What is identity merging, and how can it go wrong?

Identity stitching links records from different sources to a single person based on rules you define.

The key determines the accuracy

Email is most common. Phone numbers and names plus addresses are used as supporting information. If you have a shared customer number across your systems, that is clearly the best option.

Rules that are too lax pose the greatest risk

If you merge records based on name alone, two people named Ole Hansen will become one. One person will then be able to see the other’s purchase history, which is a privacy breach.

Overly strict rules split

If you require an exact match on everything, you’ll end up with three profiles for the same person. Less serious, but it undermines the whole point of consolidating them.

Check the result manually

Pull out a hundred merged profiles and review them. The errors are obvious when you look at them, but invisible in the statistics.