The Invisible Buyer, Part 2

An ad account presumes that whoever holds it is spending their own money and answering for the result. Objectives flex: reach, traffic, conversions, value, several running at once on separate budgets inside one account. The presumption about whose money it is does not flex. In that share of retail media, it is wrong. The money is a supplier's and the account is a retailer's. The work in between is real and continuous: deciding what gets promoted, buying the media that promotes it, answering for the result to whoever's money paid for it. It has no clean name because it was never designed as one job. It is four older disciplines, category management, trade marketing, shopper marketing, and digital ad buying, folded together by whatever org chart each company happened to have on the day retail media arrived there.

US advertisers put $60.32 billion into retail media in 2025, rising to $71.09 billion in 2026 on EMARKETER's December 2025 forecast, spread across more than 250 retailer-run ad businesses. Anyone selling into that market is selling to a counterparty whose accountability is divided by construction. The job has no agreed title and the reporting has no dominant line. Incentives stop at the channel, and the enterprise gets none.

Retail media arrived at each company on its own date

Organizations accrete the way buildings do, faster and with less ceremony. When new work arrives before an org chart can be redesigned around it, the work attaches to whichever team stands closest. It carries that team's metrics, its reporting line, its idea of what performing well means. Nobody decides this. Years later the arrangement reads as a decision.

Retail media arrived at each company on its own date, and those dates span more than a decade. Digiday traced the sequence in April 2025. Amazon's ad offering dates to 2012. Kroger, Target, Home Depot, and Walmart followed in 2017 and 2018. Nordstrom launched its network in 2021 and Sam's Club in 2022. CVS entered about five years before that reporting. Modern Retail dates Best Buy's rebuilt in-house operation, Best Buy Ads, to 2022 as well. The roster is still growing. The spread has not closed. Some companies are only now living through the arrival.

A staggered arrival means each company caught this work with a different org chart and a different amount of help. A retailer starting in 2013 had no settled name for the function, few peer examples to copy, and no hiring market to recruit from. A retailer starting now has all three. Both attach the job to whichever team stands closest, the way a building takes an addition on whatever wall will bear the load. That is why the titles vary by company and by year.

Four disciplines, four measures of a good outcome

Each of the four parent disciplines arrived with its own measure of a good outcome. Each was built to carry only itself.

Category management has the best-documented origin of the four. Brian F. Harris, then a professor at the University of Southern California, formalized it in the late 1980s. Harris founded The Partnering Group in 1990, and the eight-step process he developed with it spread as an industry standard through the Efficient Consumer Response initiative in the mid-1990s. Its objective, as the literature models it today, is shelf-space and assortment economics. Wu, Demirag, Xue, and Xu's 2024 model in the International Journal of Production Economics formalizes a discipline built to maximize a category's performance over any single brand's. The paper models the objective as currently understood. It is not a history. But the objective it models and the one Harris described three decades earlier are the same. The category is the unit that gets optimized.

The shape of the instrument is observable. Syndicated retail scanner files of the kind that reached commercial maturity in that same window aggregate millions of individual UPCs into roughly 1,100 product categories. The category is not only what the discipline optimizes. It is a field in the schema, the level at which the data was already cut before anyone was asked to manage it. That number conceals a choice. A measurement vendor set the resolution at 1,100, and the category P&Ls built on syndicated data inherited it, though nothing about groceries fixes that number. A discipline optimizes the unit its instruments resolve.

Trade marketing measures something else: money moving from supplier to retailer. Elberg and Noton's 2025 study in Management Science is rare in this literature: it works from actual transaction-level payment records, where most studies use surveys. In the two supermarket chains they observed, both owned by a single Chilean big-box retailer across a two-year panel, trade allowances were large, paid by the vast majority of suppliers, varied substantially by category, and behaved as substitutes for price promotion, which the authors read as suggesting trade allowances do not mitigate double marginalization. That is one retailer's payment data in one country. Within that scope it establishes something the other three disciplines do not touch: a large, near-universal flow of one company's money into another company's decisions.

Shopper marketing is the youngest of the three pre-digital disciplines and the most explicitly transitional. Silveira and Marreiros's 2014 literature review traces its emergence to a bridging role between brand marketing and retail category management. It grew to fill a gap between two older functions. A review synthesizes what the existing literature claims. It is the right source for where a function came from, and the wrong one for anything stronger.

