We Had 400 People Shop For Groceries. What We Found Will Shock You.

[Former FTC Chair] The holy grail for a long time has been, what if we could charge every single person the maximum amount that we know they're able to or willing to pay? [Expert] They know things about us that we don't know. They're smarter than us and we're training them in a way that is incalculable. [Eric] Walk into any grocery store and on average, everything is about 30% more expensive than it was in 2020. Inflation. Supply chains. Tariffs. We’ve heard it all.

But what if something else is happening? Something intentional? What if someone is charged $2 for these eggs, but another person is charged $2.40 at the same exact store— and it's pushing prices up? Five months ago, a researcher reached out to me about Instacart. She'd been studying their Washington, D.C. workforce. Different pay for nearly identical work. Her question: if they do this to workers, what about consumers? Before joining More Perfect Union, I worked in grocery pricing.

So I agreed to help investigate. The problem? It wasn't easy to test. Grocery prices change all the time for legitimate reasons. Receipts you can't compare. If we wanted to do this, we’d need a team. A researcher. Community organizers. Other journalists. And most importantly, regular people who care. Lots of them. Five months of digging. What we found is bigger than Instacart. Yes, inflation is real, but something else has been pushing prices higher this whole time.

There's a system, one that grocery stores and tech companies built together. It learns from every purchase. It's not just online. It's in the physical stores you walk into. With the primary goal: maximize profit. Day by day, it's moving billions of dollars, one penny at a time. Here's how we uncovered it. How it's turning the grocery business into a gold mine. And why without regulation, your grocery bill may keep climbing no matter what happens with inflation.

[Katie] I'm now going to ask you to take out your phone, and we're going to get started with shopping for 20 items at the Safeway. [Eric] This is Katie Wells. She's the Director of Research at Groundwork Collaborative. She's the one who reached out. The plan was pretty simple. Get everyone in a room. Open up Instacart. Select the same store. Same items. On phones and computers. Select pickup to remove any delivery calculations. And write our basket totals down on a note card.

All to test a simple question: were shoppers being charged different amounts for the same exact item at the same exact store and time? Congratulations guys. You have finished the test. [Eric] First, we all shared our cart totals.

[Participant] We’re in the same room, the same address ordering, and we still had different prices. [Eric] You're the highest. I'm not surprised. What I've come to understand is that a lot of things that we do online is predicated on algorithms. Different totals. There's plausible explanations. Promotions, user error, app issues. But if the items themselves had different prices, that's something else.

[Katie] The items had upwards of three different prices offered. [Eric] Katie analyzed the data. The greatest difference was about a 70 cent spread. For eggs, the prices varied from $4.59 to $4.69, down to $4.28. It was a 41 cent span. One store, one moment in time. It's a small sample size. This could be a fluke. Welcome welcome welcome. You have found your way to the Consumer Reports grocery pricing webinar. The problem with testing any of this is that these systems are so massive

that you need scale to even try. Consumer Reports joined the collaboration. They brought a volunteer network. Community organizers. A reporter. Alongside them, we recruited volunteers. Over 400 people. Same process. Everyone opens Instacart. Same store, same items. When pickup wasn't available, same delivery address. Same moment. Screenshot everything. Submit. The team verified every entry. Cross-checked every screenshot. Built the database. Katie crunched the data.

And when the data came back, I gotta be honest, I stared at it for a while, because what we found was weird. I'm going to show you some basket totals. Some people were charged about $114 for the 20 items. Others nearly $124. This wasn’t in one or two cases. This was the lion's share of shoppers for the lion's share of items. [Eric] If you zoom in, nearly three out of four products had different prices during the experiments.

Four different prices for Wheat Thins. Five different prices for eggs. Three prices for peanut butter. Wait what? The system sorted people ino price groups, where each person was charged the same for every item. For this test, there were four price groups. Same cart, same store, same moment. Same with this test. For this store, seven different price groups. This didn't seem random, but who or what was doing the sorting? Back in August, before we even ran the test, I called Instacart.

We don't set prices, retailers do. And we believed them. But the groups suggested something deeper was going on. Why would an algorithm sort people into buckets where everyone pays the exact same prices? Not just on one item, but across multiple products. All higher together, all lower together. I started researching and that's when I stumbled on surveillance pricing. Companies tracking your behavior, your purchase history, and charging you based on that. It's a type of algorithmic pricing

where companies hand pricing over to an algorithm. It's almost impossible for people outside the companies practicing the art—the dark art of algorithm price discrimination— exactly what is driving those algorithms to set the prices that we observe. [Eric] Len Sherman studies this at Columbia’s Business School. Was that what was happening here? When we showed Instacart the test results, they were clear. Retailers control all pricing. The findings were just stores putting out

random feelers to see where the pricing sweet spot is. They aren't targeting you specifically. So we verified with the stores we tested. First Albertsons-Safeway. They’re an Instacart partner. Silence. Then we contacted Target. They responded. And honestly, my jaw dropped. They told us, ‘we don't work with Instacart. We don't set prices. We have nothing to do with any of this.’ Wait, Instacart says retailers control the prices and Target says they don't even work with Instacart?

