In a large share of US grocery categories, the person recommending which products the retailer should stock works for one of your competitors.
This is not a scandal. It is the documented, mainstream operating model of modern category management, and it has been studied by the FTC, the DOJ and antitrust scholars for twenty-five years. But it has a consequence that most brand teams never price in: at the range review, the argument for delisting your SKU is built from data you cannot see, presented by a party with a direct commercial interest in the outcome.
Answer first
A category captain is a supplier, usually the leading manufacturer in a category, that a retailer relies on for advice in managing that category. The FTC 2001 staff report defines it as "an outside firm, commonly a large supplier, to whom a retailer turns for advice in managing the category," and elsewhere as "a leading manufacturer of products in the category who acts as a primary advisor for the retail chain management of the category."
The captain influences which products are stocked, and how they are displayed, promoted and priced (Wright, Supreme Court Economic Review, 2009). Its recommendations are built from retailer point-of-sale data plus syndicated data.
Here is the strategic point: that entire evidence base is observational transaction data with no counterfactual and no stated reason. It can show that your SKU sells slowly. It cannot show why. The one evidence class the captain does not have, and that you can obtain independently in days, is the shopper own account of the decision.
How common is this, actually
Prevalence is better documented than most brand teams realise:
- Gooner, Morgan and Perreault (2011) surveyed 347 retail managers across 35 food and nonfood categories in US supermarket chains representing approximately 60% of US supermarket sales and 46% of all US grocery sales. Approximately half of the settings relied upon category captains.
- The FTC 2003 staff study of slotting allowances found that five of the seven retailers surveyed used category management contracts for some products.
- Among national brand manufacturers, becoming a category captain "is considered crucial for success and a top priority" (Gundlach and Loff, American Antitrust Institute working monograph, 2018).
The practice also has real antitrust history. In Conwood Co. v. United States Tobacco Co., the plaintiff alleged that the defendant used its category captain position to exclude a rival, including by removing competitor display racks and point-of-sale materials. The FTC has repeatedly noted the concern that captain arrangements could facilitate exclusion or collusion, while also recognising genuine efficiencies. The mainstream view is that these arrangements are usually legitimate and occasionally abused, which is precisely why the useful response is evidentiary rather than legal.
The evidence asymmetry
This is the capstone frame for the whole retail lane, and it is a structural description rather than an accusation.
You and the captain are arguing from the same class of evidence, and it is a class that cannot answer the question being decided.
| Category captain | You | |
|---|---|---|
| Data source | Retailer POS, plus syndicated data | Syndicated data you purchased |
| What it shows | Units, share, velocity, days of supply | The same things |
| Counterfactual | None | None |
| Stated reason from shoppers | None | None, unless you go and get it |
| Commercial interest in the outcome | Direct, competes in the category | Direct |
Both sides bring rate-of-sale. Neither side brings causation. The decision then goes to whoever has more of the retailer trust and more of the retailer data, and that is the captain by construction.
The way out is not to buy more of the same data. It is to bring a different evidence class entirely.
Why slow rate-of-sale is not evidence of low demand
The critical weakness in a velocity-based delisting case is that several completely different shopper realities produce identical POS data.
The out-of-stock literature makes this concrete. In the landmark worldwide study by Gruen, Corsten and Bharadwaj (2002, published by the Grocery Manufacturers of America), the average out-of-stock rate was 8.3%, and consumer responses to finding an item unavailable broke down as:
- 31% buy the item at another store
- 26% substitute a different brand
- 19% substitute a different item of the same brand
- 15% delay the purchase
- 9% do not purchase at all
Read the 26% row again. In the retailer POS file, a shopper who wanted your product, could not find it, and bought a competitor instead is recorded as a competitor sale. It is indistinguishable from a genuine preference win. Your rate of sale falls, the competitor rate of sale rises, and the captain data shows exactly the pattern that justifies giving the competitor more facings, which lowers your availability further.
Note also that the 31% who buy elsewhere are invisible to that retailer entirely. They are a cost the retailer is bearing without seeing, which is the most persuasive fact you can put in front of a category buyer, and it is unobtainable from any dataset either side is currently holding.
This is the same measurement failure described in private label switching research, where forced substitution looks identical to a deliberate value switch, and it compounds with the borrowed-volume problem in trade promotion, where a competitor promotional bump is read as a durable share gain.
The research that changes the conversation
You need evidence that answers the question the POS data cannot: among shoppers who did not buy your product, what happened?
That study is straightforward and, with AI-moderated interviews, fast:
- Recruit category buyers at that retailer, not your buyers. Screen on having shopped the category there recently. Screening on your brand is the sampling bias that guarantees a useless answer.
- Anchor on the most recent trip to that banner.
- Establish intent before availability. Did they come in intending to buy your product or the category, and what did they leave with.
- Measure the availability failure directly. Was it there, could they find it, was the size they wanted stocked. This separates a demand problem from a distribution problem, which is the whole argument.
