Scale Experimentation With a Better Developer Experience

Natively integrated analytics and experimentation enable developers to learn from every release with less effort.

Inside Amplitude
March 21, 2023
Image of Larry Xu
Larry Xu
Director of Engineering, Amplitude Experiment
Scale Experimentation With a Better Developer Experience

As part of the engineering team for Amplitude Experiment, we care deeply about building a superior developer experience to deliver experiments at scale, without adding to the seemingly never-ending burden on engineering teams.

With our latest product launch, our team was focused on what key features and capabilities we could develop to build an experience for feature management and A/B testing that developers love. We are extremely excited to share what we have built and look forward to continuing to double down on superior developer experiences as well. Read on to learn more.

Test with behavioral context—without sacrificing performance

Part of Amplitude’s secret sauce is the ability to seamlessly manage user identity and leverage this insight to build audience cohorts powered by machine learning. With remote evaluation Experiment customers could seamlessly leverage user context in A/B tests, but many also wanted a more performant experience that could enable deeper targeting without adding latency to their user experience.

With local evaluation support for cohorts, developers no longer have to worry about this tradeoff. Our local evaluation is far more performant, reducing latency from 200 milliseconds to 5 milliseconds delivering 4,000% improvement. Teams can now leverage deep contextual insights to connect user data to their release workflow. Now, developers can deliver much more powerful tests without needing to worry about performance tradeoffs.

New functionality for our Management API to support more use cases

Our team has also added and updated a variety of endpoints in our management API for both feature flags and experiments. These updates make it much easier to automate QA and provide customers with new auditing capabilities. These new and updated endpoints include:

  • Variants
  • Variant users
  • Edit flag/experiment
  • Get experiment version details

This new functionality will make it much easier for customers to make programmatic updates to their experiments and flags. Check out our documentation to learn more about this new functionality.

Automated assignment tracking

Previously, customers had to figure out how to send either the exposure or assignment events to Experiment in order to run a test using local evaluation. This made using local evaluation implementations somewhat complicated.

Now, customers are able to automatically track assignments providing customers with a much more lightweight implementation for local evaluation.

New SDKs available

Our team also knows that our SDK coverage needs to align with the most popular languages developers want to work in. We are extremely excited that we have added SDKs for React Native, Ruby, Python, and Go to complement our other available client-side and server-side SDKs. To learn more, check out our documentation.

Want to learn more about what our team has been up to? Read Wil Pong’s blog post highlighting how Amplitude customers can scale experimentation when self-serve analytics is natively integrated with experimentation.

About the Author
Image of Larry Xu
Larry Xu
Director of Engineering, Amplitude Experiment
Seasoned technology executive with over 16 years in developing disruptive software technologies. Experienced in building engineering organizations enabling companies in the data space to quickly reach product market fit and grow from 0 to $100M+ in ARR.

More Inside Amplitude