Who runs the supermarket?

What aggregating demand means

A single consumer has no power at all. They buy one pack of yoghurts a week. No manufacturer changes the recipe for them and no chain changes its range.

Power appears when someone gathers millions of these small purchases and manages to speak on their behalf. The manufacturer did it with the brand: millions of people asking for the same name in different stores. The retailer did it with stores: millions of people walking down the same aisle every week. In each era, whoever aggregates sets the terms for whoever does not.

The research confirms that reading the market pays

In the 1990s, two groups of researchers, Ajay Kohli and Bernard Jaworski on one side, John Narver and Stanley Slater on the other, defined what a market-oriented company is^{12, 13}. It is a company that gathers information about customers and competitors, circulates it internally and acts on it.

Ahmet Kirca, Satish Jayachandran and William Bearden brought together 114 studies on this idea, with more than 61,000 responses from companies in several countries⁹. The relationship between market orientation and performance is positive and consistent: an average correlation of 0.32. In the social sciences, that is a medium and very stable effect. Another meta-analysis, by Rodriguez Cano and colleagues, reached a similar conclusion¹⁴.

This says something simple: whoever reads the customer better earns more. The question for this analysis is who manages to do that reading in each period, and who ends up depending on someone else's reading.

The first shift in power was made by law

The story is usually told as a natural evolution, but the first big change had help from the legislator. Until the 1960s, in several countries, the manufacturer could impose the price at which its product was sold in the store. In 1964, the United Kingdom abolished that practice¹⁰. From then on, the store could use the price of brands to attract customers, and competition between chains came to be fought with the manufacturer's product.

Who gained from the change?

In 1995, Paul Messinger and Chakravarthi Narasimhan tested the often-repeated idea that power in the grocery channel had shifted from the manufacturer to retail¹¹. They analysed the profitability of both sides over several years in the United States. Retailers did not become more profitable at manufacturers' expense. The authors suggest that the one who gained was the consumer, with lower prices and more choice.

It is the first sign that the war between manufacturer and retailer has a third side.

How the brand gathered demand

Before the chains, the store was small, independent and local. There were tens of thousands of grocers in each country, each with its owner, its tab and its range. None had the strength to impose terms on a national manufacturer. A manufacturer alone could not visit them all either.

The brand solved both problems at once. Radio, newspapers and then television allowed the manufacturer to speak directly to the consumer, over the shopkeeper's head. When the customer walked into the grocer's asking for a brand by name, the shopkeeper had no choice. If he did not have it, the customer went next door.

This is what gave the manufacturer power: not the factory, but the fact that it was the one gathering dispersed demand around a name. The shopkeeper became a point of passage.

The quote that shows the era, and the quote that was made up

Henry Ford wrote, in his 1922 autobiography, that customers could have the Model T in any colour, so long as it was black¹⁸. The quote is real and portrays a time when whoever mass-produced could dictate supply.

The other famous quote attributed to Ford, "if I had asked people what they wanted, they would have said faster horses", does not appear in any of his writing. Quote Investigator first found it in 1999, in a cruise magazine and with no link to Ford; the first attribution to Ford is from 2001, more than 50 years after his death in 1947¹⁹. It is used to argue that the consumer does not know what they want. We come back to it in the section on numbers without a source.

The brand still matters, and matters for decades

It is tempting to say the manufacturer's era is over. The research says it is not over: it has slowed down.

Bart Bronnenberg, Sanjay Dhar and Jean-Pierre Dubé studied the shares of consumer goods brands in dozens of US cities¹⁵. In many categories, the leading brand in a city is the one that got there first, sometimes more than half a century ago. The advantage of having gathered demand early is passed down from generation to generation, through family habits.

It is the first sign that the aggregation built by the manufacturer does not disappear when the store gains strength. It stays stored in consumers' memory, and that is why the store cannot replace every brand with its own.

The war moved into the consumer's head

In 1969, Jack Trout wrote an article in an industrial trade magazine that would lead, in 1981, to the book Positioning: The Battle for Your Mind, with Al Ries²⁰. The central idea is that positioning is not a property of the product. It is what the brand manages to do in the customer's mind, compared with its competitors.

Read in the light of this analysis, the book describes the peak of the manufacturer's era and, at the same time, the beginning of the end. The battlefield was no longer the factory or the store: it was the consumer's head. Whoever managed to take a place there gathered demand. Ries and Trout wrote for manufacturers, but the principle holds for any side of the triangle. When, later, the store managed to take that place with its own banner, it started playing the same game, with the advantage of also owning the shelf.

But each euro of advertising returns less

Raj Sethuraman, Gerard Tellis and Richard Briesch brought together 751 short-term and 402 long-term elasticities, from 56 studies on advertising¹⁶. The average elasticity is 0.12 in the short term and 0.24 in the long term. In practice, doubling advertising spend raises sales by about 9% straight away and about 18% over time. The elasticity has been falling from decade to decade.

The likely cause is fragmentation. When there were two television channels, a campaign reached almost everyone. Today, attention is split across hundreds of channels and thousands of screens. The brand still speaks to many, but no longer to everyone at once, and that is what gave it power.

The manufacturer started paying the store to reach the consumer

As advertising returned less and chains grew, the manufacturer's money changed destination. In 1995, according to an annual survey by Carol Wright Promotions, promotions paid to the store already accounted for 50.7% of the promotion and advertising budget of consumer goods manufacturers in the United States. Consumer promotions got 24.2% and advertising 25.1%¹⁷.

