<b>RPM A/B Plugins: Demand the Holdout Group</b>

<b>RPM A/B Plugins: Demand the Holdout Group</b>

Most 'RPM testing' plugins (Ezoic Leap, Mediavine experiments, custom GAM splits) compare variant A vs B. The one feature that separates real tools from theater is a true holdout — a slice that gets zero changes.

What it does — isolates the test from sitewide seasonality and CPM drift.
Setup effort (3/5) — needs deterministic user bucketing (hash the user ID, not random per-pageview).
RPM lift seen — not a lift itself, but it stops you from shipping fake +10% 'wins' that were just a good-CPM week.

Pros — kills false positives that cost you on rollout.
Cons — holdout sacrifices a few % of traffic to 'no change'; needs 2+ weeks for significance.

Lift: ★★★☆☆ — Risk: ★☆☆☆☆
Best for: anyone making permanent layout calls. Skip only if your traffic is under ~50k sessions/mo — you'll never hit significance.