From a spreadsheet workbook to a single ledger
Twenty-eight months of records spread across spreadsheets, turned into a single ledger that can be regenerated without duplicating anything.
Tech stack
Context
The business kept its operational accounts in a spreadsheet workbook that grew over more than two years, with criteria that changed along the way and corrections written on top of the original data.
How it was done
- • Every migrated record keeps which sheet and which row it came from, so any figure can be traced back to its origin.
- • The process is re-runnable: running it again does not duplicate records, which makes it safe to fix and repeat.
- • Inconsistencies are not resolved by guessing: they are flagged for a person to decide.
The underlying decision
The data contained dates that do not exist in the calendar. The temptation is to push them to the nearest valid date and move on, because that way the process finishes clean. Not here: the impossible date is flagged for review, even if it leaves the report with open items.
What I learned
That a migration which corrects silently produces data that is clean and false, which is worse than data that is dirty and honest. The first gets believed; the second gets reviewed.