merchant.
Department-store merchandising remains a judgment call. Lotte is not taking that judgment away. It is attaching a graph, a data model and, later, a chat interface. Whether that widens what merchants can see, without shrinking what they will try, is the longer test.