Lifting a business-critical core to the cloud, starting with what could move

What changed

• Migration approach
  Phased, not a single big-bang replacement
• Impact on trading partners
  None (external connections unchanged)
• Historical data
  Carried over in full
• After the migration
  Infrastructure, monitoring, and operation stay with us

Challenge

Each day, sales are turned into loyalty-points data and sent to a trading partner over EDI. This processing, tied directly to the company's revenue, had run on Microsoft Access for years. It worked, but few people understood what was inside it, and when it stopped there was no way to find out afterwards what had happened. Replacing it had been discussed for some time, but because external connections were involved, nothing could be swapped out all at once — so nothing was.

Approach

1. 01 Started by reading the existing system and writing down what actually runs and how
2. 02 Decided what would move and what would stay before implementation, and agreed it as a plan
3. 03 Ran old and new in parallel and switched only once their output matched
4. 04 Mapped out the order in which the remaining parts can move later

Key success factors

Not rebuilding everything at once

The parts that connect to the outside stayed; the self-contained parts moved first. Trading partners were never involved, and the impact of the migration stayed on our client's side.

Never stopping the business

Old and new ran in parallel and the switch waited until their output matched. Nothing was riding on a single cutover day.

Not walking away at delivery

The resulting infrastructure is run as our own service, with monitoring and day-to-day operation on our side. We stay responsible for it continuing to run.

Leaving room for what comes next

We mapped out how the parts left behind can move later, so there is a next step rather than a dead end.

Solution

Rather than replacing everything in one go, we moved to the cloud starting with what could move. The parts that are self-contained went to the web; the parts that connect to the outside kept working exactly as before, in a hybrid setup. From the trading partner's side nothing changed, so we never had to ask them to accommodate our schedule. All historical data carried over, keeping reporting continuous. Old and new ran in parallel until their output matched, and only then did the switch happen — so the business never had to stop. The infrastructure that resulted is run as our own service, with monitoring and day-to-day operation continuing on our side.