The example consists of an A/B test comparing the performance of two variants of a webpage. A NodeJS server serves these variants behind a feature flag using Split's NodeJS SDK. The webpages are built with Webpack and utilize Split Browser Suite (and internally Split RUM Agent) to capture performance metrics such as page load time and Web Vitals.

One of the webpage variants (treatment 'on') includes performance optimizations, while the other (treatment 'off') doesn't. For example, the 'on' variant loads Split Suite using dynamic imports and is built with Webpack's "production" mode, which minifies the code. In contrast, the 'off' variant loads Split Suite synchronously, blocking the page load, and is built with Webpack's "development" mode that doesn't minify the code by default.

An automation script navigates the page multiple times, generating events and impressions for both treatments.

This example assumes you have set up a feature flag in Split UI, with traffic type 'user' and two treatments: 'on' and 'off'. (The name of this feature flag should be provided to the FEATURE_FLAG_OPTIMIZE_PAGE variable in your .env file.)

1. Take a copy of .env.example and re-name to .env.
2. Add your Split SDK keys and feature flag names to .env. (Defaults will be used for any flags that are not created in Split UI or named in the .env file.)
3. Run npm install to install dependencies.
4. Run npm run serve to build the app and start the server. The Web page will be served at http://localhost:3000/?id=<user-id>, where <user-id> is a unique identifier for the user, used by the Split SDK to bucket the user into a treatment.
5. Run npm run automation to run the automation script. The script can take some time to complete, as it will generate events and impressions by navigating to the webpage multiple times with different user IDs, using Puppeteer and Chrome. You can grab a coffee. :)
6. Open the Split UI, click on a feature flag's Metrics impact tab, and analyze the metric results. You can click 'View more' on a Metric card to visualize metric measurements. See more about the "Metrics impact tab".

• /client: source code of the web application and its two variants.
  • /client/index-on.js: entry point for the optimized variant, served for treatment on of the feature flag identified by FEATURE_FLAG_OPTIMIZE_PAGE.
  • /client/index-off.js: entry point for the default variant, served for treatment off of the feature flag identified by FEATURE_FLAG_OPTIMIZE_PAGE.
• /webpack.config.js: Webpack configuration to build the two variants of the application.
• /dist: built static assets of the application, generated by Webpack.
• /server/index.js: source code of the NodeJS server that implements the endpoint GET /?id=<user-id> that serves the two variants of the application behind a feature flag.
• /automation.js: automation script to run Puppeteer, navigate to the application and generate events and impressions.

• npm run dev: starts Webpack with watch flag and NodeJS server with Nodemon, so that the application is rebuilt and the server restarted on file changes in the /client and /server folders respectively, for development purposes.
• npm run build: builds the two variants of the application.
• npm run serve: builds the two variants of the application and starts the NodeJS server.
• npm run automation: runs the automation script.

If you'd like to see how to set up the feature flags in Split UI and automatically create metrics for the Web Vitals measurements captured in this code example, see the blog: 'Instant feature impact detection for webpage performance: Split’s hidden gem'. Since you are already peering into the code, check it out and see Split's powerful IFID capabilities in action!