What it means
Pick a random portion of your list or your audience and do not send them the campaign. Everyone else gets it. The difference in what the two groups do afterwards is the campaign's real effect, because the untouched group tells you what would have happened anyway.
Why it matters to a small business
It is the only method on this page that answers the question owners actually have, which is whether the money did anything. Every other number tells you what happened after the campaign. A holdout tells you what happened because of it, and it costs nothing but the discipline to leave some people alone.
You have 1,000 customers on your list. You hold back a random 200 and email the other 800. The 800 produce 48 orders. The 200 produce 6.
6 orders from 200 is 3%, so 800 similar people would have produced about 24 anyway. Observed 48 minus that baseline
The email added roughly 24 orders, not 48. Half the sales in the report were going to happen regardless.
The bit people get wrong
The holdout has to be picked at random, not by convenience. Holding back your least engaged customers, or the ones without a mobile number, guarantees the treated group looks better and proves nothing except that engaged customers buy more, which you already knew.