Design an A/B test for WFH filter

A marketplace search page will add a displayed-but-optional 'Work from home' filter that, when clicked, narrows listings to remote-friendly options. Design an online A/B test to estimate the causal effect of showing this filter on conversion defined as bookings per visit (bookings/visits). Specify: (a) precise H0 and H1; (b) unit of randomization (e.g., visitor) and exposure/trigger definition (e.g., search-page load) including how returning users and cross-device identities are handled; (c) assignment and compliance rules (treatment sees the filter; control does not), including how to prevent cross-arm contamination; (d) metric definition: exact numerator/denominator, 28-day post-visit attribution window for bookings, tie-breaking when multiple visits occur before a booking, and whether to credit at the visit- or visitor-level; (e) guardrails (latency, bounce, listing CTR, cancellations) and pre-trend/AA/SRM checks; (f) segmentation by country and how to combine strata (e.g., inverse-variance weighting) and detect heterogeneity; (g) variance reduction (e.g., CUPED) and outlier/bot filtering; (h) sample size and duration to detect a 0.5pp absolute uplift from a 3.0% baseline at 80% power and alpha=0.05, including expected daily traffic; (i) analysis plan comparing ITT vs triggered analyses given only a subset click the filter, and how to estimate the effect among clickers without selection bias (e.g., instrumental variables or CACE); (j) risks (novelty, cannibalization, latency regressions, misattribution across visits) and mitigations.