Godwit Data
Reconciliation · 6 min read

Anatomy of a migration evidence pack

What your compliance function actually needs after the data moves — line by line.

When a migration finishes, someone in compliance inherits a question they didn't ask for: how do we know it worked? "The project team says so" is not an answer they can file. Here is what is.

1. Row-level reconciliation, not samples

Every record in the source, matched to its counterpart in the target, with the match basis stated. Sampling was a reasonable compromise when reconciliation was manual. It isn't manual anymore. If your integrator proposes checking 5% of records, ask what they plan to say about the other 95% when a client complains three years from now.

2. Financial totals, to the penny

Sum the source, sum the target, show the difference is zero — per account, per portfolio, per book. Penny differences matter not because pennies matter, but because a penny out means a rounding rule differs somewhere, and rounding rules that differ on pennies differ on pounds eventually.

3. Exceptions with dispositions

Real migrations have exceptions. The evidence pack doesn't hide them — it lists every one, what caused it, what was done about it, and who signed off. Fourteen dispositioned exceptions is a strong result. Zero reported exceptions on a million records is a red flag.

4. The audit trail

Which mapping specification version produced this load? Which transform code? Run when, by whom, against which extract? Reproducibility is what separates evidence from assertion.

5. A signature

Someone accountable signs it. Then it goes in the file, and the question "how do we know it worked?" has a one-word answer: look.

Got a cutover date? Tell us the source system, the target, and the deadline — we'll tell you within 48 hours whether we can hit it and what the Assessment will cost.

freddie@godwit.uk