Digital ad buying arrived last, carrying measures built somewhere else entirely: auction mechanics, attribution windows, return on ad spend. The first three disciplines never needed the concept of an attribution window. All three now have to argue about one. The job inherited four sets of accounts at once and no instruction for reconciling them.

The process can absorb a co-op dollar. It has nowhere to put one measured somewhere else.

The eight-step cycle that became the industry standard runs: define the category, assess its role, evaluate its performance, set objectives and targets, devise strategies, select tactics, implement the plan, review. That version was codified in the 1990s and diffused through the ECR consortia. It remains the closest thing the discipline has to a statement of what it thinks its own job is.

Media is not missing from the eight-step process. It was bought in 1996 and long before it. Circulars and free-standing inserts, end caps and shelf talkers, co-op radio and television, much of it funded by the same manufacturer money that funds retail media now. Co-op advertising is retail media's direct ancestor, and the eight-step process handles it the way it handles every other lever: as a tactic selected in service of the category. The process has no step for media as a business. It only knows media as a tactic. A revenue line the retailer monetizes and reports on separately appears nowhere in it, and neither does a supplier's proof-of-performance being reconciled against an attribution window set by a third party whose measurement the category manager does not control. The 1996 process could absorb a co-op dollar because that dollar was accounted for inside the category P&L. It has nowhere to put a dollar accounted for somewhere else.

A discipline's canonical process is its most durable statement of scope. When work arrives that the process has no step for, that work does not route itself somewhere sensible. It goes to whoever is standing closest. Then it gets measured by that person's existing metrics, because those are the only metrics the org chart has for them.

Four companies, four resolutions, one instinct

Julie Liu's title changed twice while the job underneath it barely changed at all. At Ghirardelli she moved from shopper marketing into retail media and then into a broader ecommerce leadership role. She offered her own gloss on why the boundary keeps moving, in the same February 2025 Forbes piece cited in Part 1: "Retail media is just that. It's retail plus media."

Priya Shenoy's team at Tillamook renamed itself mid-stream too, to "omnishopper," on the reasoning that the label described how people were actually shopping. She is candid that the renaming settled nothing underneath: "are you customer marketing or shopper marketing?" Michelle Boisvert, who leads shopper marketing and media at Georgia-Pacific, described her team's parallel rebrand from shopper marketing to omnichannel marketing, undertaken in part so the company's sales reps would understand what omnichannel growth looks like.

A fourth company in the same piece solved it structurally instead of nominally. Alicia Ponziani, an ecommerce manager at Premier Nutrition, described a recent restructure that placed the ecommerce team inside sales, split into an Amazon team and an omni team, with the omni side working closely with sales leads to plan online and in-store media together. Four companies, four separate resolutions, one instinct. The old label no longer fits and nobody has agreed on the new one. The decisions were unrelated. The reporting is not, since one journalist found all four for a single article. Each company discovered separately that the job underneath its label had already changed.

The org chart split, on both sides

On the brand side, the function straddles two departments. A 2021 Path to Purchase Institute survey, cited in the same Forbes piece, found 53% of shopper-marketing teams reporting into marketing and 33% into sales, with the rest scattered elsewhere. The survey's sample size and methodology are not published alongside the figure. Read it as directional.

The retailer side now has a direct measurement, and it shows something stronger than a split. Bain & Company and EMARKETER surveyed 73 US sell-side retail media leaders between 11 July and 21 August 2025, most at director to EVP level, and asked which leader their retail media team primarily reports to. Among the 36 who answered on operating model and team structure, the most common answer was the head of media or advertising, at 28%. Then chief marketing officer at 22%, chief digital officer 11%, chief financial officer 8%, head of ecommerce 8%, chief revenue officer 6%, chief merchandising officer 3%, chief executive 3%. The count comes to eight named destinations. All eight fall below thirty per cent. The report's own reading: RMN teams report into a wide range of executive functions with no single dominant owner. Two departments dividing a function is a fault line. Eight destinations with no majority is an absence of settled ownership, three to eight years after most of these networks launched.

Two retailers have addressed the same seam by consolidating upward. Speaking at Shoptalk in March 2025, as reported by Modern Retail, Nordstrom's VP of media Aaron Dunford, whose remit now spans enterprise media, retail media, and media enablement, gave a blunt reason for the restructure: a siloed org structure produces a disparate experience for the customer. Sam's Club made a structurally similar move, naming its first chief experience officer, Diana Marshall, in February 2025, with retail media sitting alongside membership, product, design, and marketing under one executive. Harvey Ma runs the club retailer's media business, the Member Access Platform, on four decades of first-party membership data. Read against that scatter, Nordstrom and Sam's Club are one available response, not the industry's direction of travel. Brands rename the function to describe what it has become. Some retailers move it up the org chart until it sits under someone senior enough to arbitrate between the merged disciplines. On the survey evidence, neither has become the norm. All of them are responses to inheriting a job nobody drew, and none of them changes what the job is.

