Data & BI

Data migration plan: change systems without losing data

A data migration plan in seven steps: inventory, archive, field mapping, cleaning, trial imports, reconciliation and cut-over, with a checklist of key totals.

You are replacing a CRM, an ERP or an application that has run the business for years. The new system is chosen, the screens look right and a go-live date is pencilled in. Then someone asks the question that matters: what happens to our customers, invoices, open orders and balances? A data migration plan answers it before the first morning on the new system, not during it.

In this guide we walk through the data migration steps we follow, from the first inventory to the cut-over weekend. We also explain who signs off what and give you a reconciliation checklist to reuse.

Why a data migration needs its own plan

In most replacement projects, data appears as a single line: “migrate the data”. Our guide to implementing an ERP in an SME sums it up in a few lines, and our article on what custom software costs lists it as one of the factors behind the budget. Here we go into the detail: field mapping, reconciliation and cut-over.

Data deserves its own plan for three reasons:

  • Users notice it first. A missing customer or a wrong balance on day one costs more trust than a misplaced button.
  • Old systems hold unwritten rules. Status codes nobody documented, a “notes” field used for delivery instructions, a spreadsheet beside the application that holds the real prices.
  • The new system has its own structure. A migration is a translation, not a copy. Every field needs a destination and a rule.

In his previous roles, our founder, Anass Chabbab, led projects where the data had to follow the change. At Innova Home in Morocco (2016–2018), as a freelance engineer, he moved the business from paper and Excel to one order-to-invoice tool, which saved more than 50% of the time per order. That saving only holds when the information people rely on arrives complete in the new tool.

A word of proportion: if your old system holds a few hundred records, a carefully checked spreadsheet import may be all you need. The method below scales down; use only the steps your volume justifies.

The data migration steps, from inventory to cut-over

1. Take an inventory of your data

List every source, not just the old application’s database: the spreadsheets beside it, shared folders, mailboxes used as filing cabinets. For each source, note what it contains, roughly how many records, who uses it and who can explain each field.

2. Decide what to migrate and what to archive

Not everything needs to move. A common split:

  • Migrate: active customers and suppliers, products, open orders and quotes, unpaid invoices, current stock, opening balances.
  • Archive: closed history, kept in a read-only store that can still be searched and exported.

How long you must keep invoices, contracts or personal data depends on the country and the type of document; as of October 2026 there is no single rule that applies everywhere. This article gives practical guidance; it does not replace the advice of your accountant or lawyer on your retention duties. Personal data you no longer need is a good candidate for not being moved at all.

3. Map every field

The field mapping document is the heart of the plan. For each source field, it records the target field, the transformation rule, the default value when the source is empty and the person who approved the rule. Typical rules: status codes “A”, “I” and “X” become “Active”, “Inactive” and “Closed”; one “Name” field is split into first and last names; dates stored as text become real dates.

Some values will have no home in the new system. Decide what happens to them: a new custom field, a note, the archive. Never let them disappear silently.

4. Clean and deduplicate

The same customer often exists twice, under slightly different names. Mandatory fields are empty; postcodes are in the phone column. Clean in the source where you can, or in a staging area with written rules, never by hand in the final import, because a manual fix will not be repeated on the next run. The business, not the technical team, decides which duplicate wins. This clean-up is a one-off; keeping data reliable once it is in the new system takes rules at the point of entry and a short monthly routine.

5. Run trial imports

Plan at least two or three full trial imports into a test copy of the new system. Each run produces a log: records read, loaded, rejected, and why. Key users then check samples in the real screens. The idea is the same as parity testing: run the same cases through the old and the new system and compare the results.

6. Reconcile the key totals

Counting records is not enough: the totals must match too. The checklist in the next section lists what to compare.

7. Freeze, cut over and keep the old system readable

Agree a freeze point after which nothing new is entered in the old system, or a short, defined list of entries to re-key. Then run the final extraction, the final import and the reconciliation, and hold a go/no-go meeting against criteria agreed in advance. Keep the old system read-only for a set period, and write down how you would roll back if the reconciliation fails.

The data migration checklist: reconciling key totals

Use this checklist on every trial import and again at cut-over. Each line is signed off by the person who owns that data.

  • Record counts per entity: customers, suppliers, products, open orders, open invoices. Source records minus deliberate exclusions must equal records loaded plus records rejected.
  • Receivables and payables: total and per customer or supplier, ideally by ageing band.
  • Ledger balances at the cut-over date, account by account, if accounting moves too.
  • Stock quantities and values per site or warehouse.
  • Open documents: orders, quotes and contracts, with their amounts and statuses.
  • Samples checked by users: ordinary records and awkward ones (the oldest customer, several delivery addresses, credit notes).
  • Rejected records: each one fixed, reloaded or consciously excluded, with a reason.
  • Reports: the reports and the indicators on your executive dashboard give the same figures before and after.
  • Attached documents: the number of files linked matches, and a sample opens.
  • Users and access rights: each person sees what they saw before, and no more.

Illustrative example (fictitious figures). The old system holds 10,000 customers. The business decides to archive 1,500 that have been inactive for many years, so 8,500 are expected. The trial loads 8,480 and rejects 20 for a missing tax number: they are completed in the source and loaded on the next run. Receivables total 1,200,000 in the old system and 1,199,000 in the new one; the 1,000 gap is traced to a credit note mapped to the wrong document type, the rule is fixed, and the next run matches to the cent.

Who signs off what

Make the roles explicit:

  • Business owners (finance for balances, sales for customers, operations for stock) own the meaning of the data. They approve the mapping and cleaning rules and sign the reconciliation for their area.
  • Key users test samples during each trial import.
  • The technical team or provider writes the extraction and import scripts, keeps them repeatable and produces the logs. It does not decide which duplicate wins.
  • Your accountant, if accounting moves, confirms the opening balances.
  • The project sponsor makes the go/no-go decision at cut-over, against the agreed criteria.

Keep each sign-off in writing, with the run it refers to. When we support a migration as part of our data and BI work or an ERP project, this is the document we build first.

Frequently asked questions

Should we migrate all our historical data to the new ERP?

Rarely. Migrate what the business needs to work from day one: active records, open documents and opening balances. Older history can stay in a searchable, read-only archive, which keeps the migration smaller and easier to test. Check first how long you must keep each type of record in your country, and whether your teams need history inside the new system for comparisons, for example last year’s sales.

Can we migrate the data ourselves?

For a small volume and a simple structure, often yes: export, clean in a spreadsheet, import with the new system’s own tool, then reconcile the totals. Bring in help when there are several sources, complex transformations, accounting balances, or a business that cannot stop for long. Even then, your teams must own the rules and the sign-off; a provider cannot know what your data means.

What should we do with the old system after the cut-over?

Keep it read-only for an agreed period, so people can check a past record and you can roll back if a serious problem appears. Then retire it in a planned way: final archive export, retention rules confirmed, licences and hosting stopped. If the old system still serves other tasks, our modernisation page explains how we approach what comes next.

In short

A data migration plan turns “move the data” into seven visible steps: inventory, migrate or archive, field mapping, cleaning, trial imports, reconciliation and cut-over. The business owns the meaning of the data and signs off each area; the technical team makes every run repeatable. Reconcile counts and totals, not just records. And keep the old system readable until everyone is confident.

Replacing a CRM, an ERP or an old application and want your data to arrive complete? Book a free 30-minute discovery call on Google Meet: we look at your sources together and outline the migration steps your project needs.

Related servicesData & BIERPModernisation

Have a project in mind?

Let’s talk about it on a 30-minute discovery call, free and with no obligation.

Book a discovery call

Further reading

All articles