The report from a sampling day usually says something like this: 1,400 shoppers engaged, 900 samples distributed, positive sentiment observed, strong interest in the new variant.
Every word of it is probably true. None of it survives contact with a trading meeting, because none of it answers the question the person across the table is actually asking, which is whether the money moved any product.
The four things a demo day cannot tell you
Who tried it
A demonstrator hands a cup to a person who walks away. There is no identity attached to that trial, so there is no way to know whether they were already a buyer, whether they were in your target audience at all, or whether they were the same person coming back for a third sample.
Without identity there is no follow-up, no re-contact, and no possibility of ever knowing what happened next.
Whether they bought it
Sales data from the day records the store's total sales. The store was also running a promotion, the weather was good, and it was pay week. Attribution in a single store on a single day is guesswork dressed up as a number.
Whether they bought it again
This is the only question that matters in a repeat-purchase category, and a demo day has no mechanism for answering it. Trial is worth having because it can turn into habit, and nobody at the table can see whether it did.
What they actually thought
Sentiment observed by a demonstrator who is being paid to be enthusiastic is not evidence. People are polite. Somebody who takes a sample, says lovely, and never buys it looks identical in that report to somebody who takes a sample and switches brands for good.
Why the number still gets approved
Because it is a big number, it arrives quickly, and it feels like activity. An estimated shopper-engagement count can look more impressive than a smaller verified-trial count, until somebody asks what happened to sales.
It is also familiar. The people approving the spend understand the process, and nobody gets criticised for running the same activation as last year.
What a defensible read looks like
A trial you can defend has four properties, and none of them require a store.
- Identity. You know who tried it, at least at the level of a matched profile.
- Verification. The purchase is proven rather than assumed. Proof of purchase is matched to the customer and to the content they post.
- A record. Something exists after the campaign: content, a verbatim, a piece of evidence that can be shown to a buyer.
- A comparison. The result is measured against your own advertising in the same window, not against a category benchmark from another market.
That last one gives you a relevant comparison. In a New Zealand campaign for Kiwicare Weed Weapon in December 2025, customer content and the brand's own paid ads ran during the same campaign period. Customer content returned 2.78% click-through against 1.07%, at NZ$0.20 a click. Those are results from ads running alongside each other, rather than a test that isolates what caused the difference.
Source: Kiwicare Campaign Report, January 2026.
The trade nobody explains
Moving from a demo day to in-home sampling means accepting a smaller headline number. A smaller verified-trial count in place of a larger estimated shopper-engagement count can look like a downgrade on a slide.
The value is in knowing which trials happened and having a record to follow up. A tally from a folding table does not give you that on its own.
What to ask for instead
Three questions will separate an activation report from an evidence report.
- How many trials were verified, rather than estimated?
- What exists after the campaign that we own and can run?
- What did it do against our own advertising, in the same window?
If the answers are none, nothing and unknown, the activation may still have been worth running. It just cannot prove it, and that is a different thing from having worked.
The product sampling page sets out how the purchase-first version runs, and household and cleaning covers the categories where a week at home tells you more than a spoonful in an aisle.