Ask a brand team what their share of shelf looks like in Texas and you’ll get a number. Ask them when that number was last true and the room goes quiet.
That gap is the subject of this post. Competitor shelf tracking isn’t one job. It’s four, and they tolerate staleness very differently. Confusing them is how brands end up paying real money for data that answers a question nobody asked.
We’re gmware, a software and data engineering firm in Austin, TX, with delivery centers in Bangalore and Mohali, India. We build and run Shield Suite, retail intelligence for beverage-alcohol brands across 60,000+ storefronts, so the observation problem is our day job. Four methods below, compared without flattering ours, including the two you can run yourself without buying a thing from us.
Pick the job before you pick the method
Four separate questions hide under the phrase “competitor shelf tracking.” Decide which one you’re funding before you talk to a vendor.
| Job | What you’re really asking | Staleness budget |
|---|---|---|
| New-item detection | Did a rival put a SKU on shelf that wasn’t there before? | Days |
| Share-of-shelf shift | Did their facings grow at my expense in this chain? | Weeks |
| Distribution gain or loss | Which specific stores did they win, and which did they lose? | Weeks, store-level |
| Promo and price response | Did they discount or take an endcap against my program? | Days |
The two “days” rows are where most programs quietly fail. A quarterly report can tell you a competitor gained distribution last quarter. It cannot tell you in time to do anything about it.
Method 1: field-rep and broker audits
A human walks the aisle, counts facings, photographs the set, notes prices, and files a form. Nothing else gets you this depth. A rep sees the endcap that isn’t in any database, hears the buyer say a competitor bought the reset, and can tell you the cold box got reorganized last Tuesday.
Tooling is mature. Retail-execution platforms like Repsly exist to structure this work, covering “planogram compliance, on-shelf availability, competitor tracking” through a rep mobile app with photo capture, per their own product claims.
The failure mode is the observer, and the best documentation of it is still Gruen and Corsten’s Procter & Gamble-funded Comprehensive Guide to Retail Out-of-Stock Reduction, which built its findings on exactly this method and then catalogued what’s wrong with it. Their list is worth reading if you run a field program:
- Audits get done “first thing in the morning when the shelves are more likely to have been restocked,” so the shelf you record is tidier than the one shoppers meet at 6pm.
- “The accuracy is diminished by unavoidable counting and other human errors due to inattention and fatigue of the auditors.” Hour four is not hour one.
- Whoever pays picks the categories: “the selection of the categories is often unbalanced, being influenced by a supplier who sponsors the audit.”
- Sustained measurement gets “prohibitive,” and the method is “not feasibly scalable to a large number of categories or a large number of stores.”
None of that makes rep audits bad. It makes them a depth instrument. If you use one rep’s count in Dallas and another’s in Sacramento as comparable numbers in the same deck, you’ve turned an observation problem into an arithmetic problem.
Method 2: syndicated distribution measures
The panel providers don’t sell you a picture of the shelf. They sell you a weighted statement about which stores carry an item. That’s a different thing, and it’s worth being precise about what you’re buying.
NielsenIQ defines %ACV as “Total ACV For Stores Carrying Your Product/Total ACV For All Stores” in its CPG dictionary, which means each store contributes in proportion to its total sales rather than counting as one unit. Their own worked example: a product in three of four outlets can score 67% ACV rather than 75% numeric, because it missed the biggest door. Total Distribution Points stacks that up. NIQ describes TDP as combining “the number of retailers your products are in (breadth) and the number of products you’re selling in those stores (depth),” calculated by adding the %ACV for every item you sell.
Two honest limits follow directly from that.
First, these are measures of the measured universe. Panel coverage grows when retailers agree to contribute point-of-sale data, which is why coverage expansions get announced as news. In beverage alcohol specifically, off-premise volume runs heavily through independents, and that’s where panel coverage thins out. We wrote about where that goes wrong in independent liquor store data coverage, and the broader provider comparison lives in our Circana vs NielsenIQ buyer’s guide.
Second, distribution is not shelf presence. %ACV tells you a store carries the item. It says nothing about how many facings it holds, whether it sits at eye level or on the bottom shelf, or whether a competitor’s display is standing three feet away. For share of shelf in the literal sense, this method structurally cannot answer.
The related depletion aggregations sit further back still. WSWA’s SipSource builds from distributor transactions covering “over 70% of wholesale products by volume” across all 50 states and publishes quarterly, with metrics on a rolling twelve-month basis. NABCA’s control state results come monthly, in 9-liter case volume, dollar volume, and price mix. Both are excellent for trend. Neither was built to catch a new SKU.