- Quantify the leakage. Of those who did not buy yours, how many bought a substitute, how many bought nothing, and how many bought it elsewhere. The last group is the retailer own lost sale.
- Capture the reasoning in shopper words so the deliverable includes verbatim quotes alongside the numbers.
A Koji study runs all six structured question types inside the same conversational interview: yes_no for whether the product was found, single_choice for what was bought instead, multiple_choice for every barrier encountered, scale for purchase intent before the trip, ranking for what would make them switch back, and open_ended with automatic AI follow-up for the account of the shelf. The result is a quantified leakage estimate with the reasoning attached.
Two analysis cautions, because this study will be read adversarially by people who do not want its conclusion:
- Pre-commit to the analysis plan. A study run to defend a listing is highly exposed to confirmation bias, and the buyer will discount it if the specification looks chosen after the fact.
- Where the stakes justify it, blind the analysis or have a second analyst work the same extract. If two analysts disagree, that disagreement is diagnostic rather than embarrassing, as the many-analysts literature shows.
What to bring to the range review
The strongest position in a category review is not more velocity data. It is a claim the captain cannot make and cannot rebut from its own evidence base:
- "X% of shoppers who intended to buy this category at your stores left without buying anything." That is a retailer loss, not a supplier loss, and it reframes the meeting.
- "Y% could not find the product although it was listed." This converts a delisting conversation into an execution conversation.
- "Z% bought it at a competing retailer that week." This is the number that gets attention, because it is the retailer own leakage and nothing in their POS file contains it.
- The verbatim account of the shelf, in shoppers own words, which reads very differently from a slide of index numbers.
None of these require access to the retailer data. All of them come from asking a few hundred of the retailer own shoppers what happened. And unlike a syndicated subscription, this evidence is yours exclusively, which is the entire point when the alternative is arguing from a dataset your competitor knows better than you do.
Build your shelf case with Koji
Koji is the AI-native customer research platform that turns a category argument into evidence. Launch an AI-moderated voice or text interview study, reach several hundred of a specific retailer shoppers in days, and get quantified leakage figures, segment cuts and a one-click report you can take straight into the range review. Structured questions produce the numbers a buyer will act on; the conversational AI captures the shelf experience that no point-of-sale extract will ever contain.
Legacy tools cannot do this in the time a category review allows. Survey platforms cannot probe why a shopper walked away. Panel vendors sell you reach without depth. Traditional qualitative delivers depth at a sample too small to quantify leakage. Koji does both, in hours rather than weeks, with no research headcount required.
If a competitor is advising your retailer on your shelf space, the shopper own account is the one thing they do not have. Start a free study on Koji and bring it to the next review.
Related reading
- Price Pack Architecture (2026) - why retail price setters invert your pack-price ladder on your best-selling size.
- Shrinkflation Research (2026) - why scanner data and fairness experiments disagree, and which shoppers each one measures.
- Personalized Pricing in 2026 - what individualised prices do to the reference price your research assumes.
Frequently Asked Questions
What is a category captain in retail?
A category captain is a supplier, usually the leading manufacturer in a category, that a retailer relies on for advice in managing the whole category. The FTC 2001 staff report describes it as an outside firm, commonly a large supplier, to whom a retailer turns for advice in managing the category. The captain typically influences which products are stocked and how they are displayed, promoted and priced, including competitor products.
How common are category captain arrangements?
Common. Gooner, Morgan and Perreault (2011) surveyed 347 retail managers across 35 categories in US supermarket chains representing about 60% of US supermarket sales, and roughly half of those settings relied on category captains. A 2003 FTC staff study found five of seven surveyed retailers used category management contracts for some products.
Is being a category captain legal?
Generally yes. Category management arrangements are a mainstream and largely legitimate retail practice with recognised efficiencies. Antitrust concern arises in specific cases involving exclusion or collusion, as alleged in Conwood Co. v. United States Tobacco Co. The FTC has recommended focusing scrutiny on situations that may involve anticompetitive exclusion rather than treating the practice as inherently unlawful.
Why is slow rate-of-sale not proof that shoppers do not want my product?
Because point-of-sale data cannot distinguish low demand from poor availability. In the Gruen, Corsten and Bharadwaj worldwide out-of-stock study, 26% of shoppers facing an unavailable item substituted a different brand and 31% bought it at another store. Both outcomes lower your recorded rate of sale without any change in shopper preference for your product.
What research should I bring to a category review?
Bring evidence the retailer POS file cannot contain: the share of category shoppers who intended to buy and left with nothing, the share who could not find a listed product, and the share who bought at a competing retailer that week. These are retailer losses rather than supplier complaints, which is what makes them persuasive.
How fast can I run a shelf availability study?
With AI-moderated interviews, a few days. The constraint used to be moderator capacity, which made several hundred interviews impractical inside a category review window. Running them in parallel and analysing them automatically removes that constraint, so the study fits comfortably inside the review cycle.