The money that used to go on talking to the consumer started going to the store: launch discounts, leaflets, gondola ends, listing fees. It is the clearest sign that the manufacturer had stopped being the one gathering demand.

The side effect of promotions

Carl Mela, Sunil Gupta and Donald Lehmann followed purchases of a packaged consumer good for more than eight years²¹. Promotions made consumers more sensitive to price and to promotions. Advertising did the opposite: it reduced price sensitivity.

The manufacturer who swaps advertising for promotions is teaching the consumer to wait for the discount. It gains volume straight away and loses part of the reason the consumer paid more for the brand. The one who benefits is whoever controls the discount, and whoever controls the discount is the store.

Portugal is an extreme case. In 2018, Nielsen measured 46% of consumer goods sales in Portugal as made on promotion, against 16% in Spain²². In 2024, 63% of the items in Portuguese chains' leaflets were on promotion, with an average discount of 28.8%²³. A market trained for years to buy on promotion is a market where the brand is worth less and the store is worth more.

Fiat lux: how it began in Portugal

Modern retail in Portugal is little more than fifty years old. The story is told in three layers, which coexisted for some years: the wholesaler and the cash and carry, the first supermarkets and, finally, the hypermarkets.

First, the wholesaler

Until the 1980s, consumer goods reached the consumer through a long chain. The manufacturer sold to the wholesaler, the wholesaler sold to the grocer, the grocer sold to the customer. In 1990, Portugal had 19.2 stores per 1,000 inhabitants, against 11.3 on average in the European Community²⁶. In 1993 there were 173,300 retail businesses, and only 35,752 had employees; half the stores were under 50 m²²⁶.

In this world, the one who gathered the grocers' demand was the wholesaler. From him came the cash and carry: a wholesale warehouse where the shopkeeper went to choose, paid in cash and took the goods away. Recheio was born this way, in 1972, in Figueira da Foz²⁷. In 1988, Jerónimo Martins bought 60% and, the following year, the rest^{28, 27}. In February 1990, Makro opened the first store in Portugal in Alfragide²⁹. Even today, Recheio sells about half to the food service sector and half to small grocers³⁰. The cash and carry is the link that connects modern retail to traditional retail.

Then, the supermarkets

The first generation of supermarkets arrived in the 1970s. The Brazilian group Pão de Açúcar, founded by an emigrant from Guarda, opened the city's first modern supermarket in the early 1970s, on Avenida Estados Unidos da América, in Lisbon³¹. In 1975, after the revolution, the State intervened in the company. In 1985 it had 49 stores, and in 1987 it was the national supermarket leader, followed by Modelo, from Sonae, by Pingo Doce, from Jerónimo Martins, and by the Inô supermarkets³¹.

Inô, from Grupo Inovação, reached 54 stores³². The same group launched the Feira Nova banner in Braga in 1989, with hypermarkets and supermarkets. In 1993, Jerónimo Martins bought the first four Feira Nova stores from Grupo Inovação and, around that time, the Inô supermarkets^{33, 32}. Between 2008 and 2010, Feira Nova disappeared and the stores became Pingo Doce³³.

Jerónimo Martins itself had returned to retail in 1978: Pingo Doce began operating in 1980, with Belgium's Delhaize as partner from 1985 and the Netherlands' Ahold from 1992²⁸. In 1989 it bought 15 supermarkets from Pão de Açúcar²⁸. There were also regional chains, such as Expresso, from the Entreposto group³².

Concentration began like this: not with new stores, but with chains buying chains. Even so, it was a world of grocers. In 1991, out of 38,910 food outlets, only 443 were supermarkets or hypermarkets: 1.14%³².

Finally, the hypermarkets

The change, when it came, was very fast. The country's first hypermarket opened in Cascais in 1973, under the Pão de Açúcar banner, and is now an Auchan. But the wave began in 1985, with Continente in Matosinhos³⁴. Then came Makro in 1990, Intermarché in 1991, Lidl in 1995, Aldi in 2006 and Mercadona in 2019²⁹. The hypermarkets of 1985 to 1989 were built on the outskirts of cities, next to the roads; from 1991 they started coming with shopping centres³⁵.

The Nielsen figures show the speed²⁵. In 1984, small grocers were 75% of food stores and made 32.5% of sales. Ten years later, they made 13.2%. In 1987, hypermarkets had 5.4% of sales. In 1994 they had 40.4%, with less than 0.1% of stores. In less than a decade, a few dozen stores came to sell almost half of what was sold in the country.

When three chains hold 62% of the market², a consumer goods manufacturer left out of one of them loses about a fifth of its potential consumers in one go. The reverse is not true: a chain that stops stocking a brand loses only some of the customers in that category, and many settle for the alternative on the shelf.

This asymmetry is at the heart of the book Store Wars, by Judith and Marcel Corstjens, from 1995⁷. In the 2012 edition, continued by other authors, the idea is put this way: the mind and the shelf used to be fought over between manufacturers, and the battle became one between retailers and manufacturers, who have to see retail as a separate and dominant force³⁶. The manufacturer fights for space in the consumer's mind, the store fights for space on the shelf. For decades, the first front decided the second. With concentration, the second came to decide the first.

