How we measure what sampling does to sales at a bar, and what it's shown so far
Last week we published an audit of twenty-two companies in this category. Four case studies between them carried a sales figure, and none measured what sampling did to sales in a bar. The article ended by admitting the audit applied to us too: our site quotes a sales result, and we'd never said how we got it.
This is how. Paid sales of the sampled product rose 13% to 115% during event weeks, at every venue where we had a baseline. The method first, then what it has found, then what it can't tell you.
The instrument
Our platform connects to a participating venue's point-of-sale system. That gives us every line on every receipt: what was ordered, when, and at what price. It isn't a count someone wrote down at the end of a shift. The sample itself rings on the guest's tab at zero, so it sits in the same record as everything they paid for.
Each line carries more than the item: the quantity, the price actually paid, any discount applied to that line, when the check opened and closed, who rang it, and how many guests were on the check. It reaches us as the tab closes, not in a monthly file.
Because the record is the venue's own, it exists before we arrive and after we leave. That's what makes a before-and-after possible. Separately from our pilots, we hold on-premise receipt-level sales data from more than 500 bars.
The design
A standard pilot runs one event a week at each participating venue, for four weeks. We then compare paid unit sales of the sampled product, with the samples themselves taken out, across three windows.
The four weeks before the first event. That's the baseline: what the product sells at that venue when nobody is giving it away.
The four weeks of events. Did paid sales move while sampling was running?
The month after the last event. Did anything stick once the free pours stopped? This is also where a halo shows up, if there is one: other products from the same brand selling more than they did before.
Every pilot uses the same three windows, so a result in one market can be set against a result in another.
What it has shown
We've run this design on brands from major suppliers including Diageo, Constellation Brands and Boston Beer Company, across spirits and beer, in restaurants and bars.
During the events, paid sales of the sampled product rose at every venue where we had a baseline. The smallest lift was 13% and the largest was 115%. Those are changes against each venue's own baseline, not proof that the events caused all of it; the limits are set out below.
The month after is less tidy. In one program, sales held 13% above baseline for the full month. In another, the sampled expression fell below where it started while a different expression of the same brand sold several times its earlier rate. That could be a halo, or guests trading across the range once they'd tried one. We can't yet say which.
Where a venue didn't carry the product before the pilot, it kept selling it at full price after the free pours ended.
Lift was consistently larger at busy nightlife venues than at daytime and brunch spots, which matters more for choosing venues than any average does.
Of the people who registered for a sample, between 83% and 96% ended up with a drink in their hand, and events typically delivered 30 to 45 verified samples over two to three hours. That completion rate is a number this category never reports, because without a registration step there's nobody to lose. In one program, most of the people who took a sample said they'd never tried the brand before. In another, every guest who answered our follow-up survey already drank the brand, and most called it their go-to. Same mechanic, different job.
What it can't tell you
We compare each venue against its own past, not against similar venues that didn't sample over the same weeks, so a busy month can look like lift. In one program, some venues that ran a single event lifted about as much as other venues that ran four, which tells you some of what we measured wasn't the events.
The numbers at any one venue are small. When a product moves a pour a day, a handful of extra drinks reads as a triple-digit percentage lift.
A month afterward isn't the long run, and a bar's receipts mostly stop at the door. We do see some of what happens next. When sampling guests are offered a $6 to $8 rebate on a $35 bottle, between 5% and 11% buy the bottle and redeem it. About a quarter of guests who opt in answer a short survey afterward. And at venues that run a loyalty program, the venue's own records show whether a guest comes back for the product. For everyone a rebate doesn't reach, whether a bar sample led to a bottle in a shop is still more than receipts can show.
Any one of those would be a reason to discount a single result. Together they're why we're publishing ranges and a method rather than a headline.
What comes next
The step we keep telling brands to demand is the one we haven't taken yet: matched venues that don't sample, measured over the same weeks, so a busy month can't pass for a good program. Our next pilots are being built that way, and we'll publish them on the same terms: every number traceable to a receipt, the limits stated before the findings, and the comparison named.
An agency that publishes nothing can say its results are confidential. A platform that says measurement is the whole product doesn't get that excuse.
If you're a brand and want to see a full pilot readout, including how each window was built, ask and we'll walk you through it.



