is a tool that lets you perform AB testing. Like all or most other client-side AB testing , it uses the following mechanism: Optimizely tools A script tag is included on your page which alters it on load for a variation group of your users The variation is compared to the unchanged control group to get insights on how well it performed The main sell of this model for our team is simple: no code changes are needed to run the tests. This lets product and marketing teams launch tests in production faster and without involvement of engineers. There is one problem however: Optimizely usage is inherently problematic with React apps. React re-renders components every time state or store are changed. So, Optimizely changes don’t stick — they are removed by React diffing mechanism on re-render. What’s already out there? There has lately been a number of articles that introduced different solutions for React+Optimizely usage. engineering team designed using Store to hold experiment information. Looks like a well structured approach for client side rendering where big changes are necessary (like removing/adding components). [UPDATE: Tilt blog seems to have since been taken down] Tilt a pretty intricate system ’s idea is less complex, easier to implement, and is querying Optimizely’s globally available Data Object directly to get active experiments. Read about it . Shesh here talks about a different angle to the problem — . Not just React, but any server side AB testing will pretty much always require a deploy. To me though, this approach seems to under-utilize Optimizely — I would rather use its GUI for client-side AB testing or not use it at all. Milan server-side rendered apps and using Optimizely REST API to fetch experiment information on the server A lot of good ideas here. However, it bothers me quite a lot that all of the above solutions require a deploy for every new AB test. React/Redux + Optimizely without deploy or code changes Here is a lightweight approach. Things to note: This is for small to medium UX tweaks. If your AB test is massive, then a code change might be a better path. The success of this method in your codebase depends on your specific React implementation. This works well with our React/Redux app, because it re-renders the top level component on every change to props and actions are used predominantly (rather than local states). Step 1 React re-renders components when state or store are changed. If you use React, you are definitely familiar with the . You can use several of these methods to call a global function which acts like a hook to trigger the for desired DOM alteration. lifecycle methods JavaScript What this looks like: would be a global function that gets called immediately after the component mounts on the page. Utilize your top level component’s and lifecycle methods, like so: window.optimizelyHook componentDidMount componentDidUpdate Step 2 In Optimizely’s variation JavaScript, define the function. Then, perform all DOM alteration inside of this function. optimizelyHook() is called once on the initial mount is called on every consecutive update of props componentDidMount componentDidUpdate The result will run on every props update, resulting in experiment that persists through re-renders triggered by Redux actions. optimizelyHook() Deploy once, and use many times across multiple AB tests. 🎉 Resources Understanding server-side vs client-side AB testing React lifecycle methods