The first weapon: price

The abolition of manufacturer-imposed prices, in the United Kingdom in 1964, gave chains the right to sell brands at whatever price they wanted¹⁰. From then on, the manufacturer's brand was used by the store to attract customers, often below the normal margin. The manufacturer lost control of the price at which the consumer saw its product.

The second weapon: time

Whoever pays at 30 or 60 days and gets paid by the customer in cash is financed with the supplier's money. In its 2024 results, Jerónimo Martins reported negative working capital of €4,062m. The suppliers line alone was equivalent to 52 days of sales³. In practice, the chain gets paid by customers long before it pays suppliers, and that difference finances stores, stock and expansion.

It is legal and it is the business model of food retail all over the world. But it shows who has the power. In 2012, with the Confederation of Portuguese Farmers (Confederação dos Agricultores de Portugal), Pingo Doce cut the payment term for 402 national fresh produce suppliers, from an average of 60 days to as little as 10 days, for 12 months³⁷. The conditions were also the chain's: at least five years of relationship, 95% of orders fulfilled and Pingo Doce accounting for at least half of the supplier's business. It was presented as help for small producers, and it was. But the one deciding the term, and for whom, was the chain.

The legislator eventually stepped in. Since 2010, large companies have had to pay micro and small food suppliers within 30 days for perishables and 60 for everything else³⁸. Decree-Law 76/2021, which transposes the European directive against unfair trading practices in the food supply chain, sets maximum terms of 30 days for perishable products and 60 days for the rest, with fines that can reach €2.5 million^{39, 40}.

The third weapon: access to the shelf

To be on the shelf, the manufacturer pays. It pays listing fees for new products, contributions to leaflets, top positions, store anniversaries, year-end rebates. The US Federal Trade Commission studied listing fees in 2003 and concluded that they were common in several consumer goods categories, with amounts that varied greatly by product and by region⁴¹.

What the shelf costs

In Portugal there are no public rate cards. The amounts sit in the annual agreements between each supplier and each chain, and the most detailed picture is still the Competition Authority's 2010 report on trade relations between food retail and suppliers⁴². The total discounts asked of suppliers reached 35–45% for rice in one large retail group, 45–55% for pasta, flour and cereals in another, and between 35–45% and 65–75% for coffee, drinks and dairy in a third. Gondola ends, gondolas and islands were charged through monthly debit notes, and there were discounts backdated to January. A study by Católica Porto describes the parts of a Portuguese supply agreement: rebates, financial discounts, listing costs, campaigns and penalties⁴³.

In the United Kingdom there are rate cards, and The Grocer published those of three chains⁴⁴:

One supplier was asked for £50,000 for a branded display. The rule of thumb in the sector, according to the same source, is not to spend more than 10% to 15% of the sales you expect to make with a chain. Asda also started charging suppliers for access to customer sales data: tens of thousands of pounds per category⁴⁵. In the United States, the FTC found in 2003 listing fees between $2,313 and $21,768 per product, per chain and per metropolitan area, higher for frozen and chilled goods⁴¹.

There is a second, less visible form of access to the shelf: the category captain. In many chains, the leading manufacturer proposes what goes into the category, where it sits and what promotions it runs. An investigation by a US consumer association found in 2016 that Nestlé was category manager for frozen desserts in 22 of the 25 largest supermarkets, and the American Antitrust Institute warned in 2018 that the model can block the entry of rivals⁴⁶. It is an alliance between the big player on each side, and the one left out is the small manufacturer who could bring something new to the consumer.

Those who pay rarely complain. According to the European Commission's 2013 Green Paper on unfair trading practices, 87% of suppliers do no more than talk to the customer when they suffer these practices, and almost two thirds of them say it is out of fear of retaliation⁴⁷.

The fourth weapon: join forces even further

When the chains in each country are already concentrated, they join across countries. International buying alliances negotiate on behalf of several chains from several countries at once. The European Commission's Joint Research Centre devoted a report to them in 2020, which describes how they work and concludes that the effects on suppliers are ambiguous⁴⁸. It is the same logic as always: whoever aggregates more demand negotiates better.

But the store does not rule alone either

Three pieces of data show the limits of this power.

The first already appeared in chapter 1. Messinger and Narasimhan found no transfer of profitability from manufacturer to store¹¹. The store gained bargaining power, but much of it was passed on to the consumer in price.

The second comes from the European Commission. A 2014 study on the impact of modern retail on choice and innovation found no evidence that concentration had reduced the choice available to consumers between 2004 and 2012; at local level, choice even increased⁴⁹.

The third is in the Competition Authority's cases, and has its own section at the end: there were cases in which chains and suppliers lined up, not against each other, but against the consumer.

How the store became a manufacturer

Private label began as the cheap version: plain packaging, low price, acceptable quality. Today, chains have several own ranges, from basic to premium, and some have private labels the consumer already recognises by name.

In Portugal, the share of private label in the sales of large supermarkets went from 31.7% in 2011 to 47% in 2024, the highest value in the INE series¹. At Continente, private label already accounts for 33% of sales, against just over 20% at the start of 2022⁵². In Spain, it has reached half of consumer goods purchases, driven by Mercadona²⁴.

Private label is the proof that the store started gathering demand. The consumer trusts the banner and agrees to buy a product with its name, often without knowing who makes it. The store took what used to be the manufacturer's exclusive asset: trust.

Why the store wants private label

Private label gives the chain a higher percentage margin, but not necessarily more money per product. Because the price is lower, the absolute margin can be smaller than that of a manufacturer brand. It is one of the questions that comes up most in the comments on the reels, and the answer depends on the category.

