You want to A/B test a new color scheme for your global Cloud Run web app on real production traffic. How should you design the study?

A/B testing requires simultaneously exposing different user groups to distinct versions of an application to compare their impact on a metric. An external HTTP(S) load balancer can split incoming traffic by a specified percentage, directing it to two separate Cloud Run services (one with the original color scheme, one with the new). This allows for direct comparison of sales data between the two groups to determine statistical significance. The second option describes a blue/green deployment, not A/B testing, as it doesn't compare variants concurrently. Mirroring traffic (third option) is for testing performance or errors, not for measuring user behavior like sales, as mirrored traffic doesn't interact with the application. Feature flags (fourth option) are suitable for A/B testing, but the question specifically asks about routing traffic to app variants, implying separate deployments, which an external load balancer handles effectively for Cloud Run.