A/B Testing with TD
Each commercial web-site or application has the embedded options of product purchase, sign up, conversion to paid version, subscription, etc. The amount of such actions, initiated by end-users (which, if being measured in percents, is called conversion rate), is highly dependent on app’s visual appeal, advertising texts and other marketing ploys used. Comparison of different application versions, aimed to designate which app version is better in confines of user’s conversion rate, is named A/B testing. Such procedure can help you to find out the best combination of attractions and other product characteristics, that will allow to receive the highest revenue.
With the help of Jelastic Traffic Distributor, you can easily apply such kind of testing to your projects - find the details below.
1. To perform A/B testing, you’ll need two different application examples to analyze, and any preferable type of monitoring to track the desired actions’ calling (e.g. when a user clicks on the required button).
- Routing method - select Sticky Sessions
- Traffic ratio - set 50 to 50 to get the fair analysisNote:
- DO NOT use Round Robin routing for A/B testing, as this procedure assumes provision of different content on your backends, so some project elements may not be available on a particular server upon the request.
- in case you have the already-run production project, learn how to smartly inject Traffic Distributor in front of it without interrupting app normal work
3. Everything is set up now, i.e. new incoming requests are evenly routed to both application versions with monitored conversion score. So just share your Traffic Distributor entry point link (i.e. environment domain or your custom one) among users and wait for some period of time for your customers to try out the given app versions and collect some data on conversions.
4. Finally, analyse the results you’ve gathered to choose which project branch is more preferable for usage. Obviously, environment with higher Conversion Rate will win the race and can be boldly used in production, while the other one can be removed or just modified for another testing.