Kusum Ailawadi, Koen Pauwels and Jan-Benedict Steenkamp studied the relationship between private label use and store loyalty⁵³. They found an inverted U curve. Customers who buy private label in moderation are the most loyal to the store. Those who buy almost only private label are less loyal, because they are mainly price-sensitive and switch to whoever is cheaper.

This has an important consequence. Beyond a certain point, more private label stops holding on to the customer and starts letting them go. The store needs manufacturer brands to have reasons for a visit that the competitor does not have.

The consumer applying the brakes

The 2025 Centromarca and Kantar study shows that private label has slowed down in Portugal and that, in more than half of the categories where it is gaining share, it has reached a breaking point, beyond which it stops growing⁵¹. The same study shows that 84% of shoppers use both types of brand. The consumer does not choose between the store and the manufacturer. They use both, depending on the category.

In Spain there is an episode that readers of the reels' comments know well. In December 2008, Mercadona removed almost 800 manufacturer brand lines from its stores, among them products from Nestlé, Calvo and Pascual, to make room for its private label⁵⁴. A case study from Universitat Pompeu Fabra records that the change lost it some customers⁵⁵, and a 2015 academic article writes that many of those brands later returned to the shelves, to protect the chain's image⁵⁴. We did not find how many.

Laurens Sloot and Peter Verhoef studied what happens when a store removes a brand⁵⁶. The reaction depends on the brand. When a strong brand with no close substitute goes, some customers change store to keep buying it. When a weak brand goes, almost everyone accepts the alternative. It is the consumer's veto at work: the store can remove whatever it wants, but the consumer decides whether to stay.

What the store controls

A chain's buyer decides the range, each brand's space, shelf height, gondola ends, islands and promotions. That is why the manufacturer pays for the central aisle, gondola ends and aisle ends, one of the topics that comes up most in the comments on the reels.

Xavier Drèze, Stephen Hoch and Mary Purk ran real experiments in 60 stores of a US chain, across eight categories⁵⁹. Moving a brand from the worst to the best vertical position raised that brand's sales by 39% on average. Moving it horizontally had a much smaller effect. Adding facings returned less and less. In the authors' words, a couple of facings at eye level did more than five facings on the bottom shelf.

It is proof that the shelf moves sales. If you are a manufacturer, you have known this for a long time.

What the store does not control

Pierre Chandon, Wesley Hutchinson, Eric Bradlow and Scott Young had shoppers choose from shelves while a device recorded where they were looking⁵⁷. They brought together 348 shoppers and more than 8,000 observations.

Space changes attention. Going from four to eight facings makes many more people notice the brand: from 60% to 76%. But choice barely moves: from 7.7% to 8.8%. A central position helps little too: 9.0% against 7.7%.

What decides choice is habit. Those who already buy the brand choose it 49% of the time. Those who have never bought it choose it 1% of the time. The shelf can get the consumer to see. It cannot get the consumer to change.

The difference between planning the category and planning the brand

The most quoted figure in the sector is that "76% of decisions are made in store". It comes from a 2012 study by POPAI, the international point-of-purchase advertising association, which has a direct interest in the result and did not publish its method⁶⁰. We come back to it in the section on numbers without a source.

The reference study is older and better documented. Jeffrey Inman, Rosellina Ferraro and Russell Winer analysed about 34,000 purchases by 4,200 shoppers in 14 US cities, comparing what they said they wanted to buy on the way in with the receipt on the way out⁵⁸. 59% of purchases were unplanned. As early as 1967, Kollat and Willett had found 50%⁵⁸.

But the same study shows another figure, far less quoted. Only about 3% of shoppers switched, inside the store, the brand they had decided to buy. The store decides much of what goes into the trolley that the consumer had not thought of buying. It decides very little when the consumer already knows what they want.

And the nudges?

Over the last fifteen years the idea has spread that small changes in how options are presented, so-called nudges, change behaviour cheaply and effectively. It would be confirmation that the store manipulates the consumer.

A 2022 meta-analysis by Stephanie Mertens and colleagues, with 455 effects from more than 200 studies, found a moderate average effect, larger for food choices⁶¹. The same year, Maximilian Maier and colleagues reanalysed the same data correcting for publication bias, that is, the fact that studies with no effect stay in the drawer⁶². The average effect fell to close to zero. Nudges that change the structure of the choice, such as position or the default option, seem to have some effect. Those that only give information, none.

The same caution applies to consumer neuroscience. Martin Lindstrom, in Buyology, presented a study with more than 2,000 volunteers in brain scanners to show that brands act below consciousness⁶³. The study was never published with peer review. In 2011, when Lindstrom wrote in the New York Times that people literally "love" the iPhone, 46 neuroscientists signed a letter disputing the interpretation⁶⁴. It is a useful source for raising hypotheses, not for treating them as proven.

The stock-out punishes the store more than the brand

Thomas Gruen, Daniel Corsten and Sundar Bharadwaj brought together 52 earlier studies, 661 stores, 32 categories and 71,000 consumers in 29 countries⁴. On average, 8.3% of products were missing from the shelf. In north-western Europe, just over 7%. In the south, close to 11%.

Faced with a stock-out, 31% go to another store. Only 26% switch brand in the same store. The store loses about 4% of sales to stock-outs, and 70 to 75% of the causes are in the store itself: poor orders, wrong forecasts, late replenishment.