I started digging. It turns out, in 2017, Target spent over half $1 billion to buy Shipt, Instacart’s competitor. Why would they help Instacart? We asked Instacart about Target's response. Instacart admitted they actually do manage Target's prices on the marketplace. And that they were running tests to figure out how much they can charge on top of Target's normal prices. Their explanation shifted, but the question we started with was still the same.

How does the system sort people into different price groups? To figure that out, we needed to understand what Instacart actually does. You may think of Instacart as a delivery app, but it's much bigger than that. Instacart is more the technology. I don't think of it as a grocery delivery company. I think of it more of a technology company. [Eric] Selling technology that powers grocery websites, advertising campaigns, promotions. In 2022, Instacart bought Eversight,

a company that optimizes grocery prices. Eversight says it runs constant experiments that are not made visible to consumers and it is running tests, in part, to collect data about how much I'm willing to pay for canola oil. And it's using that information to likely create more profits for these companies. [Eric] The company calls it “AI for everyday price performance.” The website promises 2% to 5% profit increases. In early November, Consumer Reports ran follow-up experiments at Costco.

They found price gaps. 15% on a bag of chips. Derek asked Costco to explain. We got an email a few hours later from a Costco Vice President of grocery. The message described a program called “Smart Rounding.” You know, took us a little bit of time to realize what we were seeing, which was an internal communication between Instacart and Costco. Inadvertently, this Costco executive forwarded us this internal email. As far as we can tell, Instacart has only talked about Smart Rounding once,

in a 2023 letter to shareholders. The company defined it: “machine learning-driven tool that helps retailers improve price perception and drive incremental sales.” Translation: flex prices up or down based on what the algorithm says will maximize profit. [Derek] “For some of our major grocery partners, this has led to millions of dollars in annual incremental sales.” Millions of dollars. This is a very sophisticated, very, deliberate, explicit attempt to maximize profits,

and do so somewhat surreptitiously. We'd answered some questions. Different prices for different people? Yes. Random testing? Maybe not. Smart Rounding showed its designed for profit. Retailers in control? Debatable. Instacart sets at least one grocer’s prices. But we still couldn't answer: how does the system sort people into price groups? Then one store showed us the bigger picture. Schnucks. Regional Grocer. They're on every Instacart press release. No price variation at all.

One night, it hit me. What if they already moved past online testing? You've probably heard stories about electronic shelf labels. Digital price tags that can change prices instantly. If they're testing prices in store with those labels, and the app just mirrors what's on the shelf, you'd never catch it with our test. Which might explain Schnucks. The typical fear with these labels is surge pricing. Raising lemonade prices when it's hot out. What if the goal isn't surge pricing at all?

What if it's more subtle? More permanent? Would you rather charge $10 for a bottle of water during a single hurricane? Or 20 cents more every single day? Derek had a genius idea. Patents. He found one. It described an infrastructure for constant in-store price testing. Take dozens of stores, cluster them together, test different prices in different clusters. The purpose? Find out the most profitable price point. Once you do, adopt it. We asked Schnucks and Instacart. Both denied it.

But for Instacart, their website told a different story. If you go there right now, you'd see this: three features. Before we asked? Four features. And the one removed? Price optimization with Eversight via electronic shelf labels. Gone. So, Instacart denied doing it, then changed their website. Later they told us it was never a feature and the update was made for accuracy. But that didn't track. The feature seems to show up in past marketing materials and press releases. If it's operational,

the system would be a truly impressive feat of engineering. In-store? Testing across locations. Online? Testing across individuals. But we still couldn't answer how the online system decided who gets charged more. So we started building a theory. Not based on what Instacart told us, but on what Instacart told the government. Patents. The word that appears over and over in them? It's not random. It's segmentation. Derek started reading through the criteria they could use to sort people.

Some things raised red flags. Demographics. It's illegal to charge someone more based on those. Katie controlled for that. No correlation in the sample. I don't believe they would try to vary prices based on those kinds of characteristics. Nor would it be very effective if they did. What would be way more effective and perfectly legal, as far as I know, is to vary prices based on behavioral characteristics. Behavioral characteristics. That's exactly what the patents show. Buying behavior.