Method 3: retailer sites and delivery-app catalogues
This is the cheapest method and the fastest, and every brand should be doing some version of it. A competitor’s new SKU usually appears in a retailer’s online catalogue around the time it enters the system. You can see pack configurations, listed prices, and often the store-level availability toggle.
It also gets oversold, and the reason is mechanical.
An online catalogue is a projection of the retailer’s inventory record. Inventory records are wrong a lot. Gruen and Corsten’s audit of more than 20,000 items across 121 stores of a major US retailer found perpetual inventory “accurate only 45.4 percent of the time,” splitting the rest between phantom inventory (the system thinks there’s stock and there isn’t) and hidden inventory (the reverse). That result lines up with earlier work by Raman, DeHoratius, and Ton in California Management Review, which examined roughly 370,000 inventory records at a large chain and reported 65% of them inaccurate.
So “available online” means a store bought the product and the system believes some is left. It does not mean a shopper standing in the aisle can find it. For a competitive read that’s often good enough. For anything about facings, position, or display, it’s useless.
There’s a durability problem too. Third-party catalogues are business decisions, not infrastructure. Uber shut down Drizly in January 2024, roughly three years after buying it for $1.1 billion, with orders accepted through the end of that March (TechCrunch, AP). If your competitive tracking depended on one app’s catalogue, that was the week it stopped working.
Method 4: systematic storefront-level observation
The fourth approach treats observation as an engineering problem rather than a staffing one. Instead of sending fewer people deeper, you make the same structured observation at a very large number of storefronts, on a repeating schedule, using one definition of every field so the results compare across markets.
That’s the approach behind our competitive intelligence module: competitor shelf presence, competitor pricing market by market day over day, promotion and display detection, and alerts when something changes. The design goal isn’t to know one store better than a rep does. No feed beats a good rep inside a single account. The goal is that the thousandth store is recorded the same way as the first, so a change in the number means a change on the shelf rather than a change in who was holding the clipboard.
Where it breaks: it only sees what’s publicly visible. It isn’t in the buyer meeting, doesn’t know what got promised for next spring, and doesn’t replace your reps. Anyone selling observation as a substitute for field presence is selling something that doesn’t exist.
The four, side by side
| Method | Detects change in | Breaks when | Cost shape |
|---|---|---|---|
| Rep audits | Your visit interval | Observers count differently | Per visit, scales with people |
| Syndicated | The reporting period | You need facings or position | Annual subscription |
| Catalogues | Days | The record disagrees with the shelf | Near zero, plus your time |
| Observation | The refresh interval | You need what’s off the shelf | Per storefront, per period |
Read the “breaks when” column first. That’s the one that decides whether the method fails at the job you’re paying it to do.
Which method answers which job
Latency is the whole game
Most competitive-intelligence conversations skip this part, and it’s the part that decides whether any of the above was worth buying. Finding a competitor’s new SKU in a quarterly report isn’t intelligence. It’s history with a chart on it.
Every method has a floor on how old a fact can be the first time you see it, and that floor is just the reporting interval. A monthly report can be describing something that happened 30 days ago. A quarterly one, 90. A rep cycle that reaches each account every six weeks has a six-week floor, and no amount of rep quality changes the arithmetic. The intervals below are the published cadences of each source, so the “worst-case age” is the interval itself.
How old is the fact when you first see it
Put your own numbers on it. Say a competitor takes two facings from you in 400 stores, and assume, purely for the arithmetic, that a facing turns three units a week. That’s 31,200 units of shelf-driven volume over a quarter you’re no longer standing in front of. Swap in your own assumptions and the shape holds: catching it in week two rather than week twelve is ten weeks of compounding, and no report frequency fixes that retroactively.
That’s the reason we built alerting into the competitive module rather than a monthly PDF. A change you learn about after the reset is finished is a change you file, not a change you fight.
What’s fair game
Worth being plain about this, because it’s the first question a good legal team asks.
Our line is public observation of publicly displayed retail merchandise. A retailer has put products on a shelf in a store open to the public, marked with a price, for anyone to see. Recording what’s there is the same act a shopper performs, done systematically.
The rules we hold ourselves to:
- Follow the store’s policy. If a retailer’s policy or its staff say no photography, that’s the answer, and it doesn’t get relitigated in the aisle.
