Migrating from Salesforce to a "modern" platform. Vendor's data migration team gave us the usual "seamless, automated" spiel. Their pre-migration report showed a 99.8% field match. Sounded great.
We ran our own CSV diff on the sample dataset. Found the 0.2%:
* "Customer Since" dates shifted by one day for 30% of records (timezone "correction" they failed to mention).
* All historical notes appended to a single field, truncating after 255 chars.
* Their "mapped" custom picklist field was actually populating a text field, which would have broken our workflows.
The contract's liability clause capped their responsibility at the cost of the migration service—a fraction of the business impact. We forced a re-scope and got a fixed-price correction addendum *before* proceeding.
The lesson wasn't about the diff tool. It was about never trusting a vendor's pre-flight report. Always validate the sample yourself. Their definition of "success" is rarely yours.
Read the contract
Oh man, date/timezone issues are the absolute worst. They're such a silent killer because they often don't show up until a quarterly report looks funny, and by then you're debugging live data.
Your point about vendor definitions of "success" is spot on. I've seen similar gaps where a field "matches" because the data type is string and the values are strings, but the semantic meaning is completely broken. A "status" field that goes from Active/Inactive to "1"/"0" is a match to their tool, but it's a hard fail for our business logic.
Getting that fixed-price correction addendum before proceeding was the real win. It changes the entire incentive structure. Otherwise you're just paying them to troubleshoot their own mistakes.
terraform and chill