A 2003 study by ECR Europe with Roland Berger followed what happens when the stock-out repeats⁶⁷. At the first gap, most people accept a substitute. By the third, 69% do not buy and change store.

The stock-out is the purest case of veto. No one complains, no one coordinates anything. Each person solves their own problem, and the sum of those solutions moves to the store next door. The comments on the reels about stock-outs lasting 5 to 9 days are the Portuguese version of this study.

Online reviews: the trace you can see

Judith Chevalier and Dina Mayzlin compared reviews and sales of the same books on Amazon and Barnes & Noble⁶⁸. When a book's reviews improved on one site, it sold more on that site than on the other. A one-star review weighed more than a five-star one.

Michael Luca ran the same study with Seattle restaurants on Yelp⁶⁹. One extra star raised revenue by 5% to 9%, but only in independent restaurants. In chains, the effect was zero. Where the brand is strong, the consumer trusts the brand. Where there is no brand, they trust the traces left by other consumers.

Ya You, Gautham Vadakkepatt and Amit Joshi brought together 51 studies on online word of mouth⁷⁰. The effect is real and larger for products that are expensive, hard to try and consumed in private. It is smaller for products that are cheap, easy to try and have many substitutes. In other words, it is smaller in grocery.

In Portugal, a Portal da Queixa survey with about 5,000 responses says 72% of consumers name the opinion of other consumers as the main purchase factor⁷¹. It is an online, opt-in survey, so the number counts as a signal, not a measurement. Marktest found 47.5% of consumers research online before buying in a physical store⁷².

Boycotts: lots of noise, little effect, with exceptions

In 2003, after France opposed the Iraq war, sales of French wine in the United States fell 13% for about six months, with no organisation leading the boycott⁷³. It is pure stigmergy: each person reacted to what they saw around them.

But most boycotts fail. Jill Klein, Craig Smith and Andrew John studied a real boycott and concluded that only a minority took part, even among those who thought the company was to blame⁷⁴. Boycotting has a cost: you have to give up a product you like. When Goya was boycotted in the United States in 2020, social media filled up with calls to act, but sales rose 22%, driven by people who wanted to support the brand, and after three weeks everything was back to normal⁷⁵.

The exception is Bud Light. After the 2023 boycott, volume among regular buyers fell by 34% to 37%, and most moved to other beers⁶⁶. In June 2023, Modelo became the best-selling beer in the United States⁷⁶. When there is a close substitute and switching costs nothing, the veto works.

The day the consumer ran the manufacturer

In 1985, Coca-Cola changed its formula after taste tests with about 200,000 consumers, in which the new formula won⁶⁵. Complaint calls went from about 400 to 1,500 a day. After 79 days, the company went back to the original formula. In the next two days it received 31,600 calls.

The tests measured taste. They did not measure what the brand meant to the people who drank it. It is the clearest example of a consumer who could not say what they wanted beforehand, but knew very well how to say what they would not accept afterwards.

What changes with social media

Philip Kotler, Hermawan Kartajaya and Iwan Setiawan describe, in Marketing 4.0, a shift of power along three axes: from exclusive to inclusive, from vertical to horizontal, from individual to social⁷⁷. The idea is that consumers now listen to each other more than to brands. The research in this chapter partly confirms it. The consumer listens to other consumers when the brand is weak and the product is hard to judge. In grocery, where the product is cheap and quick to try, they mostly listen to themselves.

The sentence that changed retail

In 1994, Tesco tested a loyalty card in three stores and then in fourteen, with the data analytics firm dunnhumby. When the results were presented to the board, the chairman, Ian MacLaurin, said: "you know more about my business in three months than I have learned in 30 years"⁷⁹. Clubcard was launched in February 1995. In 1996, Tesco overtook Sainsbury's and became the largest chain in the United Kingdom. In 2010 it bought the rest of dunnhumby.

The quote often circulates as "you know more about my customers". The verified version says "my business". The difference matters: the card was not only used to get to know customers, it was used to decide range, prices, promotions and store locations.

The traces in Portugal

Cartão Continente was launched in 2007 and serves about 4 million households⁸⁰. The app had 2.3 million users in 2022⁸¹. Lidl Plus grew close to 70% in Portugal in two years and has more than 70 million users in Europe⁸². In 2026, almost half of grocery purchases in the country already go through a loyalty card, and each shopper holds, on average, three cards from different chains⁷⁸.

The consumer gives up their traces in exchange for discounts. The chain gathers them, reads them and makes decisions with them. It is the stigmergy of the triangle with one essential difference: with termites, the traces are visible to all. In the supermarket, they are kept by whoever owns the card.

The manufacturer no longer reads its own consumer

The grocery manufacturer sells to the store, not to the consumer. It knows how much it sold to the buying office. What happened next, who bought, how often, what they bought alongside it, is known by the chain. To find out, the manufacturer buys panel studies or pays the chain for access to its data.

It is the complete reversal of the manufacturer era. In the 1960s, it was the manufacturer who knew the consumer, through advertising and market research, and the store was a point of passage. Today it is the store that knows the consumer purchase by purchase, and the manufacturer that needs the store to see them.

But the consumer still holds the veto

The traces say what the consumer did. They do not say what they will do when an alternative appears. Private label is reaching its limit (chapter 4), stock-outs keep sending customers to the store next door (chapter 6) and strong brands keep bringing people into the store (chapter 2).