Purchase history. How frequently you shop. Whether you're shopping around. Coupon usage. Loyalty program participation. An infrastructure that tracks everything you do. Uses it to decide what to charge you. We asked Instacart: does this explain who gets sorted into different price groups? Their response? One old patent from 2015, before we purchased Eversight. Patents are always broad. This doesn't prove we're actually using this technology. Fair point. Here's the same chart

and a host of other patents, mostly about promotions, marketing material that touts the patents, and then this one. A promotional patent that describes shopper and product headroom. Meaning: how much more money could this person spend overall? And how much more could they spend on this product? The patent says they can calculate this for every individual consumer. But for efficiency, typically people are “grouped with similar consumers.” Groups. Just like our tests showed.

Every company that I know is practicing first degree price discrimination, strenuously denies that they're doing so. [Eric] And again, Instacart denies segmenting people by personal or behavioral data. What happens when companies actually do this at scale? In 2022, Uber rolled out algorithmic pricing for riders and drivers. Len studied what happened next. For the next three years, average rider price per mile went up, up, up, up, up, up, up. [Eric] Uber denies all of this.

But at about this time, they went from burning money to a cash machine. [Len] That's why more and more companies are trying as hard as they can to perfect this dark art—devilishly effective dark art, as I've called it. That's how you turn grocery sales into a goldmine. Now, algorithmic pricing sometimes offers lower prices. And the total impact on grocery prices? Impossible to measure. Instacart is just one company, a sliver of the grocery market. They claim ten clients with Eversight.

What are other companies doing? I started looking. [Stephen Mewborn] Pricing is the single most powerful lever to raise profits. Far more effective than either boosting sales or cutting costs. [Narrator] Prices optimized. Profit increases. [Eric] There’s an entire industry built around optimizing grocery prices. I called Errol Schweizer. 25 years in grocery. Former VP of grocery at Whole Foods. Why is everyone racing to adopt this? It's very much defensive. Amazon's already doing it.

Walmart is doing it. So everybody's got to keep up. Merchants pricing teams, they all rely on technology and algorithms to safeguard their profits. That's Walmart's former VP of pricing. So, Instacart has played this really vital role for the competitors of Amazon and Walmart, which is pretty much almost everybody else, to have a lot of the same tools and access to consumers as the big guys. [Eric] The systems are watching you. Learning your patterns. Using that to charge everyone accordingly.

[Errol] They know things about us that we don't know. They're smarter than us and we're training them in a way that is incalculable. This didn't happen overnight. 20 years of building this infrastructure. It was kept in check by inertia. Everyone was afraid to raise prices. Covid changed that. [Errol] There's no guardrails. It's a free for all. It's the wild, wild west. [Eric] Companies are investing millions into systems designed to outsmart you. Testing. Learning. Finding the maximum profit.

The holy grail for a long time has been, what if we could charge every single person the maximum amount that we know they're able to or willing to pay? Studies like this are just validating that firms are very much already midway through trying this, executing it. That's Lina Khan, former chair of the Federal Trade Commission. She agreed to review our findings with an eye to how we should protect consumers. [Lina] This could ultimately end up being another big form of wealth transfer

from ordinary Americans to massive corporations. I asked about the in-store testing apparatus, electronic shelf labels, the patents. [Lina] Firms have been trying to figure that out for many decades. But I think what's different now is both the sophistication of the tools that are available, just the deep information asymmetries that exist between consumers and these firms that are benefiting. But there was something else that concerned her. [Lina] When you have companies like Instacart

that could be facilitating, it could also ultimately allow firms that really should be competing against one another when setting these prices to actually engage more in something that looks like price fixing or collusion. [Eric] Because it could mean that one company, Instacart, is setting prices for retailers who should be competing with each other. Then I asked her about regulation. Her answer surprised me. I think we need to have a much more first order conversation about,

do we even want to permit some of these tactics in the first instance? [Eric] Under Khan's watch, the FTC started an investigation into surveillance pricing. This week it launched an investigation, the FTC, into the use of AI to rapidly change pricing. That investigation is now in question. The Trump administration claims it is ongoing, but Khan's replacement stopped public comments. But Lina pointed out that states don't need to wait for the Trump administration.

[Lina] The steps we've seen, even from states like California and New York, to ban certain types of algorithmic price fixing, in areas like housing and rent, for example, I think are extraordinarily important steps. We need more investigations. So, I think we need a 360 view, to try to understand what's really happening here. [Eric] Five months. Three organizations. 400 volunteers. We know the system exists for surveillance pricing, but we don't know if it's operational

because we can't see inside. And that's intentional. It's a system that's designed to stay hidden. [Lina] This work that you all have done is fantastic, but state AGs have subpoena power. They can be requesting and demanding information so they could look under the hood and be understanding, how are firms actually using this data?