How to merge a customer book without losing data.

The steps

• Take an inventory: each system, export, and feed the business runs on.
• Score the data first: what is missing, duplicated, or in conflict.
• Write match rules a person can read, and fix conflicts by rule.
• Keep households together, not only individuals.
• Run the merge on a copy, and compare before and after by business keys, not random ids.
• Explain any difference, or do not ship it.

Three mistakes that cost the most

• One lookup per row. It makes a large import take minutes, and it can freeze the app while it runs. Load what you match against once, and look up in memory.
• Trusting deltas that cannot report removals. If an export only shows active records, a cancellation shows up only as a missing row. Use the full file, with a guard if too many records vanish at once.
• Testing on a small sample. Time the run at the real size of the book before anyone relies on it.

After the merge

A clean book drifts. A daily check that lists gaps and conflicts beside the fix keeps it clean. BigStarfish sells this as Glide Data: the Clean Merge, the Scorecard, and a daily data pull.

Common questions

How do you merge duplicate customers from multiple systems?

Score the data, write readable match rules, merge on a copy, keep households together, then compare before and after by business keys.

How do you prove a data migration changed nothing by accident?

Fingerprint the old and new books by business keys and compare them. The hashes match, or each difference is intended and explained.

Should I use delta files or full exports?

Use full exports when the source cannot report what left. A delta from an active-only export does not show a cancellation.