Moving Data Into a New System: Codes, Duplicates, Balances
31 Oct 2026
Migration is not copying columns. It is one code, merged duplicates, an opening balance on a date a person signs, then a short parallel run before the cutover day. Old mess moved as it is becomes a balance people trust by mistake.
The new system opens on Monday with Sunday data. If that data is duplicate names, items with no code, and balances nobody counted, Monday is not a launch. It is the same book, faster. Migration is a piece of work with an owner and a date, not a last step on delivery night. This page is about the data itself, whether it comes from files or from an old program. The decision to leave a spreadsheet at all is a wider topic, and it has its own guide.
Start with a list of what moves and what does not. Move what you will transact after launch: a customer with open business, an item you sell or stock, an opening balance you agree. Do not move a test row, a customer with no movement for years unless you have a real reason to keep them, or a free-text note that is actually a complaint with no field. The archive can stay a read-only file outside the system. Forcing it into daily use makes search worse.
Codes before import
- One code per item, in the file and in the system.
- One code or key per customer, even if several phone numbers sit under that key after a merge.
- The unit is written: a piece, a carton, and the relationship if you need it.
- An empty column is not imported in the hope it will be filled later, if the field will be required on the next movement.
- A sample of fifty rows is reviewed by hand before the full file is imported.
A code that changes shape between branches, sometimes with leading zeros and sometimes without, will create two items. Agree the shape, write it down, and apply it to the file before any import tool. The program will not guess that two rows are one item because the names look close.
Duplicates
The same customer in three spellings: with a small difference in the letters, without it, and under the shop name rather than the legal name. Merging is a human decision by someone who knows the data, not a magic button. The output is a list: this row stays, these two fold into it, and the balance or the limit is checked after the merge, not only before. Keep the merge table. A month later someone will ask why a name disappeared, and the answer is in the table, not in memory.
Items work the same way. A wrong merge mixes two balances that have nothing to do with each other. If you are unsure, leave both rows and a visible exception rather than a rushed merge. An exception can be closed later. A bad merge sells stock you do not have.
The opening balance
- A cutover date you agree, with balances taken at the end of that date.
- A stock count or a reconciliation, not the last number in a file that nobody counted.
- A customer balance, debit or credit, as the accountant accepts it on that date.
- The name of a person who signs that this file is the opening position, and the date of the signature.
- Movements after the cutover date are not mixed into the opening file. They are entered as movements in the system, or held until launch day under a written rule.
A balance with no date is an orphan number. A week later you cannot tell whether it includes Monday invoice. A signature does not make the number true by itself. It creates a person who can be asked, and a file that can be reopened if a gap appears. Without that, the argument turns into a charge against the system when it is a charge against the file.
A parallel run, then the cutover day
For a short period you agree, often described as one or two weeks in moves of this kind, the team works in the new system and a sample is compared with the old method. The comparison is the point, not a ceremony. Choose what you compare: a count of invoices, the total of a sample, and the balance of items that move often. A difference is written the same day. If you spend the period typing everything twice and comparing nothing, you worked hard and tested nothing.
The cutover is a calendar date the warehouse, sales and accounts all know. After it, the old file is not edited. If editing stays open, work returns to the easier place and the system becomes a late copy. Reading the old file can stay allowed. Writing is closed. Someone who needs extra analysis exports from the system, and does not revive two living copies.
What to send before estimating the move
- Where the data lives: files, or the old program name and version.
- A rough count of items, customers and open documents, not a false promise of precision.
- Whether a code exists today, or only a name.
- Who will merge duplicates, and whether that person has days available, not scattered hours.
- Who signs the opening balance on the operations side and on the accounts side.
How migrations fail
Importing on launch night
The file arrives large, the columns differ from what was discussed, and the count never happened. Delay the launch. An extra day on the old tool costs less than a week of movements on a wrong balance. The scope includes time to clean. It does not assume the data is ready because someone said it is in a spreadsheet.
Copying old passwords and permissions as they are
A new system is a chance to review who sees prices and balances. Copying every old permission copies an old leak. A short list is enough: who posts a movement, who reads a report, who approves a count variance. The rest is added when a real job cannot be finished without it.
The decision to leave spreadsheets, with its signs and steps, is in replacing Excel with a business system. Carrying out the move sits in custom software development.
Describe the source of the data and who signs the balance via start your project. A sample file with sensitive names removed can be sent from the WhatsApp button under this article. It arrives tied to this page, and no phone number is printed here.
