Before any data move, I ask one question: how many customers do you actually have? A 25-person industrial fastener distributor said “about 15,000.” The real number was 6,200. The rest were duplicates, dead accounts, and one salesperson’s eleven creative spellings of the same buyer.
Nobody did anything wrong. Every system collects duplicates the way a garage collects paint cans. The problem is what happens when you carry all of it into a new one.
Why 8,400 extra records is normal, not a scandal
Accounts get created at the counter when someone is in a hurry. A customer changes names after an acquisition and keeps the old number. A credit hold gets resolved by creating a fresh account instead of fixing the old one.
Move that mess as-is and your new system will simply be a faster way to be confused.
Seven questions to settle before you move a single record
- Who owns each data set — customers, vendors, items, open orders — by name, not by department?
- What makes two records “the same”? Match on tax ID or a normalized address, not on company name.
- Which records will nobody miss? Define it plainly: no order in 36 months and no open balance.
- What do you do with historical transactions when the account gets merged?
- How will you handle the 200 records that are genuinely ambiguous, and who breaks the tie?
- What is the cutoff date, and does everyone know that nothing gets entered after it?
- How many times will you run a full test load before the real one? Three is the minimum.
The rule that saves the most pain
Freeze the data before you clean it. Pick a Friday, stop new account creation entirely, and clean what exists. Cleaning while people keep adding makes the job twice as long and never quite finished.
Second rule: the people who use the data clean the data. A 62-person distribution team I worked with assigned customer cleanup to the two order-entry clerks who knew every account by voice. They finished 6,200 accounts in nine working days.
An outside team can write the match rules and run the loads. That is legitimate ERP Consulting work. But the judgment calls — is this the same company, is this account worth keeping — belong to the people who talk to those customers every week. Every ERP Consulting project I have seen fail at migration failed on that handoff, not on the technology.
This week, export your customer list, sort by name, and count how many times the same company shows up. That number is your real project scope.