Know what your on-premise dollars actually bought.
Circuit reads what actually sold in bars, restaurants, and independent retail — and separates the lift your money bought from the lift that was going to happen anyway.
Advertising and membership, sold off the back of transaction data. Kroger, Target, and Albertsons all built the same business. It is among the highest-margin work those companies do.
Bars and restaurants ring up the same data. Almost none of them earn a cent from it.
Not because they don't want to. Building a data business takes a science team, a compliance function, and a sales organization that has nothing to do with running stores.
Circuit pools venues into cohorts until the cohort is big enough to be worth buying. That's the whole trick.
Source: public company reporting. Not a representation of Circuit earnings.
You're spending it. Nobody can prove it worked.
Brands pour money into bars and restaurants — well positions, menu features, samplings, promotions. Then they guess at what came back.
The report lands after the party
Weeks or months late. By then the promotion ended and someone else moved on it.
It counts the grocery shelf, not the bar
Syndicated panels were built on grocery scan data. On-premise runs on menu audits, depletions, and recall surveys — what one industry leader calls "a black hole for data."
"Did it work?" gets a model
Depletions tell you what left the warehouse. They can't tell you whether anyone ordered your brand instead of the one beside it.
Roughly 20% of gross sales — the second-biggest line on the P&L after making the product. Most of it has never been measured against a baseline.
Trade spend: Strategy& (PwC). Break-even rates: published industry studies ranging from roughly 25% to 70% ineffective depending on method. Illustrative, not a Circuit projection.
Two things. Both built on real receipts.
We measure trade spend — the money committed to features, well positions, samplings, and promotions. We are not a retail media network and we sell no advertising.
How you're doing against the brands next to you
We sort venues into cohorts by what people actually order there — craft cocktail, sports, fast casual, convenience. Pick the cohorts that matter and see who's gaining, who's slipping, and what gets ordered alongside you.
Whether your promotion actually sold anything
A well switch, a menu feature, a sampling, a coupon drop. We look at what people bought before, during, and after — and tell you how much of the change was you.
Cohorts are built, not collected. Each cohort — and each venue admitted to it — is selected to represent its slice of the market. Otherwise a benchmark is just noise with a label on it.
You get a dashboard that refreshes daily, plus written reads against your brands and your competitive set. Every number arrives with the venue count and margin of error behind it.
You're sitting on an asset. Nobody has ever tapped it.
It's oil under your parking lot. What rings up in your venue is exactly what brands can't see anywhere else — and on your own you're too small for anyone to come knocking.
Restaurants keep 3–5 cents of every dollar they take in. By partnering with Circuit to monetize your data, you don't have to buy any product, there's nobody to schedule, nothing to spoil or over-pour — so at a 4% margin, one dollar from Circuit does the work of about twenty-five dollars in extra food and drink sales.
It's the only line on your P&L with nothing sitting underneath it.
Plug in once
A secure connection to the POS you already use. Minutes to set up, and nothing changes about how you run the place.
Get selected for a cohort
Cohorts stay deliberately small and every venue in one is chosen — not everyone who applies gets in. That's why your venue counts for something instead of being one row among thousands.
Get paid
One published formula, the same for every venue, paid out of Circuit's own revenue. Never tied to what you stock, sell, or promote.
Four inputs. One honest number.
Sales jump for reasons that have nothing to do with you — it was 95 degrees, the game went to overtime, it was the Friday before a long weekend. Ignore that and a before-and-after chart is not proof.
The venues
A panel in the market you're promoting in.
What people bought
Every item rung up, before, during, and after.
What you ran
You tell us where and when. We don't guess it backwards out of your sales.
The signal layer
Weather, holidays, paydays, big games, concerts, local events — the things that move volume without you.
What the 8% was actually made of
± 0.9 margin of error
- Your program 3.0
- Heat wave 3.4
- Holiday weekend 1.6
Illustrative. Every Circuit read ships with its venue count, sample composition, and margin of error attached — and is withheld when the panel is too thin to support it.
The Retail Language Model. The part nobody else can copy.
