GLOSSARY

What is a holdout test?

A holdout test measures the true impact of a marketing intervention by randomly withholding it from a slice of the audience. The held-out group experiences the site or campaign without the treatment, at the same time as everyone else, and the difference in outcomes between the two groups is the lift the intervention actually caused.

Why a holdout is the only honest baseline

Every alternative baseline is confounded. Before/after comparisons absorb seasonality, traffic-mix shifts, and whatever else launched that month. "Model-estimated baselines" are a vendor's model estimating the vendor's own contribution. A concurrent, randomly assigned control group is the one design where the only systematic difference between the groups is the treatment itself — which is why clinical trials, ad-platform lift studies, and honest personalization measurement all converge on the same shape.

The practical tell when evaluating a tool: ask to run a holdout. A confident vendor helps you set one up. A nervous one explains why holdouts are unnecessary, or hard, or "already baked in."

How AXO runs holdouts

The holdout is always on, not a special study. Roughly one visit in ten is assigned to control by hashing the session — deterministic, so a visitor never flips between arms mid-visit — and that session sees no personalization anywhere on the site. Treatment and holdout conversion rates are compared with a standard two-proportion z-test.

Two honesty rules sit on top. Below minimum sample sizes the verdict reads learning, not winning. And below a traffic floor where significance is mathematically out of reach, the scorecard hides itself entirely rather than imply a verdict is coming.

QUESTIONS PEOPLE ASK

What percentage of traffic should a holdout be?

Enough control sessions to reach significance in a reasonable window, small enough that you are not withholding value from too much traffic. Around 10% is a common working point for always-on personalization holdouts — AXO holds out roughly one visit in ten. Lower-traffic sites need either a larger holdout share or a longer measurement window.

What is the difference between a holdout test and an A/B test?

An A/B test compares variants against each other — hero A versus hero B. A holdout test compares the whole program against its absence: personalization on versus personalization off. A holdout is effectively an A/B test where one arm is "nothing," which is what makes it the right design for answering "is this tool paying for itself?" rather than "which variant is better?"

Can you trust a lift number without a holdout?

No. Without a randomized concurrent control, a lift number is a correlation dressed as a result — seasonality, traffic mix, and parallel campaigns all move the metric on their own, and a dashboard comparing against last month or a modeled baseline cannot separate those from the treatment. The holdout gap is the only number that isolates what the intervention caused.

KEEP GOING
If your traffic is thin, we hide the verdictWhat is incrementality?How AXO measures lift
See the agentic experience layer on your own traffic.
One tag, live in about 15 minutes.
Start free trial →
MORE FROM THE GLOSSARY