graphical user interface, diagram, application

Progressive Delivery and Experimentation Work Better Together

Claire spoke about Optimizely’s belief that progressive delivery and experimentation belong together. By combining these two disciplines, you can help the entire product development team from engineering, product management, data science, and growth marketing place customer data at the center of the product development and delivery process. This spans from doing rapid prototyping at the ideation stage to doing feature free feature flagging at the build and release phase to doing optimization, testing, and personalization in the launch and measurement phase. All of this is supported by statistically significant real-time results and analysis for data teams that ensure your company is making the best decisions possible for its customers. Doing progressive delivery and experimentation ensures that you can move fast with confidence. Progressive delivery helps teams go fast and experimentation helps ensure you are building the right thing. 

Speed comes from letting engineers ship code multiple times a day and advanced feature flags provide granular control over how and when to release to customers. Quality comes from making it easy to roll out and roll back features without deploying, this allows you to catch and fix bugs in production before they become a problem. Confidence comes from knowing you’re building the right thing because you use experimentation and data to validate your product and features with customers. Most importantly, your entire team can work together with less friction and a customer-centric approach backed by data.

Our Roadmap for Product Development Teams

Jon focused on our roadmap for product development teams which covered: features in consideration, in development, and launched. Jon began by discussing Optimizely Agent, an easier way to deploy Optimizely Feature Experimentation as a microservice inside of your architecture. This affords you the ability to centralize requests, standardize implementation, and meet security and compliance requirements. It also makes upgrades easier as there is a single instance to upgrade versus having to manage multiple SDK instances. Optimizely Agent is already in beta and can be tried out by Optimizely Feature Experimentation or free free feature flagging users. The docs are now live for Optimizely Agent.


Deploying Optimizely Agent in a Service-Oriented Architecture

Next, Jon spoke about targeted free feature flagging, a new capability for Feature Experimentation that is coming soon which provides even more granular control over your release process. Targeted free feature flagging allow users to gradually roll out a feature for several different audiences to differing rollout percentages while monitoring the impact on performance and business metrics.


Targeted free feature flagging example: Roll out a feature to beta testers and 25% of US users

Jon also provided an update on enriched event export and Data Lab, which makes it even easier for data teams to consume and join Optimizely experiment data with their own 1st party data and use industry-standard Jupyter notebooks to do even deeper analysis on experiment data.