So you're ready to replatform your ecommerce website, but aren't sure where to start. I hear you. This question comes up all the time. As always, the right answer depends on priority and context. However, there is an underlying best practice I believe can apply to every use case, and that is to move incrementally and avoid making a big jump at once.
When moving to a composable stack, the strangler pattern is often a good choice. It's important to remember that composable begins with two components and that you're embarking on a journey, not a project, when getting started. The process involves splitting your architecture into multiple domains, replacing each domain with a dedicated tool or micro service, and iterating until the monolith has been completely "strangled" or reduced to one component.
Even though this feature-by-feature approach is very reasonable, if you're selling internationally, there's another strategy, which is to move country-by-country instead. In my career, I have seen this pattern applied multiple times, and it is becoming the de facto standard for all international brands.
In this approach, you identify one of your countries, and implement an entirely new stack for that country, keeping the others on the old monolith. In essence, the country selector becomes your router. It is possible to host the new country site on a different subdomain or tld. Alternatively, you can map a country/language subpath to the new country instance using a load balancer.
Selecting a country that represents only a small portion of your GMV will reduce the business impact should anything go wrong during the transition. Once the new stack is live in the selected country, you can measure the new conversion rate and other KPIs to ensure it performs better than the previous site.
In case something does not work as expected, you can quickly adapt, identify any issues, and fix them before rolling out the new stack to more countries. If all else fails, you can always roll back to the old platform and resolve any issues offline before trying again.
For a more sophisticated approach, you can also consider A/B testing the new stack versus the old platform within the same country. This will allow you to validate the effectiveness of the new solution with real data before rolling it out.
Compared to a feature-by-feature approach—which can be considered a horizontal split of your platform—a country-by-country rollout can be seen as a vertical split, which can be easier to digest and implement. There is also a possibility that the first country could be an entirely new market to you. By doing so, not only do you reduce the risk to zero, but also leverage the new platform to open up a market that wasn't served before, hitting two birds with one stone.