How much would you pay for this mayonnaise? Artificial intelligence wants to find out.

How data, algorithms and consumer behavior are transforming one of the oldest decisions in commerce: setting the price of a product.

Picture two people buying exactly the same product, in the same online supermarket, at the same moment.

One sees a certain price. The other sees a different one.

The product did not change. The supplier did not change. Neither did the production cost.

So what changed?

Perhaps only the version of the price each shopper was shown during a commercial experiment.

Sounds strange? For anyone working in marketing, statistics or business intelligence, this situation opens a fascinating discussion: what is a product worth, and how does a company find that out?

The answer is no longer coming only from cost spreadsheets, preset margins and traditional market research.

We are entering the era of data-driven pricing.

Price was never just a math problem

For a long time the commercial formula looked fairly simple: calculate costs, add a margin and put the product on the shelf.

But that formula never fully explained why someone happily pays R$ 15 for a coffee in one place while finding R$ 8 expensive in another.

Marketing has known this for decades.

Price also involves perceived value, convenience, positioning, experience, urgency, competition and even the way an offer is presented.

A bottle of water has one value in a supermarket. In an airport, the sense of convenience is different.

The product may be essentially the same. The commercial context is not.

Artificial intelligence adds an important dimension to this discussion: the ability to analyze thousands of signals at once and spot patterns that would be practically impossible to see by hand.

And that changes the way a price can be built.

The supermarket became a pricing lab

In December 2025, an investigation by Consumer Reports, together with other organizations, revealed that American shoppers found different prices for the same products, in the same stores and at the same times, during experiments run on the Instacart platform.

For some products, the difference reached 23%.

The tests used pricing technology to assess how consumers reacted to different price points.

Instacart stated that the experiments did not use personal information, demographic data or individual behavior to set prices and did not amount to personalized pricing. After the public reaction, it announced the end of the tests on December 22, 2025.

Whatever one makes of the controversy, there is a very interesting business question in this episode.

What if a product’s price could be continuously studied as a marketing hypothesis?

Instead of assuming that R$ 19.90 is the ideal price, a company can investigate how different price levels affect sales, conversion, margin and loyalty.

In this context, the supermarket is no longer just a place for transactions; it also becomes a source of learning about the market.

The science behind willingness to pay

In economics and marketing there is a concept called willingness to pay.

It is the maximum amount someone would be prepared to spend on a product or service under certain conditions.

That amount is not fixed.

It can vary with need, competition, convenience, brand reputation and the alternatives available.

This is where statistics, predictive models and artificial intelligence come in.

By analyzing sales history, aggregate behavior, seasonality, response to promotions and shifts in demand, a company can estimate how sensitive certain products are to price.

That sensitivity is known as the price elasticity of demand.

One product may sell far more with a small discount. Another may barely change in volume, even after a price increase.

Knowing the difference makes for better commercial decisions.

And the best decision is not always to raise prices.

Selling cheaper can mean earning more

Consider a hypothetical example.

A company sells 100 units of a product at R$ 100. The variable cost of each unit is R$ 60.

Total contribution margin is R$ 4,000.

Now imagine that by cutting the price to R$ 90, the company sells 140 units, with the same unit costs.

Total margin rises to R$ 4,200.

Even charging less per unit, the business improves its result by R$ 200, before fixed costs.

It is a simplified example, but it shows a fundamental lesson: the highest price does not always produce the highest profit.

Pricing intelligence is precisely about understanding these relationships.

Instead of relying on intuition alone, managers start testing hypotheses, comparing scenarios and tracking results.

It is the difference between simply picking a number and building a commercial strategy.

Where does marketing come in?

Almost everywhere.

Smart pricing can guide promotional campaigns, loyalty strategies, product positioning and even the development of new offers.

A company may discover, for example, that certain customers do not really need discounts. They value fast delivery, an extended warranty or better service more.

Other segments may be more price-sensitive and respond better to economy packs, subscriptions or bundles.

Notice the shift.

It is not only about deciding how much to charge for the same product.

It is about understanding which value proposition makes sense for each audience.

That insight can shape the whole marketing mix: product, price, place and promotion.

Price starts talking to the other commercial decisions.

The next step: algorithms that recommend decisions

As artificial intelligence systems evolve, price analysis can become more sophisticated as well.

Statistical models estimate demand. Algorithms identify patterns. Recommendation systems suggest adjustments. Generative AI tools explain scenarios and help managers interpret the results.

In more advanced operations, AI agents can monitor market signals and recommend commercial actions within limits set by the company.

That does not eliminate human strategy.

Quite the opposite.

The work starts to demand better questions.

What is the goal of pricing? Increasing margin? Gaining market share? Clearing inventory? Winning new customers? Improving repeat purchases?

Each goal can produce a different recommendation.

And an algorithm without a clear commercial strategy may simply optimize the wrong metric.

Is there a limit to this intelligence?

Yes. And it is a commercial one too.

A strategy that increases the profit of one transaction can damage trust in the brand when customers do not understand the criteria being used.

Offering loyalty discounts, subscription benefits or promotional prices with clear rules is one thing.

Changing prices opaquely based on personal characteristics, without weighing the legal, reputational and competitive effects, is another.

In Brazil, the processing of personal data and automated decisions must comply with the LGPD, the country’s data protection law, as well as consumer protection rules.

The topic is also already under regulatory scrutiny: a study by the US Federal Trade Commission found that different kinds of data are used to support individualized prices and offers. It is important to distinguish those practices from the randomized tests described by Instacart.

Technology has to serve a sustainable value proposition.

Because winning an extra margin today may not make up for losing a customer tomorrow.

And loyalty has economic value too.

The real revolution is not on the price tag

For decades, companies have tried to find the right price for their products.

Now, with analytical tools becoming ever more accessible, they can investigate something far more interesting: how perceived value changes with context, the offer and the behavior of the market.

That shift opens opportunities for supermarkets, manufacturers, service companies, digital platforms and small businesses.

There is no need to start with giant systems or extraordinary algorithms.

A company that organizes its sales history, knows its margins, compares periods and measures how customers respond to promotions can already turn data into better decisions.

Artificial intelligence expands that capacity. But the foundation is still data quality, understanding of the business and clarity of goals.

In the end, perhaps the most important question is not how much a machine can get someone to pay for a jar of mayonnaise.

The commercially smarter question is a different one:

How much value can a company create once it truly understands why its customers buy?

Because the future of pricing should not simply be about charging as much as possible.

It should be about finding the best balance between perceived value, competitiveness, profitability and relationships.

Reality sends signals. Knowing how to read them is what turns data into decisions.

---

Sources: Consumer Reports — investigation into Instacart price tests · Instacart — announcement ending the tests (Dec. 22, 2025) · Federal Trade Commission — study on individualized pricing (2025).