The incentives follow the same pattern. Asked to describe their network's internal incentive structure, 53% of the same respondents called it revenue-focused. Thirty-six per cent had cross-functional KPIs spanning shopper, brand, and ecommerce. Twenty-eight per cent had incentives aligned with merchandising, the function whose category economics retail media spending most directly moves. Nineteen per cent described their own incentive structures as inconsistent or unclear. All four figures come from the subset of 36 who chose to answer on operating model and team structure. Read them as a self-selected group.

The asymmetry follows from how each side experienced the arrival. A retailer experiences retail media as a launch: a dated decision with a budget and an announcement attached, and therefore a moment to reorganize around. Nordstrom launched in 2021 and Sam's Club in 2022, on the Digiday timeline above, and both had restructured around the function within three or four years. A supplier experiences the same phenomenon as a rolling series of other people's launch dates, one per retail partner, none coordinated, none theirs to schedule. There is no moment to reorganize around. A team can be renamed on any Tuesday. Reorganizing wants an occasion.

The job description is the merger, written down

A job posting is an organization's attempt to write down a role it has no word for. In April 2025 Digiday reported on what that produces, and the reporting is direct about the arithmetic. Larisa Dumitru, head of commerce for Europe at GroupM, on postings demanding fifteen to twenty years of retail media experience: "you can't be asking for 20 years of retail media experience because that just doesn't exist." She is right on the timeline. Even dated from Amazon's 2012 start, the generous reading, the discipline is still younger than the seniority being demanded of it.

The postings themselves read as inventories of the merger. One retail media specialist listing from Crayola, current at the time of that reporting, asked candidates to run campaigns across Amazon Ads, Walmart Connect, and Target Roundel, three separate retailer ad platforms each with its own console and reporting conventions, plus manage paid search and work across sales, brand, and ecommerce teams internally, on a minimum of three years' digital marketing experience. That is category management's cross-functional coordination, trade marketing's supplier interface, shopper marketing's channel logic, and digital ad buying's platform mechanics, addressed to one person, at a level of seniority the four parent disciplines would each have called junior.

Some employers have concluded the specification is the problem. PriceSpider's chief revenue officer, Florencia Schiavon, told Digiday the company deliberately left a retail media manager description open-ended. A broader posting attracts candidates with more varied backgrounds. The constraint is not the supply of talent but how employers frame the need. The pipeline is only now catching up. Eric Brackmann, VP of commerce media at the commerce tech company Koddi, said it had been about six months since the company started hiring college graduates who had been through retail media training programs. Until that year those programs did not exist. Parbinder Dhariwal, who runs CVS Media Exchange, described a hiring posture built around compromise: hire the adjacent parts and train the rest. He offered his own career as the argument. He came out of data and insights in digital advertising and did not understand retail when he entered retail media. His division's hiring plans at the time ran to a 25% to 30% increase, weighted toward product roles.

Sometimes the composition runs the other way. Lisa Valentino left Disney, where she was an EVP, to become president of Best Buy Ads in October 2024: a retailer's media business staffed at the top by an executive imported from entertainment media. The fourth discipline came in through the hiring channel.

The same Bain and EMARKETER survey found the difficulty is not only technical. Asked where sourcing or developing talent is hardest, 48% named data and platform expertise and 48% measurement and analytics. But 36% named cross-functional leadership and alignment, ahead of sales, strategy, and marketing. The hardest thing to hire for is partly the seam itself. When a role has no coherent predecessor and no training pipeline, recruiting becomes an act of composition. Employers assemble a person out of adjacent parts the same way the job itself was assembled out of adjacent disciplines.

The objection, and what it does not reach

The strongest objection is that this is an ordinary job, and the merger is a story being told about it after the fact. The people closest to the hiring make that case directly. They are not naive about the work. In the same Digiday reporting, Brandy Alexander, founder and managing director of the performance marketing and omnichannel commerce firm Tandemtide, argued that anyone with a foundation in media buying and an understanding of the digital shelf can do this job given training. Kiri Masters, quoted in the same piece as an analyst, put a number on it: roughly three months of training and hand-holding was enough to make someone with baseline media-buying skills a passable junior buyer at a single retailer. Dumitru lands in the same place, describing the work as media and sales and something people can learn. Hiring managers in that reporting decline the phrase talent shortage. They call it a training shortage instead.

