The short answer
- Migrate customers, artwork and open orders. Everything else is optional
- Run both systems in parallel for a fortnight rather than switching overnight
- Move your ten highest-value customers first, not your easiest ones
- Historic orders are usually better exported to a file than imported
- Redirect every old URL. A migration that loses your search rankings has cost more than it saved
Changing the software your shop runs on — whether between platforms or from a general cart to a dedicated one — is not a technical project. It is an operational one, and the failures are almost never about data — they are about trying to do it all on a Monday. A migration that takes six calm weeks costs less than one that takes a chaotic fortnight and loses two customers.
What actually has to move?
Less than you think. There are three categories, and only the first is genuinely non-negotiable.
Customers and their artwork. This is the asset. Years of logos, saved designs and contact records are what make a repeat order easy, and rebuilding them is impossible rather than merely slow. Get a real export before you commit to anything — a sample file you can open, not a promise that one exists. If artwork only lives inside the old platform in a proprietary form, that is the single most important thing to solve.
Open orders. Anything in production or awaiting approval has to exist in the new system or be finished in the old one. Deciding which, per order, is a morning's work and prevents a job falling into the gap.
Historic orders. Optional, and usually best left. You need them to be readable, not to be live records. A clean CSV export kept somewhere findable answers every question anyone will actually ask, and importing five years of orders into a new system to be searched twice a year is rarely worth what it costs.
Migrate what you will use on Monday. Archive what you will only ever look up.
What order keeps you trading?
Sequence matters more than speed. The aim is that at no point does an incoming order have nowhere to go.
| Stage | Do this | Roughly |
|---|---|---|
| 1 | Export everything from the old system and open the files | Week 1 |
| 2 | Build the new catalogue and pricing; place test orders | Weeks 1–2 |
| 3 | Import customers and artwork; check a sample by hand | Week 2 |
| 4 | Move your ten highest-value customers, and watch closely | Week 3 |
| 5 | Run both systems in parallel; new orders go to the new one | Weeks 3–4 |
| 6 | Finish remaining open orders in the old system | Week 4 |
| 7 | Redirect URLs, re-check every storefront, close the old system | Week 5 |
Stage four is counter-intuitive and it is the important one. The instinct is to move easy customers first because they are low risk. That is exactly why it teaches you nothing.
Your most valuable accounts have the most complicated requirements — the locked company stores, the odd pricing, the artwork nobody else has to deal with. If the new system handles them, it handles everything. Find that out in week three, not week six.
What should you deliberately leave behind?
A migration is the only realistic chance you get to clean up, because everything is being touched anyway. Take it.
Worth not bringing across
- Products you have not sold in two years — they are a maintenance cost, not a catalogue
- Customer records with no order and no valid contact detail
- Pricing rules nobody can explain the origin of
- Discount codes that were meant to expire
- Duplicate customer records — merge them before the import, never after
That last one is worth doing properly. Duplicates are far easier to merge in a spreadsheet than in a live system, and a customer who exists three times will get three different price levels and eventually notice.
What usually goes wrong?
Four failures account for most bad migrations, and all four are avoidable by knowing about them in advance.
The export is discovered to be incomplete after the old system is closed. This is the unrecoverable one. Export early, open every file, and check that artwork is actually included rather than referenced by a link that will stop working.
Pricing rules are rebuilt from memory. Somebody reconstructs the quantity breaks from what they remember, and a customer notices three months later that their price changed. Export the price list as data and reconcile it line by line.
Nobody owns the switch. Migrations run by everyone in the gaps between orders do not finish. One person needs the schedule and the authority to say when the old system closes.
Training happens after go-live. Staff learning a new system while serving customers will revert to the old one wherever they can, and you end up running both indefinitely. Train on the test data in week two, before anything real depends on it.
What has to happen the week before you switch?
Two things, and the second is the one people forget until it is too late.
Tell customers, once, plainly. Not a change management announcement — a short note saying the ordering site is changing on a date, what stays the same, and who to call if something looks wrong. Send it to everyone who has ordered in the past year.
Map every URL. If your storefront has been live for years, its pages are in search results and in other people's bookmarks and emails. Every old address needs a permanent redirect to its new equivalent. A migration that quietly drops that work loses traffic it took years to earn, and the loss shows up two months later when nobody connects it to the switch.
If you are moving to us, the migration and the back-office setup are part of onboarding rather than something you do alone. Before any of this, make sure you know what leaving the new system would look like. Asking for a sample export during the sales process is the cheapest insurance available, and it is one of the questions a demo will not answer unless you ask it. The rest of the software archive assumes you are choosing deliberately rather than under pressure.



