Two people in the same brand meeting can read real data and disagree about whether the brand is growing. One is looking at depletions weighted toward bars and restaurants. The other is looking at scan numbers from grocery and liquor. Neither is wrong. They’re measuring different channels, and in beverage alcohol those channels behave almost nothing alike.
On-premise and off-premise aren’t two views of one dataset. They’re two separate measurement worlds with different vendors, different grains, different refresh cadences and different structural blind spots. Confusing them is how a brand cuts a SKU that was quietly building trial, or celebrates a display program nobody built.
We’re gmware, a software and data engineering firm in Austin, TX, with delivery centers in Bangalore and Mohali, India. We build the pipelines that blend these feeds, and we run Shield Suite, our own retail-intelligence platform for beverage-alcohol brands across 60,000+ storefronts. Shield Suite is off-premise only. We say that early and we’ll say it again further down, because knowing where a data source stops is worth more than any vendor’s coverage map.
This is written for the brand analyst, the new category manager, or the founder who just got quoted a five-figure data subscription and wants to know what questions it will and won’t answer.
| Term | What it means in one line |
|---|---|
| Off-premise | Sold in a store, consumed somewhere else |
| On-premise | Sold and consumed on the same licensed property |
| Depletions | Cases shipped from a distributor to any licensed account, either channel |
| Syndicated scan | Register sales from a measured retail panel (Circana, NIQ) |
| On Premise Measurement | CGA by NIQ’s brand and category sales tracking for the on-premise channel |
| Storefront observation | What’s physically on the shelf: facings, price tag, display, gap |
Where the line actually falls
Off-premise covers grocery, liquor stores, convenience, club, drug and increasingly delivery marketplaces that a physical store fulfills. On-premise covers bars, restaurants, hotels, stadiums and arenas, casinos, nightclubs, cruise lines and airlines.
The definitions are easy. The edge cases are where analysts get burned.
Off-premise isn’t one uniform universe. Which class of store may sell beer, wine or spirits is set state by state, and in control states the state itself sits in the supply chain. So a national off-premise “footprint” is really a patchwork where the same brand competes against a different set of shelves in every market. Any dataset that flattens that into one number is hiding the interesting part.
Tasting rooms and taprooms straddle the line. A winery pouring a glass at the bar is doing on-premise business. The same winery selling a bottle to go from the same counter is doing off-premise business. Same building, same staff, two channels, and most datasets will see one of them or neither.
Delivery muddies the off-premise read. An app order fulfilled by a liquor store is off-premise volume, but whether it shows up in that store’s reported POS depends entirely on how the marketplace integration was built. Sometimes it lands as a normal transaction. Sometimes it lands somewhere nobody reports from.
Hotels, stadiums and airlines are on-premise with strange plumbing. They often buy through procurement structures that look nothing like a bar ordering from a rep, which means they surface in distributor data as account types your mapping logic wasn’t built for.
The channel split everyone quotes and nobody sources
You’ll see an on-premise/off-premise split repeated constantly and almost never with a source attached. The most-cited figure comes from NIQ, which has stated that on-premise accounts for 61% of global beverage alcohol value sales, excluding wine, a figure also carried in the Business Wire release. Read the qualifiers, because they do all the work: global, not US. Value, not volume. And wine is excluded outright.
Value and volume splits diverge hard in this industry, because a serving poured in a bar carries several times the price of the same serving bought in a store. A channel can be a minority of your cases and a majority of your dollars at the same time. Category makes it worse: a craft brewery’s draught mix looks nothing like a vodka brand’s, and an RTD brand’s looks like neither.
So don’t quote a split. Ask four questions of anyone who does: which category, which measure, which geography, which year. Without all four it’s decoration.
What is documented cleanly is that the on-premise venue base is still moving. CGA by NIQ reported that the US on-premise universe grew 1.9% year-on-year in 2024, which matters for distribution math even when total volume is soft.
The two numbers you can actually cite
What you can actually measure on-premise
Four families of source, roughly ordered by how many brands can actually get at them.
Distributor depletions come first because they’re already yours. Every case a distributor ships to a licensed venue gets recorded whether or not anyone runs a panel, which gives depletions the widest on-premise footprint most suppliers will ever see. It also stops at the venue’s back door. We went through the mechanics in our guide to depletion data, VIP, iDIG and SipSource, so we won’t repeat that here.
Specialist on-premise measurement is a much smaller field. CGA was acquired by NielsenIQ in June 2022 and now trades as CGA by NIQ. Its On Premise Measurement service “measures brand and category sales performance in On Premise across multiple territories,” and NIQ said in July 2024 that it had pushed the service out from long-established UK and US coverage into Australia, Canada, Germany and France, with South Korea to follow. If you want on-premise share rather than only your own shipments, this is the category-defining option, and there isn’t a crowded field behind it.
Foodservice measurement sits adjacent rather than native. Circana’s CREST tracks restaurant and foodservice market share and consumer behavior, and its SupplyTrack dataset is described by partner Datassential as “one of the most widely trusted sources for foodservice operator purchasing, sales, and volume trends”. Foodservice-shaped, not bev-alc-shaped. Still answers occasion and operator-purchase questions no scan panel touches.
Menu data is the fourth and the least obvious. Datassential tracks new menu items and how “flavours, pricing, and positioning adapt to local preferences,” across a stated 730+ chains and 67,000+ menu items. That’s the nearest thing on-premise has to a price feed, and it’s chain-only by construction.
Then your own sales team, which isn’t really a family so much as a fallback. Rep visit logs and survey apps are the cheapest on-premise data anyone has, the most biased, and often the only read available on independent venues.
