Repeat-visit analysis cannot rely on ticket numbers alone. Customer, actual product, and service event must be linked correctly in persistent operational memory.
A durable customer identity is required. Practical match fields such as phone numbers can be used, but the system needs a stable customer ID so number changes and duplicate records can be managed.
Product identity comes from serial number and model code together. Different products for the same customer must be tracked separately.
The service event sits under the product — ticket number, complaint, diagnosis, parts, technician, duration, outcome, and customer feedback belong to that event.
With this structure, the model can learn recurring complaints on the same product, part risks by model, and first-time fix performance by team.