Independent Liquor Store Data: Where Your Coverage Goes Blind

Here’s a scenario that shows up in almost every conversation we have with a growing beverage-alcohol brand. The distributor’s authorized-account list has several hundred stores on it. The panel report shows a respectable %ACV distribution number and a flat trend line. The sales team swears the brand is moving. And nobody in the room can reconcile the three, because they’re describing different populations of stores.

The panel isn’t lying. It’s answering a narrower question than anyone realizes it’s answering. Syndicated retail measurement is built from register data that retailers agree to hand over, and the independent liquor store on a corner in Queens or Austin or Tampa has neither a reason nor a mechanism to hand anything over. That store still sells your product. It just doesn’t exist in your file.

We’re gmware, a software and data engineering firm in Austin, TX, with delivery centers in Bangalore and Mohali, India. We also run Shield Suite, our retail-intelligence platform for beverage-alcohol brands across 60,000+ storefronts. We spend a lot of time reconciling distributor account lists against measured-store lists, which is a boring job that produces the single most uncomfortable number in a brand’s data stack: the share of your own volume that nothing you subscribe to can see.

This post explains the mechanics of why that happens, then gives you a reconciliation you can run yourself in about a week.

The vocabulary that hides the problem

Start with ACV, because everything downstream inherits its shape. All Commodity Volume is the total annual sales volume of a retailer across its whole inventory, “rather than sales for a specific category of products,” and it aggregates from individual store level up into larger geographic sets (Universal Marketing Dictionary). It’s a store-size yardstick. It has nothing to do with how much wine or tequila that store sells.

Then %ACV distribution puts that yardstick to work. FMI’s food industry glossary describes it as a measure of the breadth of a product’s distribution, similar to the percentage of stores selling an item except that larger stores carry more weight (FMI glossary). Two details on that page do more damage to naive readings than anything else in the industry’s vocabulary. First: “just because an item is authorized does not mean it is in distribution.” The item has to scan at least once in the period to count. Second, in capital letters on FMI’s own page: “Percent ACV Distribution is NOT additive across products, markets, or periods.”

Now put the two together and you can see the trap. Say a supermarket does $40M a year all-commodity and an independent liquor store does $1.5M. Those are illustrative numbers, not measured ones, but the ratio is the point: winning that supermarket is worth roughly twenty-six of the independent in %ACV terms. A brand can add a hundred independents and watch %ACV barely twitch. The metric is behaving exactly as designed. It’s just answering “how much shelf-adjacent retail dollar-weight can reach my product” rather than “how many places sell me,” and those diverge violently in a channel full of small doors.

The third piece of vocabulary is projection. A syndicated panel is a sample of cooperating stores, cleaned and pooled and then scaled to represent a universe the provider defines. Census data, by contrast, is every store in scope with no scaling. Most brands treat their panel file as if it were a census, and that assumption is where the real errors get born. A store that isn’t in the sample doesn’t appear in your file at all. It doesn’t appear as a zero, either. It’s simply outside the frame.

“All-outlet” is the fourth term, and it’s the most misleading, because it sounds like a promise about completeness. In practice it means the provider has stitched multiple channels into one view. It does not mean every outlet in the country is measured.

Why independents don’t get measured

This is structural, and it’s worth being fair about. Panels don’t skip independents out of laziness.

Register data arrives through a licensing agreement. Somebody at the retailer signs it, legal reviews it, IT wires up a nightly export, and an account manager maintains the relationship as POS versions change. A regional chain has all four of those functions. A single-store operator whose “data team” is the owner’s nephew has none of them, and no commercial upside to creating them. There’s nobody to sign.

Second, the POS heterogeneity is genuinely brutal. Independent liquor retail runs on a long tail of systems, some of them decades old, with product files maintained by hand and UPCs that don’t always match the bottle. Even when an owner is willing, getting a clean, consistently structured feed out is a project rather than a switch. We’ve written about what that looks like from the retailer’s side in our liquor store analytics guide.

Third, ownership is fragmented by design in much of the country, and control-state structures add another wrinkle. NABCA’s own directory is internally inconsistent about the count: a section heading says “18 Control Jurisdictions” while the introductory text says seventeen, and individual state profiles say both. Thirteen of those jurisdictions “also exercise control over retail sales for off-premises consumption” (NABCA control state directory). Where the state runs the store, the data path is a government reporting system, not a commercial panel, and it behaves nothing like a scan feed.

You can see all of this reflected in how providers describe their own liquor coverage. When NielsenIQ launched its expanded Liquor Channel Open State view, it said the view “expands the base of stores measured, including a wide array of Liquor chains and independent stores,” extends to “an aggregate of 31 states and Washington, D.C. (non-control markets),” and “refines universe statistical projections” (NielsenIQ, July 2024; same release via Business Wire).

The most useful numbers in that release are the coverage lifts, because they describe the size of the hole the expansion filled.

Read that the other way around and it’s sobering. Whatever a brand concluded from the older liquor channel view, it concluded from a file covering a fraction of the beer sales the newer one covers. And the expanded view is still non-control markets, with named channel detail for eight states. That’s not a criticism of NIQ. It’s the clearest public evidence we’ve found of how much of this channel goes unmeasured, published by a provider being straightforward about it. If you’re comparing providers on exactly this axis, our Circana vs NielsenIQ buyer’s guide is the companion read.

What projection does to a small brand’s read

The mechanism deserves a careful statement, because the internet is full of confident error-magnitude claims that nobody can source.

