Backend API Feature Flags: Percentage User Targeting for Checkout Cost Attribution

Backend API feature flags enable percentage user targeting for checkout cost attribution without placing account identity into client-controlled requests. The backend assigns each account to a deterministic cohort and records the flag key, revision, and cohort alongside every failed checkout. Configuration polling refreshes the setup, with the most recent validated snapshot surviving a control-plane outage.

This approach allows a small SaaS team to perform simple percentage releases without letting a browser decide who experiences payment-path behavior.

A global error-rate graph may indicate that checkout performance declined, but it cannot pinpoint whether new paths caused failures for high-volume accounts or how much those failures cost the business. Therefore, the flag decision must become part of operational evidence, while keeping account identity out of client-controlled requests. The backend API should handle feature flags and percentage rollouts by treating them as allocation rules rather than explanations.

When a checkout_v2 moves from 5% to 20%, and declined payments increase, the evaluated variant and stable account bucket per failure are necessary to separate rollout exposure from processor issues or tenant-specific integrations. A compact failure record containing an internal account identifier, order identifier, flag revision, assigned variant, failure class, and integer amount in minor currency units should be used.

The revision and unit matter, and both signals should be considered when defining alerts in SLO terms. User targeting should use immutable server-known attributes like account ID, plan, or an explicit allowlist, rather than trusting a React client to announce its own cohort.