The ground they are claiming is real and should be conceded. The individual competencies are learnable, on evidence, in months. The four-discipline account agrees. Every new function in every industry gets assembled out of adjacent predecessors. That is how organizations absorb novelty, and an objector can point to any number of fields that look in retrospect as though they put themselves together the same way. Treating an ordinary process of accretion as a design failure is retrospective tidiness. A discipline six years old that has already produced college training programs looks healthy by the normal measures.

What the objection leaves untouched is accountability. Training settles whether one person can run campaigns across three retailer consoles. It leaves untouched which department's definition of performance that person answers to. That is not a skill held by an individual at all. It is a property of the org chart, and no amount of onboarding relocates a reporting line. Alexander and Masters are describing the executional layer, where the evidence says they are right. The reporting-line and incentive data describe a different layer. A trained buyer answering to a chief marketing officer and an equally trained buyer answering to a chief financial officer will do the same work against different scoreboards and both be performing well by their own lights. Just over a third of these networks report incentives that span shopper, brand, and ecommerce. Fewer than a third have incentives aligned with merchandising at all.

One note on the sourcing, since the claim of independence does real work. The three pieces of reporting are separate: three reporters at three outlets, published within about eight weeks of each other in 2025. There is one crossover. Kiri Masters, who wrote the Forbes piece, appears as a quoted retail media analyst in the Digiday piece. It is an overlap of people, not of reporting.

Whose advice the category manager trusts

Category captainship adds a second, harder-to-see fault line: whose advice a category manager is supposed to trust. The theory runs in three directions at once. The efficiency case comes first. Subramanian, Raju, Dhar, and Wang's 2010 model in Management Science finds that handing one supplier authority over a whole category can lower coordination costs and improve category outcomes, because that supplier holds category-wide knowledge the retailer lacks. Delegation to a well-informed party has a real economic argument behind it. The suspicion case has a name. Timoumi, Esseghaier, and Koçkesen model captainship as a strategic information-transmission mechanism, in Production and Operations Management (2023): a captain's nonprice decisions can reveal information that lets the retailer price better, so delegation can make the retailer better off rather than simply inviting discounting. In the one category where transaction-level data has made a direct empirical test possible, ready-to-eat cereal, the record is more forgiving than the theory. Kim, Shen, Basuroy, and Beldona's 2023 study found captains generally grew the whole category, with no disproportionate favor to their own products. That is one study of one category. It counterweights the game-theoretic suspicion. It does not rebut it. All three hold at once. Delegation is efficient, the incentive to self-favor is structurally present, and where it has been directly tested it usually is not acted on. A category manager operates inside that uncertainty permanently, without knowing in advance which category, which captain, or which year is the exception.

Distrust changes how carefully the data gets read

Trust changes not just what a category manager believes, but how carefully they read in the first place. This is a different mechanism from the measurement disagreements in Part 1. Those were a reporting-pipeline problem: two systems producing different numbers for the same campaign. This one is cognitive. Prince, Kwak, and Priporas's 2019 study in the Journal of Retailing and Consumer Services, on what they call the Diogenes effect, found that when retail buyers searched for a vendor with an ethical orientation, they processed information more centrally — more intensively, not more shallowly. Whether distrust of a vendor thins processing in the opposite direction is not what the study tested, and the extension to platform dashboards is the essay's own inference. The study does not show the mirror image, that a distrusted source gets only a shallow read. That step is an inference, and a weaker one, since a manager who distrusts a dashboard may scrutinize it harder, not less. The study examined general buyer-supplier relationships. It did not test platform dashboards. A category manager reading one is a step beyond what it measured, though a plausible one, since a platform's reporting is itself a claim from an interested party. If it holds, it stacks a second pressure on top of the measurement problem Brian Monahan described in Part 1. Even where the underlying data is accurate, the working relationship with whoever reports it shapes how carefully the report gets read, and the direction of that effect is not settled.

