Migrate your Excel data into your application: a guide without breakage

The application is ready; the historical file is waiting. Migration happens in four stages: clean in Excel, turn formulas into rules, import with a discrepancy report, and run both in parallel.

Migrating from Excel to an application takes 4 stages: clean the data in the file, turn formulas into approved rules, import with a discrepancy report accounting for every row, then run both tools in parallel over a full period. On those terms, zero rows lost and zero surprises in the first month.

What hides in your columns

A long-lived file accumulates invisible traps: dates entered as text, ‘March 2023’ beside ‘01/03/23’, customers duplicated with slightly different spellings, merged cells breaking sorts, amounts with and without VAT in the same column, and test rows never deleted. The application will reject these ambiguities: that is its reason for being. Migration is when years of tolerance are settled.

Step 1: clean in Excel, not afterwards

Make corrections in the tool you know. Deduplicate customers, unify date formats, separate things sharing a column and delete dead rows. Devote a proper session to it, with two people if possible: the person who knows the history spots the odd 2019 row in a second. Every anomaly corrected here is one the import will not turn into a mysterious rejection.

Step 2: formulas become rules

Your formulas are business rules in disguise: that column calculating discounts, the other flagging delays. Before any import, they must exist in French in the application's requirements, ‘the discount applies above €100, never to deposits’, and you must approve them. In Maestro, the file is attached to the project and agents extract these rules for your review; elsewhere, demand the same stage. A formula migrated without being understood reproduces its errors exactly.

Step 3: import with a discrepancy report

The non-negotiable principle: an import must account for itself. A report counting rows read, rows created and rejections, each with its reason and row number, makes the import verifiable; you correct rejections in the file and rerun until every row is accounted for. An import answering ‘done’ without details is an act of faith: reject it. Keep the original file untouched, dated and safe. It remains your evidence.

Step 4: the parallel period

Run both tools in parallel for a complete cycle: a month-end close or billing period. Enter data in the application, maintain the file as a mirror and compare totals at period end. Discrepancies are lessons: a poorly phrased rule or forgotten case, to correct in the requirements and then the application. Once a whole period ends without discrepancy, disconnect the mirror. The old file becomes read-only, available to consult forever.

Two refusals that protect you

Reject an import without a discrepancy report and a switchover without a parallel period. These two requirements each fit in one sentence in your requirements document, cost almost nothing to someone building properly, and make the difference between a migration and moving in the dark. Your history deserves this care: it is the numerical memory of your business.