The same customer buys on a marketplace today, through social media next week, and walks into the shop the month after. And the back-office system counts them as different people. Everything built on top of that is wrong as a result: the segmentation, the customer value figures, and the promotions that go out. This project merged those identities into a single record, then built segmentation, per-channel analysis and person-by-person product recommendations on top of it. ThinkVerse AI delivered it as an AI Customer Data Platform / CRM, used entirely in Thai.
Built for
Brands and retailers selling through several channels at once that still cannot say where the same customer has bought again
Merging several identities into one customer
At the heart of the system is the Golden Record: establishing that several buyer entries from different channels are in fact the same person. The search-and-link screen finds buyers by name, phone number, membership ID or address, and lets several entries be selected and merged into one record. AI proposes the matches first and a person makes the final call, because merging the wrong two records does more damage than leaving them apart. Buyers that cannot be matched yet are held in a separate list to be worked through, rather than disappearing quietly into the database.
- Search buyers by name, phone number, membership ID or address
- Select several buyers and merge them into a single record
- AI-suggested matches, confirmed by a person
- A list of buyers not yet linked, to work through
The customer base and every channel on one screen
Once the data is merged, the dashboard summarises the state of the membership base at the top: total members, members still active, new members that month, churn rate and growth rate. Below that come segments by product category with the average order value of each, a view that often reveals that a smaller segment carries a far higher value per order. The channel-performance screen separates the number of buyers and the average order value for each platform, and counts buyers who have bought on more than one platform as a figure of its own.
- Membership metrics: total, share still active, churn rate and growth rate
- Segments by product category, with average order value
- Channel-by-channel performance: marketplace, social and in-store
- Buyers who purchase on more than one platform counted separately
Segment by behaviour, then recommend person by person
Customers are placed into behavioural segments on RFM lines: champions, cannot-lose-them, loyal, new customers, about to lapse, at risk, hibernating, and lost. The marketing team can then choose who to talk to and with what message. The recommendation screen proposes products drawn from that customer's own purchase history, each carrying a match score, so the sales team sees the reasoning behind a suggestion and not just the result, and can put follow-up questions about the customer to the AI in Thai on the same screen.
- RFM segmentation by purchase behaviour
- Segment filters on a single screen
- Product recommendations from purchase history, each with a match score
- Ask the AI about a customer in Thai