- Never misrepresent an observer. No pretending to be a rep for another company, an auditor for the chain, or anyone other than who they are.
- Respect site terms for anything collected online. A site’s terms of use govern what you may do on it, full stop.
- Don’t touch non-public information. Buyer conversations, distributor systems, and internal planograms belong to whoever owns them.
Store policy and site terms are the governing documents, and the legal question past that belongs with your counsel, not a vendor’s blog. But if a method requires deceiving someone to work, it isn’t a method. It’s a liability with a dashboard.
A DIY protocol you can run next week
If you’re not ready to buy anything, here’s a program you can actually run. It borrows its measurement structure from Gruen and Corsten’s best-practice approach to planogram compliance, which breaks a shelf check into distribution (do the items on the shelf match what should be there), space (facings per SKU), and arrangement (correct shelf, correct brand order, correct item order within the block).
Four steps to a comparable audit
Start with the sample frame, because it decides everything downstream. Pick a fixed panel and keep it fixed. Stratify by chain and by market so a difference between them means something, and hold at least a third of the panel in independents if independents carry your volume. Don’t let reps choose which stores to visit. Self-selected samples drift toward the accounts where the rep is welcome.
Per visit, the fields worth capturing are your facings and each tracked competitor’s facings in the set, shelf number counted up from the floor, shelf price as marked rather than list, any secondary display or cold-box door or endcap and who owns it, any shelf tag with nothing behind it, and one wide photograph of the whole set from a fixed distance. That photo is your tiebreaker later, when two observers disagree and both sound certain.
Comparability across observers is the piece everyone underestimates. Write down what counts as a facing. One sentence, with a picture. Decide in advance how you’re counting a stacked case on the floor, a bottle turned sideways, and a competitor’s SKU sitting in your tagged space. Then fix the visit window: same weekday range, and after the morning restock, because auditing at open flatters the shelf every time. Rotate observers so no single person permanently owns one market’s numbers.
Cadence: every two weeks per store in the core panel, monthly in the long tail. Any slower and you’ve built a history project. Budget an hour a month to put the wide photos side by side, because the facings count tells you how big a change was and the photograph tells you what it actually was.
Two side notes. If you’re also verifying your own display and feature programs, that’s a related but distinct workflow, and we covered it in retail execution tracking. If you’re using shelf data to hold distributors accountable, the metrics that actually work sit in our distributor scorecard guide.
When DIY stops working
Usually you notice all of this at once, in the same quarter.
Your visit interval is longer than the change you care about. This is the hard one. A rep cycle that reaches each account monthly cannot detect a new competitor SKU in under a month. If new-item detection is a strategic priority, the program has already failed at it and no coaching fixes that.
The counts stop being comparable. When you can’t tell whether Nevada’s share-of-shelf drop is real or an observer change, you’ve lost the measurement. This is what Gruen and Corsten mean about counting error and fatigue, and it gets worse as the panel grows.
The cost curve turns. Rep audits scale with people, so they scale linearly with stores. Fine at 50. Painful at 2,000, which is what Gruen and Corsten mean by not feasibly scalable.
What a scaled feed changes is comparability and interval, nothing more. You get the same field recorded the same way across markets, refreshed on a schedule you don’t staff. What it doesn’t change is the need for reps in the accounts that matter, and any vendor who tells you otherwise hasn’t been in a liquor store lately.
What we’d recommend
Run methods 3 and 4 as your always-on layer and method 1 as your targeted layer. Catalogue monitoring is nearly free and catches listings early. Systematic observation gives you facings, price, and display on a schedule with one definition of every field. Reps then go deep in the accounts where a relationship, a buyer conversation, or a reset decision is actually in play. Syndicated distribution measures stay in the deck for share and trend, where they’re genuinely the right tool, and out of the conversation about what’s on the shelf this week.
Don’t buy any of it below roughly $5M in revenue with distribution in a handful of markets. At that size a fixed 20-store panel, a phone camera, and a disciplined spreadsheet will beat a platform, because your real constraint is attention rather than data. Buy when you have more stores than you can visit, competitors making moves you learn about too late, and a share-of-shelf number in a deck that nobody in the room can defend.
If that’s where you are, tell us which chains and markets matter and what you’re trying to catch, and we’ll give you a straight answer on whether observation solves it or whether you just need a better rep protocol. Sometimes it’s the second one. We work with beverage-alcohol brands and distributors on exactly this, and the competitive intelligence module is where the shelf-presence and pricing work lives.