Rebuilding a payments platform while it's live

A card machine has one job, and it has to do it while the customer is standing at the counter. It has to be fast, and the connection has to hold. For the merchant behind that counter, it's their living.

In September 2022 iKhokha launched the iK Flyer, its first Android smart card machine, and I was Product Owner on the proof of concept. In the launch release I said:

We started with a POC with this new Android-based hardware and set out to work on the most important features: speed and reliability. We then ran a trial on the product with trusted merchants who we know would give us good feedback in terms of the transaction and connectivity reliability.

Later came the bigger job. We rebuilt the merchant platform while it was still live, moving it from a monolith into microservices so the app, the card machines and the dashboard would finally work together. Inside iKhokha it was called Project SWARM, and the CEO later wrote that it probably defined 2025 more than anything else. Merchants kept taking payments the whole time, and the dates didn't move.

Not every release went smoothly. When a rushed one went out, complaints spiked, and it was the support team who took the calls. So I turned my attention to them. I wrote them detailed training and opened a direct Slack channel, so they could get an answer while the problem was still happening.

Then I owned the last phase of the migration, end to end: connectivity add-on billing and subscription management. Until then, a merchant's connectivity had been tied tightly to buying the card machine. We linked it to actual usage instead. For the first time, support, merchants and our fin ops team could all see that detail, and fin ops had a more accurate way to measure what connectivity cost.

This time I staged the change so it reached support first, roughly 30 to 50 people, in smaller steps, before it went any wider. We co-built a support portal with them, shaped around the queries they actually fielded, reviewed it often, and changed the plan sprint by sprint until switch-over. Earlier releases had sent the calls escalating fast. This one landed quietly.

What stays with me from that phase is pride. Pride in the quality of the work, and in the support teams who monitored the release in the field and kept telling us how it was performing.

In 2025 Nedbank bought iKhokha, and in its 2025 results it named iKhokha's technology and more than 54,000 card machines as part of why.

If I did it all again, I'd agree the guardrails and the quality bar before anyone fixed the dates. I'd give support a seat from the very first sprint. And I'd make sure the people absorbing the change were ready before it reached them, which is exactly what that last phase taught me.