When a brand’s volume is distributed roughly like the panel’s sample, projection works. The measured stores stand in for the unmeasured ones and the scaled number lands close. When a brand’s volume is concentrated in stores the panel doesn’t measure, projection has nothing to scale. Your cases in those stores are not underestimated. They’re absent. The trend you read is the trend of your measured minority, and it can move in the opposite direction from your actual business without anything in the file looking broken.

There’s a subtler version that catches brands the other way. Suppose you’re strong in a handful of large measured accounts and thin everywhere else. Projection assumes your measured performance is representative and scales it across the universe. Now the file can overstate you. Direction depends entirely on how your distribution skews relative to the sample, which is why the honest answer to “how wrong is my panel number” is always “depends on your footprint, and here’s how to check.”

What about the national size of the gap? We went looking for a citable figure and couldn’t find one worth printing. Even the count of US off-premise alcohol retailers doesn’t settle: American Beverage Licensees describes its membership of on-premise and off-premise retailers as numbering “nearly 20,000” (ABL), while third-party industry publishers citing Census County Business Patterns report roughly 36,000 beer, wine and liquor store establishments (VantaInsights) and elsewhere 31,835 (Vertical IQ). Those figures are measuring different things: association members, NAICS-classified establishments, and a vendor’s own store file. None of them tells you how much category ACV sits outside a panel’s universe, and providers treat their universe definitions as proprietary.

So the share of the category living in unmeasured independents is not a published number. Anyone quoting you one is guessing. Which leaves exactly one credible move: measure your own footprint.

How to find out how blind you actually are

This audit takes a competent analyst about a week, and it produces a number you can take into a budget conversation.

A few specifics that will save you a wasted pass.

Match on a standardized address key, not a name. The same store shows up as “Joe’s Wine & Spirits,” “Joes Wine and Spirit,” and “JOE’S W&S #2” across three distributors. Run the addresses through USPS-style standardization, key on standardized street line plus ZIP+4, and keep the raw strings for auditing. Name matching on this channel is a trap that eats a month.

Then the aggregation is a five-line query in shape:

select
  coalesce(m.is_measured, false) as in_reporting,
  count(distinct a.account_key)  as accounts,
  sum(d.cases_90d)               as cases_90d
from dim_authorized_accounts a
left join dim_measured_accounts m
       on m.address_key = a.address_key
left join fct_depletions d
       on d.account_key = a.account_key
      and d.week >= current_date - interval '90 days'
group by 1;

Two rows come back. If the in_reporting = false row holds a minority of accounts but a minority of cases too, your panel read is broadly trustworthy and you can stop here. If it holds most of your cases, every trend, share, and promotional readout you’ve built on that file describes a subset of your business, and you should stop making decisions from it until you’ve fixed the coverage.

The third bucket is worth isolating separately: accounts on the authorized list with zero cases in ninety days. Those are distribution voids, stores that should be carrying you and never started, and they’re a different problem with a different fix, which our out-of-stock and void reporting work is built around.

Buy or build the coverage

Once you know the size of the gap, you have three real options and they’re not equivalent.

Field-rep audits are the oldest answer and still the most detailed. A rep in the store sees things no feed captures: where the bottle sits, what the owner thinks, whether the competitor’s distributor rep was in yesterday. The cost is linear in visits, so it scales the way headcount scales, which is to say badly. Reps are the right tool for a concentrated priority list, not for national coverage.

Retailer direct feeds are the best data you can get and the hardest to get. Real register truth, units and price, no projection. But you only get them from retailers willing to sign, which loops you straight back into the reason independents are unmeasured in the first place. Chase these with your top accounts and don’t plan your coverage strategy around them.

Third-party storefront observation is what we built Shield Suite to do, and it’s worth being precise about what it can and can’t tell you. It observes the shelf: whether your SKU is present in a given store, the real price on that shelf rather than a list price, whether a display you paid for actually went up, and whether an authorized store has gone empty. What it does not do is count units sold. It’s presence and price and execution, not a register. Brands that expect it to replace scan data end up disappointed; brands that use it to cover the stores scan can’t reach get exactly what they came for. That trade-off is the same one we lay out for the whole stack in our first retail data stack guide for emerging brands.

When you should skip all of this

If your volume is concentrated in measured chains and you already pull direct feeds from your top three retail partners, adding independent coverage buys you very little. Run the audit, find that the unmeasured bucket holds a small slice of your cases, and go spend the budget on execution instead. Same answer if you’re pre-scale in one or two markets and your sales lead can still name every account from memory. Coverage tooling replaces knowledge you can’t hold in your head. Below that line it’s an expensive way to confirm what you already know.

The uncomfortable middle is the brand doing real volume across many markets with distribution skewed toward independents, still reading its business off a panel file. That brand is making national decisions from a partial view and doesn’t know the size of the partial. That’s who this audit is for.

What we’d recommend

Run the reconciliation before you buy anything, including from us. It costs a week of analyst time, it uses data you already own, and it converts an argument about data quality into a single defensible percentage. We’ve watched that number change budget conversations in ways no vendor deck ever does, in both directions. Sometimes it says the panel is fine.

If it says most of your cases live in stores nobody reports on, then the question becomes which of the three coverage options fits your gap, and the answer usually turns out to be a blend: reps on the priority accounts, direct feeds where a partner will sign, and observed storefront coverage for the long tail neither of those reaches. If you’re also weighing how much of your read should come from bars and restaurants versus stores, our on-premise vs off-premise breakdown sorts that out.

Send us your distributor account list and whatever reporting you have today, and we’ll map which portion of your footprint currently goes unmeasured and we’ll give you a straight answer on whether closing it is worth the money. We do this work with beverage-alcohol brands and distributors constantly, and the honest answer is sometimes no.