This is where the thesis of this analysis comes together. The consumer aggregates without knowing it, through shared needs and the traces they leave with each purchase. Whoever reads that aggregation best, rules. First it was the manufacturer, with the brand. Then the retailer, with the store. Today both fight over the reading, and the retailer is ahead, because it keeps the traces. But neither of them rules the consumer. The consumer holds the veto. Not the command.

A note on words

George Lakoff, a linguist, showed how the frames we use, the metaphors we think with, decide what we are able to see⁸³. When we say that "argument is war", we start to see attacks and defences. Lakoff wrote about politics, not brands, and applying his idea here is our own extrapolation. But it is useful. As long as manufacturer and retailer think of the relationship as a war between the two of them, they will keep seeing only the opponent in front of them. This analysis proposes another frame: a triangle in which the third side, the consumer, decides who wins.

Before: the pushed order

In the old model, the manufacturer sold to the store and the store sold to the consumer, and the two sales barely talked to each other. A manufacturer's sales rep was measured by what they could get into the store, the sell-in. What happened afterwards, the sell-out, was known late and poorly: through leftover stock, the next order that did not come, the returns.

In this model, the art of selling was convincing the buyer. Whoever negotiated best filled the store's warehouse, and the problem of selling through moved to the other side of the table.

The turning point: the barcode and sales data

The first barcode scan at a supermarket checkout happened on 26 June 1974, in Ohio: a pack of chewing gum⁸⁷. By the end of 1984, more than 10,000 US supermarkets had scanners. Emek Basker measured the effect: store productivity rose and labour costs fell, and the author allows for the possibility that scanners contributed to the power of the chains⁸⁸. It was the chains that imposed the code on suppliers: Kmart in 1983, Walmart in 1987⁸⁹. In Portugal, the barcode arrived in 1985, with 137 founding companies of what is today GS1 Portugal⁹⁰.

The barcode gave the store something the manufacturer had never had: the sale, product by product, store by store, day by day. It was the first big trace of the consumer, and it stayed on the store's side.

The pulled order

The next step was to use that trace to order. In 1988, Procter & Gamble proposed replenishing stock based on sales, first to a regional chain and then to Walmart. According to the account at the time, Sam Walton replied that if it helped him sell more soap, yes⁹¹. Walmart opened its sales data system to suppliers in the early 1990s⁹².

In 1993, the Kurt Salmon report that launched ECR (Efficient Consumer Response) estimated possible savings of 30 billion dollars in US grocery if manufacturers and chains worked from real demand⁹³. In 1996, Walmart and Warner-Lambert tested joint order planning for Listerine for five months: on-shelf availability rose from 87% to 98% and stock fell about 20%⁸⁴.

Theory explains why. Hau Lee, V. Padmanabhan and Seungjin Whang described the bullwhip effect in 1997: small variations in consumer demand turn into large swings in orders as they move up the chain⁹⁴. The cure they proposed was to share consumer sales data with the manufacturer and replenish continuously. In other words, let the purchase drive the order.

Today: every sale triggers the next order

In Portugal, a University of Minho thesis on Sonae's logistics spells out the mechanism: replenishment is automatic and orders "are generated automatically after a given number of sales", with minimums and maximums set centrally⁹⁵. In May 2026, MC chose Blue Yonder for allocation and replenishment in more than 400 stores, with AI demand forecasting "based on actual sales" and "less manual intervention"⁹⁶. Minipreço moved to RELEX in 550 stores in 2021⁹⁷. Mercadona uses its own system that adjusts fresh produce orders daily based on sales history⁹⁸. Pingo Doce shares with suppliers, in real time, each store's sales, range and shelf data⁹⁹.

The names change, the logic is the same. The order is no longer decided by the conversation between sales rep and buyer. It is decided by the product's rotation on the shelf, read by a machine.

The consequence: trade marketing starts to decide

If the order comes from the sale, the place where the order is won is the store. Everything that makes the product leave the shelf, being available, visible, in the right place, at the right price and with the right facings, feeds the algorithm and decides the next orders. Whoever sells more today buys the place in tomorrow's orders.

Trade marketing was born in Europe in the 1980s precisely as a response to the power of the chains and the lack of shelf space, with the aim of increasing rotation at the point of sale¹⁰⁰. Automated ordering gave it an importance it did not have before.

The numbers confirm that in-store execution counts. In a 105-week study across 90 stores, keeping the planogram intact delivered 21% more units, 12% more sales and 7% more gross margin⁸⁶. A new product took between 3.6 and 7.2 weeks to be properly distributed. One ketchup that reached the shelf in four weeks sold 118% more; another that took seven weeks, 57% more. In a world of automated orders, the weeks lost at launch are weeks in which the algorithm learns that the product sells little.

The role of the sales rep changed with this. In 1994, Ralston Purina cut 125 field sales jobs and moved to key account teams, while adopting continuous replenishment¹⁰¹. Bain wrote in 2019 that manufacturers' key account managers still often work in spreadsheets, without basic consumer sales data, while the best chains come to the negotiation with data product by product and shelf by shelf¹⁰².