Four consequences for anyone selling into this market

The first concerns who is being addressed. Whatever the objective mix inside it, an account resolves to one payer and one accountable party. On the retailer side no reporting destination holds more than 28% of these teams. The person holding the login is one of at least eight possible principals, and which one it is varies by company. Knowing the segment does not tell you. Treating "the advertiser" as a fixed role mistakes a variable for a constant. The survey evidence says that role exists in no dominant form.

The second concerns what the reporting satisfies. Fifty-three per cent of these teams are measured on channel revenue, and twenty-eight per cent on anything aligned with merchandising. Reporting expressed as return on ad spend therefore answers the incentive of the person reading it and leaves the supplier's category economics unaddressed. Both parties can be satisfied by the same report while the number in it answers only one of their questions.

The third concerns where the difficulty concentrates. Amazon and Walmart between them take roughly nine dollars in ten of US retail media spend, and they are also the two most organizationally settled counterparties in the market. The several hundred networks holding the remainder are where reporting lines are least fixed and arrival dates most recent. The largest two counterparties are the least representative two.

The fourth concerns the counterparty's own hiring. Thirty-six per cent name cross-functional leadership and alignment as a top talent difficulty, ahead of sales, strategy, and marketing. The role that would make a cross-functional product legible inside a retailer is the role retailers report the most trouble filling. Adoption friction that presents as product complexity may instead be an absence of anyone positioned to absorb it.

Two limits on the argument

The evidence here is overwhelmingly American. A US survey, US job postings, US retailers at a US conference. The two non-US anchors pull in the same direction without settling anything. Elberg and Noton's trade-allowance panel is Chilean, and the most pointed testimony about the timeline comes from GroupM's head of commerce for Europe. Whether the four-way merger takes this shape where retail concentration, trade-funding norms, and labour markets differ is a question none of these sources can answer. A reader outside the US should treat the pattern as demonstrated at home and untested abroad.

The second limit is definitional. The four-discipline account describes companies that already had category management, trade marketing, and shopper marketing in place before digital ad buying arrived. A digitally native marketplace did not. Whatever its category function looks like, it did not form by merger with three pre-digital predecessors. The essay is about inheritance. It applies only where there was something to inherit.

The two limits turn out to be one limit seen twice. The spending figures above are American. eMarketer's forecasting put worldwide retail media at $140 billion, with much of the non-US total running through Alibaba, JD.com, Meituan, and Mercado Libre. Those are marketplace-native businesses. They never carried three pre-digital disciplines, so there was nothing to merge. The same asymmetry appears inside the American figure. Amazon Ads took 79.7% of US retail media spend in 2025 and Walmart Connect 8.0%, with Target Roundel third at 1.5%, and EMARKETER expects Amazon and Walmart between them to absorb 89% of the incremental spend in 2026. The field is numerically crowded and financially concentrated at the same time. The places where this argument has least purchase are, on the money, the largest. The concentration conceals a mismatch: dollars and headcount are not distributed the same way. The population this essay describes is the org charts on both sides of those several hundred networks and the supplier teams answering to them. A single marketplace taking four-fifths of the spend does not employ four-fifths of the people arguing about what to call this job. The scope limit is real. It is about which organizations the inheritance story fits, not about how many practitioners live inside it.

The merger holds

The job lacks a clean name because it is a merger of four disciplines with four original mandates, inherited unevenly. The arrival was staggered across more than a decade and caught whatever org chart happened to be standing at each company when it came. A process document codified before the fourth discipline existed still governs. The evidence shows up in three places that do not depend on each other. In the renamings, where Liu, Shenoy, Boisvert, and Ponziani each arrived separately at the same instinct. In the org charts, where a 53/33 marketing-versus-sales split on the brand side sits against a retailer side with eight reporting destinations and no owner above 28%. In the postings, which ask one person for four disciplines' worth of range and then demand more years of experience than the discipline has existed. The objection that all of this is ordinary and learnable is correct about the executional layer and silent on the accountability one. Category captainship and the Diogenes effect are two further pressures on top of the merger. One is about whether the advice reaching a category manager can be trusted. The other is about whether trust itself changes how carefully a report gets read. What this calls for is precision about the discipline. This is a clearly-defined-in-pieces job, each piece specified for a narrower predecessor role, assembled by accretion. Nobody designed it. The condition is familiar to anyone who has surveyed an old building before touching it. It stands, every element was competently detailed for the use it was drawn for, and no one has surveyed the whole since the last three alterations. Part 3 takes up the money on the same terms. Whose it actually is, and what strings travel with it from a supplier's joint business plan into a retailer's ad account.

References