A sampling program ends, sales at those venues are up, and the program is written up as having driven growth. Some of that is usually true. The useful question is which part.

A free sample does three different things

The cleanest account of this comes from a 2004 paper by Kapil Bawa and Robert Shoemaker in Marketing Science. Rather than asking whether sampling worked, they pulled apart what it actually did and found three effects tangled together.

Acceleration. Some people bought sooner than they otherwise would have. The sale was always coming; the sample moved it forward. Over a long enough window this contributes nothing — it borrows from next quarter.

Cannibalization. Some people took a free unit of something they were going to buy anyway. The brand paid to give away a sale it already had.

Expansion. Some people bought who genuinely wouldn't have. This is the only new money in the program.

A venue-level sales bump four weeks after an event contains all three and looks like only the third.

Why the split matters commercially

Because the three have different implications for what you do next.

If a program is mostly acceleration, running it more often doesn't compound — it keeps pulling the same demand forward. If it's mostly cannibalization, the program is a loyalty spend wearing a trial program's clothes, and it should be budgeted and targeted that way. Only expansion justifies the targeting logic most brands apply, which is to find people who have never bought the product.

This is also the strongest argument for sampling in rooms where your product isn't already selling. In an account that already moves your brand well, the balance tilts toward cannibalization almost by definition: the people in that room are disproportionately people who already buy you.

The part nobody measures

The same paper found effects on purchasing observable as long as twelve months after a sample. Twelve months is a ceiling rather than a typical result, and it sits awkwardly beside a four-week readout.

How much of a sample's total effect lands in that first month is a separate question, and the honest answer is that nobody has quantified it. It may be most of it. The problem is that no one can tell you, and the convention treats the question as settled.

What to ask of your own report

Three questions that separate a growth claim from a number.

What would these venues have done anyway? If there's no comparison — venues running something else, or nothing — the answer is unknown rather than zero.

How much of this is people who already buy us? If the venue list was chosen for relationship reasons, the honest expectation is a high cannibalization share.

How long did we look? And was that window chosen because of the behavior or because of the invoice?

Growth is the word doing the most unearned work in this category. We'll tell you what a program of ours can establish about it, and what it can't.