What is the best approach to test conversion rates between three different website designs deployed on Compute Engine?

Using App Engine with traffic splitting is an effective approach for testing conversion rates across different website designs for several reasons.

App Engine’s built-in traffic splitting feature allows developers to distribute incoming requests among multiple versions of an application seamlessly. This means that you can deploy all three different designs as different versions in App Engine, and then use traffic splitting to direct a specific percentage of your users to each design. By doing this, you can accurately measure how each design performs in terms of conversion rates in real time without needing to set up complex routing logic or additional infrastructure.

Additionally, App Engine is designed for scalability and automatic handling of workload variations, which is ideal for testing different versions under real-world traffic conditions. This facilitates obtaining reliable data on user interactions across the different designs, making the comparisons fair and manageable.

The other options offer valid deployment methods, but they lack the seamless integration of traffic management that App Engine provides. Deploying the website as separate services or using Cloud Functions can increase complexity and make it more challenging to ensure a fair distribution of traffic among the designs. Thus, App Engine with traffic splitting stands out as the most efficient and user-friendly method for conducting such an experiment.