The algorithm's blind spot: phantom stock

The algorithm only knows what the system tells it. Nicole DeHoratius and Ananth Raman analysed about 370,000 stock records in 37 stores and found 65% were wrong⁸⁵. A 2025 study in a British supermarket, with 24,000 products in 11 stores, found errors in 64.7% of products; in 39.4%, the system had more stock recorded than actually existed¹⁰³. That is the worst case: the system thinks there is product, does not order, and the shelf stays empty. A full store count raised sales by about 11% in two months, and the effect on perishables was three times larger.

There is a second blind spot. When the shelf is empty, the sale that did not happen is not recorded. The algorithm learns that the product sells less than it does, and orders even less¹⁰⁴. The stock-out feeds itself.

That is why physical work in the store has become power again. The shelf stacker, the merchandiser and the promoter are not just tidying shelves. They are making sure the trace the consumer leaves reaches the machine that decides the next order intact.

The limits

The algorithm does not decide everything, and there are three limits to bear in mind.

People still intervene. Karel van Donselaar and colleagues showed, with 19,417 product-store combinations in Dutch supermarkets, that store managers systematically change the proposed orders, for reasons the system does not see, such as the cost of shelving¹⁰⁵. It does not always go well: in a recent study at a European chain, managers cut about 22% of the system's proposals, and those cuts after a stock-out caused a third of the avoidable stock-outs¹⁰⁶.

Space and promotions deliver less than they seem. A meta-analysis of 1,268 estimates found an average shelf space elasticity of 0.17: 10% more space gives about 1.7% more sales¹⁰⁷. According to Nielsen, about three in four grocery promotions in the United States do not pay for themselves¹⁰⁸, and promotions are what most mislead forecasts¹⁰⁹.

The algorithm only replenishes what the buyer decided to stock. Negotiation still decides what goes into the range and what comes out in range reviews. Mercadona contracts product by product, and can remove a manufacturer from part of a category¹¹⁰. Chains also sell access to their own data, like Asda (chapter 3). The algorithm works within a list that is still negotiated.

And when it fails, it fails big. Target closed 133 stores in Canada between 2013 and 2015, with empty shelves and full warehouses, after it began ordering automatically with badly loaded data¹¹¹. In March 2020, toilet paper sales in the United States rose 845% in two days, and no system was ready for it¹¹².

What this changes in the thesis

The consumer still does not command, but their purchase has gained a direct consequence it did not have: every sale is a replenishment order. It is the stigmergy of the triangle in its most literal sense. Every purchase leaves a trace, and the trace decides the next step of the whole chain, without anyone giving orders.

Whoever gets the product off the shelf no longer convinces only the buyer. They convince the algorithm. But the algorithm only sees what the store shows it, and only replenishes what the negotiation let in.

The store started selling access to the consumer

On 26 June 2024, Sonae's MC presented Endless Retail Media Solutions, "powered by Cartão Continente", to advertisers and agencies¹¹⁷. Retail media is advertising sold by the chain itself, on its own channels, targeted using its customers' data. Endless combines three kinds of data: who the customer is (age, gender, location, segment, preferred store), what they buy (products, quantities, prices, promotions used) and what they look for online (searches, clicks, what they put in the basket). At launch, it announced more than 4.3 million monthly visits to Continente Online, more than 1.5 million active app users and more than 4.5 million store visitors.

What Endless sells the manufacturer is not space. It is the chance to reach a chosen consumer: someone who stopped buying the brand, someone who buys the competitor, someone who never tried it. It is chapter 7 taken all the way. The store keeps the traces and sells the manufacturer access to them.

It is not a one-off. Jerónimo Martins opened a retail media unit at Pingo Doce and Biedronka. At Pingo Doce it works with more than 40 suppliers and expected more than 200 targeted coupon campaigns by the end of 2024; it says the Portuguese market is "at an early stage"¹¹⁸. Auchan presents about 2 million weekly store visits as an audience that "compares well with television"¹¹⁹. In Spain, Schwarz Media, from the group that owns Lidl, sells advertising on the Lidl Plus app, the online store and in stores¹²⁰.

Abroad, the scale is different. Amazon made more than 68 billion dollars in advertising in 2025¹²¹. Walmart, about 6.4 billion, up 46% in one year¹²². According to WPP Media, commerce-linked advertising overtook television worldwide for the first time in 2025¹²³.

Why the chains want this

Because it is the highest-margin business they have ever had. McKinsey estimates cited in the Harvard Business Review put retail media margins at 50% to 70%¹²⁴. Food retail runs on margins of a few per cent: Ocado's supermarket business, for example, had an EBITDA margin of 3.8% in 2025¹²⁵.

When one line of business gives 15 times more margin than the core activity, it changes the priorities of those who decide. The buyer who negotiated prices and terms now has a colleague alongside who sells campaigns to the same supplier.

Who pays, and how freely

In the United States, executives from four agencies told Digiday that chains put retail media investment commitments into the annual business plans with brands, and hint that there are shelf space consequences for those who do not invest¹²⁶. "It is never written down." "It feels like paying twice." A survey in Australia found 54% of brands feeling pressure to increase spend, 73% struggling to measure return and more than half investing to protect shelf space¹²⁷.

The measurement problem runs deep. Whoever sells the campaign is also whoever measures the result, with data only they have. In an IAB Europe survey of 180 advertising buyers, 85% name the chain's data as the big opportunity and 53% the lack of common standards as the biggest obstacle¹²⁸.

