## Overview Make the GLEX experiment engine more powerful and less error-prone, evaluate build-vs-buy for our experimentation platform, and automate testing infrastructure for both self-managed and SaaS changes. This epic combines two related streams of platform/infrastructure work: 1. **GLEX Framework Evolution** — core improvements to experiment assignment, caching, and clearing + platform evaluation (GLEX vs Optimizely/GrowthBook) 2. **UAT Infrastructure** — documentation and automation for testing self-managed and SaaS changes ## Goals - Fix GLEX assignment pain points (caching override, forced variants, clearing process) - Complete platform evaluation to make an informed build-vs-buy decision - Document and automate self-managed testing environments to remove engineer bottleneck from UAT - Reduce the engineering overhead of implementing and testing experiments ## Success Metrics - Platform evaluation decision documented with clear rationale - GLEX core improvements reduce experiment implementation complexity - Self-managed UAT can be performed without dedicated engineer setup time ## Target Quarters **FY27 Q3–Q4** - Q3: GLEX core fixes + platform evaluation decision + UAT documentation - Q4: Implement platform decision + SM environment automation ## Issues — GLEX Framework - [ ] gitlab-experiment#91 — Allow cached assignment to override exclusion logic - [ ] gitlab#591174 — Add forced variant assignment for backend-only experiments - [ ] gitlab#577708 — Improve the GLEX assignment clearing process ## Sub-Epic — Platform Strategy - [ ] Epic 19994 — Assess other possible options outside GLEX (Optimizely, GrowthBook) ## Issues — UAT Infrastructure - [ ] gitlab#579158 — Document how to manipulate an existing GitLab license for testing/verification - [ ] team-tasks#923 — Explore automating a self-managed environment for UAT