What's new in Optimizely Experience Optimization: June 30, 2026

Idea builder: Skip the blank page. Idea Builder generates experiment hypotheses based on your page, goals, and live competitive research. Go from "we should test something" to a structured, prioritized experiment backlog in minutes.

Visual editor enhancements:  Building page-level experiments and personalize experiences just got faster. The visual editor now automatically detects elements you want to change, handles selectors more reliable, and gets you from idea to live variation with fewer sets in between.

Remote MCP server: for Experimentation. Many teams are already using tools like ChatGPT, Claude, or Cursor in their day-to-day work. Developers can now connect Web and Feature Experimentation directly to those tools via the Model Context Protocol. Manage flags, query experiment data, and take action without context switching.

Fixed horizon stats engine: Some teams want to define their sample size upfront, run for a fixed duration, and get a clean result at the end. That's exactly what Frequentist stats engine delivers. It sits alongside our existing Sequential and Bayesian stats engines so you choose the model that fits the test.

Bayesian stats engine: Prefer probability over p-values? The Bayesian Stats Engine updates continuously as data comes in and tells you the probability that each variation is best. Make faster decisions without sacrificing rigor. No waiting for a fixed sample size to call a winner.

Flag status: Running 20, 50, or 100+ flags? Knowing what's live, stale, or archived shouldn't require a team meeting. Flag Statuses automatically calculates lifecycle labels (Draft, Testing, Live, Paused, Stale, Archived) so the state of your program is always visible at a glance.

Copy a flag rule: Stop rebuilding flag rules from scratch when promoting across environments. Duplicate test and delivery rules between dev, staging, and production in seconds. 

Custom fields: For Feature Experimentation and Web Experimentation Your flags and experiments always carried context that lived outside the tool: in spreadsheets, in Confluence, in someone’s head. Custom fields gives that context a permanent home. Add structured metadata to every flag and filter to your entire program by team, initiative, page optimized, quarter, Jira ticket, Figma link or whatever matters to your org.

How to set up custom fields in Web Experimentation 

How to set up custom fields in Feature Experimentation 

Flag team & owner field: Every flag now has a clear owner when you create a new flag. For customers in Enhanced or Advanced tier, you can set up team names and filter by teams. No more Slack threads asking who set this up. No more orphaned flags sitting in production indefinitely. Ownership is explicit, visible, and filterable. 

Contextual bandits (CMAB): Traditional A/B tests find one winner for everyone. But your users aren't everyone. What converts for one segment can actively underperform for another. Contextual Multi-Armed Bandits uses AI to automatically serve each individual the variation most likely to convert, based on their real-time attributes. It learns as it goes, adapts automatically, and does all of it without manual segmentation or post-test reconfiguration.   

Configure contextual bandits in Feature Experimentation  

Configure contextual bandits in Personalization 

New results page: The results experience has been fully redesigned! It surfaces the metrics that matter most, makes statistical significance easier to read, and gives teams a single view to understand an experiment and share it across the org. Less time decoding data. More time acting on it.

Remote MCP server for analytics: Teams can now query warehouse data directly from the AI tools they already use: Claude, ChatGPT, Cursor, VS Code with Copilot, and more. Explore behavioral trends, read experiment results, and build dashboards on the fly using plain English.

Conversion drivers analytics: Now automatically surfaces the behaviors driving conversion and dropoff between any two funnel steps, ranked by correlation and statistical significance. Instead of manually hypothesis-testing why 40% of users drop off between Add to Cart and Checkout, your teams get a ranked list of what converted users did differently, then drill straight into a path analysis, segmented funnel, or cohort to dig deeper.

Chart annotations, targets & KPI comparison: Mark key moments directly on time-series charts, set goal values with target dates, and add period-over-period comparisons to tiles. Your dashboards now answer "what happened on June 12th?" without a Slack thread, track progress without a companion spreadsheet, and look presentation-ready for executive reviews.

New AI-powered analytics capabilities: Optimizely's agentic platform now lets your teams build custom dashboards from scratch, pin explorations as tiles, add context with text headers, and freely arrange and undo layout changes — plus it automatically interprets experiment scorecard results and surfaces the explorations behind each experiment's Explore tab — so customers spend less time chasing numbers and more time acting on them.