What you can measure off-premise
Syndicated scan is the backbone. Circana and NielsenIQ measure what rings at the register across a measured retail panel, which gets you real consumer takeaway with category context wrapped around it. We’ve compared the two against store-level alternatives in our Circana vs NielsenIQ vs store-level buyer’s guide.
Underneath that sit two direct-relationship options. Retailer feeds: chains running supplier portals will hand you your own item performance, sometimes at store level, on their schedule and in their format. Store-level POS: raw transactions from stores you’ve struck a deal with, so exact price, exact basket, exact sell-through, for exactly the stores that said yes.
Storefront observation is the odd one out, because it doesn’t measure sales at all. It measures the shelf. Facings, the price tag, whether the display exists, whether the SKU is simply missing. That’s the layer that sees what the shopper sees, and it’s the layer Shield Suite runs across 60,000+ storefronts.
Control-state reporting is the piece most brands forget. NABCA “gathers, validates, and organizes data” from control jurisdictions, with availability varying by state and store-level or account-level detail in some of them. Underused, and a completely different animal from scan.
The decisions each channel can and can’t support
Business questions don’t map neatly onto channels, and the expensive mistakes happen when someone asks a dataset something it structurally cannot answer.
Which channel answers which question
The rows that need both channels triangulated are the ones with money attached.
“Should I cut this SKU?” Off-premise scan will tell you it’s a slow mover in grocery. On-premise depletions might tell you it’s a bartender’s default pour and the highest-margin line in the portfolio. Cutting on one channel’s read alone is the classic own-goal, and it happens most often to brands whose growth started behind a bar.
“Is the brand building trial?” On-premise usually leads here. Someone tries a cocktail out, then buys the bottle six weeks later. If you measure only off-premise, you’ll see the echo and credit it to whatever marketing ran that month.
“Is my price holding?” Off-premise answers this directly and on-premise mostly can’t, which we’ll get into next.
What neither side sees
Where the map ends matters more than where it’s dense. Five structural blind spots, none of which a vendor will lead with.
On-premise has no shelf-price analogue. The price a venue charges for a pour is set per venue, moves with happy hour, and lives in a POS you have no relationship with. Menu data catches the chains and goes dark on independents. So “is my price holding” is close to unanswerable on-premise, while off-premise has a direct read on it, which is exactly why price to consumer is a storefront problem rather than a survey problem.
Pour data is largely invisible. A keg depletes into a bar as one unit and then vanishes from your view. Nothing in your data says whether it poured in four days or four months, how much went down the drain, or whether the bartender free-poured 2.5oz where the spec said 1.5. Beverage-control systems exist, but that data belongs to the operator and rarely leaves the building.
Off-premise scan goes thin exactly where beverage alcohol is fragmented. Syndicated panels are strongest where chains report consistently. Independent liquor stores are neither consistent nor small in this category, so a brand that over-indexes there can look flat in scan while depleting fine. We mapped that gap in detail in our post on where independent liquor store coverage goes blind.
Neither channel tells you why. Both are outcome data. You’ll watch the number move and you won’t see the competitor’s new shelf tag, the resets that pushed you to the bottom shelf, or the fact that you got cut from a back bar in favor of a house pour. Causes need an observation layer or a human in the room.
And there’s no shared key across the two. Same SKU, different distributor item codes, different retailer item numbers, and no account master that spans a bar and a grocery store. Every “total brand view” deck you’ve admired sits on top of a mapping project somebody paid for.
Where Shield Suite stops, and why we say so
Shield Suite is off-premise. All four modules are about what’s happening at a store: competitor shelf presence and pricing market by market, out-of-stock detection and distribution voids at store level rather than regional rollup, real shelf price instead of list price, and verification that a display program went up where the plan said it would.
If your growth question is about back-bar placement, draught handles or cocktail menu listings, that isn’t us. We’d rather tell you that than sell you a coverage map that quietly omits a large slice of your volume. For that side of the house, your distributor’s on-premise account data and CGA by NIQ’s OPM are the honest starting points.
The corollary is a moment not to buy. If most of your volume pours in venues and you’re pre-scale, off-premise storefront intelligence is not your first purchase. Get your on-premise account data clean, then come back when the shelf starts mattering.
A triangulation recipe that survives contact
Blending the two channels without lying to yourself
Step 01 does more work than the rest combined. Pull twelve months of your own invoices, tag every account as on-premise or off-premise, and look at the mix in both cases and dollars. Most brands are more lopsided than they assume, and the answer usually tells you which single channel to go deep on rather than buying thin coverage of both.
Step 05 needs a real rule, written down. Something like: depletions own the demand-timing view, off-premise scan or storefront owns anything about price and sell-through, and where they disagree by more than an agreed threshold, someone goes and looks. Teams without that rule spend their quarter arguing about which report is right.
What we’d recommend
Pick the channel that carries your volume and go deep there before you go wide. A brand with 80% of its cases going through stores should spend on off-premise depth, because the questions that will cost it money in the next year are all off-premise questions with off-premise answers: which SKU to cut, whether price held, whether the display actually went up. A brand built behind bars should spend on on-premise account data and specialist measurement first, and treat off-premise as a later chapter.
Buy the second channel when a decision genuinely spans both. SKU rationalization across a mixed portfolio is the clearest trigger. Brand-building measurement is the second. Neither is a reason to buy two subscriptions on day one.
And when you do run both, budget most of the project for reconciliation rather than the connectors. The pipes are easy. Making a bar and a grocery store agree on what your product is called takes longer than anyone plans for, and it’s the part that determines whether the blended number is trustworthy.
Tell us your channel mix and which feeds you have today, and we’ll give you a straight answer on what’s worth buying and what isn’t, including when the answer is “not us.” We work with beverage-alcohol brands and distributors on exactly this problem.