Academic research is just starting. Koen Pauwels and Ladipo Fagbola concluded, in 2025, that retail media gives chains a strategic advantage, but creates a conflict between being an advertising platform for suppliers and competing with them on the shelf¹²⁹. A Toulouse School of Economics model shows that competition between manufacturers for ads can erode their profits, and push some towards selling direct¹³⁰.

Some see it differently. Pingo Doce says retail media and trade marketing are "complementary, not substitutes"¹¹⁹. And there is no public figure for Portugal on the money already flowing through here.

Online weighs little in Portugal

Retail media is growing, but Portuguese online grocery is not. According to Kantar, online was worth about 2.1% of grocery in 2019, rose to 3.4% in 2021, with the pandemic, and went back to 2.5% in 2022¹¹³. In 2025, the channel fell 13.9%, while in Europe it grew 6.8%¹¹⁴.

Continente Online has existed since 2001 and claims more than 300,000 active customers¹³¹. Pingo Doce opened the country's first online supermarket in 1998 and closed it in 2003. Between 2018 and 2025 it sold online through Mercadão, which Glovo bought in 2021. In May 2025 it once again had its own online store¹³².

Spain is further ahead. According to NIQ, online reached 7.3% of the basket in 2025 and grew 17.7%. But the leader is not a supermarket chain: Amazon has 36.6% of grocery sold online, Mercadona 9.8% and Glovo 4.8%¹³³. Mercadona sold €1,061m online in 2025, 2.5% of its sales¹³⁴. For comparison: in the United Kingdom online is worth about 13% of grocery shopping, in France 11.9% and in the United States 19%¹³⁵.

Endless itself admits it, in a 2026 sponsored article: more than 90% of purchase decisions happen in physical stores¹³⁶. The most valuable digital trace in Portugal does not come from the website. It comes from the card swiped at the checkout.

Is the online consumer different?

Research says yes, in ways that help and hurt both sides. Junhong Chu, Pradeep Chintagunta and Javier Cebollada studied the same households buying online and in-store at a Spanish chain¹³⁷. Online, they were less price-sensitive in all 12 categories studied. Alexandru Degeratu and colleagues had already found, in 2000, that brand weighs more online, especially when the site shows little product information¹³⁸.

But Andrea Pozzi showed that trying new brands happens mostly in the physical store¹³⁹. Online, the consumer repeats the list. That protects the brands already there and closes the door on new ones. And Arce-Urriza and Cebollada, with data from 30 categories at a Spanish chain, found more online loyalty to all brands, but only one gaining new customers: private label¹⁴⁰.

Online favours whoever is already settled in the basket: the leading brand and the store brand. The loser is the new, small brand, the one that most needed to be discovered.

Delivering costs more than selling

McKinsey did the maths for every 100 dollars of shopping¹¹⁶. In the physical store, the chain makes about 4. Online, with the order picked in store and delivered to the home, it loses about 13: picking alone costs about 8 and delivery another 8. A Capgemini study of 500 executives concluded that companies only recover about 80% of the delivery cost from customers, and 97% admit the model is not sustainable at scale¹⁴¹.

Ocado, the company that bet most on automated warehouses for online, had an adjusted loss of £353m in 2025. Kroger, in the US, paid it 261 million to close centres it had ordered¹²⁵.

Logistics is where the store's old power becomes physical. Sonae MC invested about €50m expanding the Azambuja and Maia centres; Azambuja covers 70,000 m² and ships more than 9,000 pallets a day. Jerónimo Martins opened the Valongo centre in 2017, with about 100,000 m² and a 75 million investment¹⁴². Whoever has warehouses, fleets and stores close to people has the last word on what arrives and when.

Whoever delivers: the fourth side of the triangle?

Delivery platforms arrived to take the last contact with the consumer. In 2020, Continente teamed up with Uber Eats, with about 1,000 products delivered in up to 30 minutes. In 2022 it launched its own quick delivery app, Quico, which by 2025 already reached 33 towns¹⁴³. Getir, which had bought Gorillas, left Spain, Italy and Portugal in 2023¹⁴⁴. Glovo bought Mercadão and did not keep the model; Pingo Doce went back to selling online on its own¹³².

In the United States, where platforms grew more, the pattern repeats with one difference. Instacart processed 37.2 billion dollars in orders in 2025, and about 29% of its revenue already came from advertising¹⁴⁵. The platform that delivers also keeps the traces, and also sells them. In Portugal, platform commissions charged to restaurants reached 15% to 35% in 2020, and DECO complained to the Competition Authority¹⁴⁶. We did not find the commissions charged to supermarkets.

The fourth side exists. In Spain, the biggest online seller of grocery is Amazon. In Portugal, for now, the chains have held it: they have the warehouses, the stores and the card.

And the consumer?

The Portuguese consumer did what they have done in every chapter: used the veto. They tried online during the pandemic and went back to the store. The CTT 2025 e-commerce survey shows that only 28.4% of online shoppers have ever bought supermarket groceries online, against 35.3% in Spain; and only 8.7% prefer to receive their shopping in less than two hours¹⁴⁷. Ultra-fast delivery was a promise few asked for.

Digital did not give the consumer the command. It gave them more ways to be read, and created more players who want to read them. Whoever delivers reads. Whoever reads sells what they read. And the manufacturer, which tried to sell direct to the consumer, rarely managed it: McKinsey concluded that only "a handful" of these attempts work, and in 2023 Unilever sold the majority of Dollar Shave Club, which it had bought for about a billion dollars¹⁴⁸.