Every register in America spells things its own way. Four systems, four spellings, one product — and that’s one product, in one venue, on one night. The Retail Language Model is our name for the whole system that reads it — raw export in, written finding out. Not a general-purpose chatbot. A system that speaks one language: what gets rung up in a bar.
Reads the file
Point it at a raw export and it works out which column is the clock, which ones are the products, which one is the money — and at what grain. Nobody sits down and maps a schema first.
Resolves the language
TTOS/SODA LIME, TITOS DBL and Tito’s Handmade Vodka 2oz become one product, one brand, one category. Every venue that joins adds spellings to the library.
Picks the cut
Two hundred SKU-level slices are each too thin to mean anything. The system finds the level where the sample can carry a finding, instead of handing you a chart that can’t.
Finds what moved
Trend and seasonality come out first. Then three passes for three different kinds of change: one-off spikes, stretches that break their own pattern, and shifts that never come back.
Writes the read
Correlates across cohorts, draws the charts, and writes the finding in words — with the venue count and the margin of error attached to it.
Most companies put this in the fine print. For us it's the architecture.
No individual is ever identified, resolved, or inferred.
No identity graphs. No loyalty matching. No card-linked measurement. No household panels. We care what sold, in which kind of room, at what hour, at what price. The data model has no place to put a person.
Four ways to measure the on-premise. Only one reads receipts.
| Method | Menu audits | Consumer surveys | Distributor depletions | Circuit |
|---|---|---|---|---|
| What it sees | What's on the menu | What drinkers recall buying | What shipped to the venue | What people actually bought, item by item |
| Refresh | Audit cycles | Often quarterly | Monthly | Daily |
| Well & hidden volume | Invisible | Unreliable | Volume, not choice | Captured at the line item |
| Proves incrementality | No | No | No | Yes, against a matched baseline |
| Controls for weather & events | No | No | No | Yes |
| Publishes coverage & error | Rarely | Rarely | No | Yes — on every cut |
| Pays the venue | No | No | No | Yes |
The things people actually ask.
Tracking what consumers buy in bars, restaurants, and other venues where the product is consumed on site — as opposed to off-premise retail like grocery and convenience. On-premise is roughly half of all US beverage alcohol dollars, but it has been measured through menu audits, depletions, and surveys rather than transactions. Circuit measures it from item-level POS data, refreshed daily.
No — and not as a policy we could revise. Circuit does not resolve consumer identity, build or license identity graphs, join to loyalty, card, device, or household identifiers, or infer individual-level attributes. The unit of analysis is the basket, the venue, the cohort, and the market.
This holds in every category we operate in, and it is not a rule we inherited from alcohol regulation — no US privacy statute treats alcohol purchases as a protected category. If anything the reverse is true: a grocery basket reveals more about a person than a bar tab does. That is exactly why the commitment is architectural rather than vertical, and why it survives a change of control.
If a retailer runs its own loyalty program, that is their relationship with their own shoppers. Circuit neither requests nor receives those identifiers.
No. Data is anonymized and pooled before it enters any product, and every cut must clear a minimum venue count before it renders. No buyer receives data attributed to an identifiable venue — not by name, not by inference, not by breakout — and there is no consent-based exception.
No. Circuit measures programs; it does not place, administer, fund, or deliver them, and it does not route or hold money moving between a supplier and a licensed retailer. Circuit buys venue data with its own revenue on a published formula that is identical for every venue.
Tied-house law is what makes this mandatory in beverage alcohol. It is not why we do it. Circuit is never paid for an outcome it also measures — a company that places a program and then grades it is marking its own work, and every buyer knows it. That does not change by category, so neither does the rule.
Their panels were built on retail scan data from grocery, drug, and mass. Bars don't scan a pour at checkout, so on-premise products from traditional providers rely on menu audits, depletions, and recall surveys.
Yes. Large retailers built high-margin data businesses on their own transaction data; independent venues never could, because no one venue has the scale to interest a brand. Circuit pools venues into cohorts that do, and buys their data out of its own revenue. No cost to the venue, no new hardware.
Three good reasons to get in touch.
Selling a brand into bars and restaurants. Trying to prove a promotion worked. Or running a venue and wanting to get paid for data you already generate.