Brian Lehrer: It's The Brian Lehrer Show on WNYC. Good morning again, everyone. Now we turn to a conversation about groceries. We all need them, we all buy them, and these days, a lot of people buy them online. Regardless of where you shop, you probably felt some sticker shock recently, that moment when you realize, like, wait, how is a tub of yogurt almost double the price it was a few years ago? According to a new investigation from Consumer Reports and the nonprofit's Groundwork Collaborative, and More Perfect Union, the grocery delivery company Instacart may have been making the problem at least a little bit worse.

They found that the company has been running AI-driven pricing experiments on their customers. That is, they've been charging people different amounts for the same product from the very same store. Now, we've actually got some pretty big breaking news that came out yesterday. I see that as a result of their investigation, Instacart announced that they would stop charging different prices for the same groceries at the same time. With us now to unpack this news and tell us about the investigation that they undertook is Derek Kravitz, reporter and deputy editor at Consumer Reports. Derek, thanks for joining us. Welcome to WNYC.

Derek Kravitz: Thanks for having me.

Brian Lehrer: What kinds of experiments were they conducting exactly?

Derek Kravitz: We enlisted 400 volunteers from across the country to do our own testing to find out which prices they were being charged for a variety of different goods on the Instacart app for Target, Costco, Kroger, Safeway, some of the big retailers in the United States. We looked at a lot of different products. 12 pack of eggs, Skippy peanut butter, Cheerios, some major name brands. We saw subtly different prices for a good amount of the products we tested; about 74% had some type of algorithmically changed price attached to it. The variance, or the difference between the lowest price and the highest price for some of these products was as high as 23%.

When you add that up over the course of a week, a month, a year, it can add up to real money. When we looked at Instacart's own calculations for how much a household of four might spend on groceries in a given year, the difference between the low and the high baskets could result in $1,200 in terms of a cost swing between, again, the low and the high.

Brian Lehrer: Over the course of the year?

Derek Kravitz: Right. For us, even though we saw marketing materials that Instacart really made this price experimentation known and public to their business clients, they described these price experiments as negligible, having a negligible impact on the end consumer, but what we found was something a little bit different. The prices were time substantially different from one another.

Brian Lehrer: What was the point? Again, listeners, if it isn't clear what we're talking about, our guest is not talking about different prices in different neighborhoods or cities. He's talking about two people shopping online at the same store at the same time, paying different prices for the same item. Were they just trying to test how much profit they could take out as the online middleman between the store and the shopper before the customers started to turn away from the product?

Derek Kravitz: Right. When we looked at marketing materials that they, again, provide to their business clients, they market this as a profit generator. A maximization of profit, both sales revenue, more orders, and also profit margin. Increasing individual product margin and rounding up in some cases, but not just $0.10 or $0.50, up to $2.56, we found on some items. It adds up to real money over the course of a fiscal year for a major retailer. This can be hundreds of millions of dollars. Grocery industry is typically a low-margin business, and so these percentage points really do matter to this type of company. That's why, in short.

It's really an elaborate form of A/B testing, how a particular price works, or how people respond in group A versus group B. This is way more elaborate, way more turbocharged in the AI era. This data that they have, not just personal data, but also your behavior, your shopping history, all these metrics, they add up to really sophisticated output where you can manipulate or toggle prices and see how you might react. That's the point of this experimentation. You need more training data, you need more data in order to fully understand or explain how a particular price point might work for this type of consumer or that type of consumer.

Brian Lehrer: Do I have this breaking news right, that Instacart announced just yesterday that it's ending these AI-driven experiments on price?

Derek Kravitz: Yes. There was a lot of backlash to this story. We published on December 9th, the same day in Congress, Ruben Gallego introduced a bill that would ban surveillance pricing, pricing that uses your personal data nationwide. I think we counted 12 members of Congress that wrote letters to the FTC or Instacart itself asking for answers or an investigation. The FTC then weighed in and wouldn't comment on an investigation to us, but Reuters reported that they were probing this, and they said that they were disturbed by our findings.

That all happened the last two weeks, and then Instacart announced yesterday that they were going to be ending price experiments for base prices, for item prices is the term they used. We're unpacking that a little bit. We have an update to that story going out this afternoon. The distinction here is that they are ending price experiments for base prices, the core price to a product. They are not ending price experiments for promotions and discounts.

Now, that's what they say. That's long industry practice. Even before the online era, grocery stores would experiment with how our promotion might work. Buy two, get a dollar off, or buy one, get one free, how that might work for different customers, and how they might react to it. In this case, it's just much more turbocharged. The tech is just much more advanced and allows them to really see how these offers work and how the final price points really influence a shopper's buying decision.

Brian Lehrer: Did these experiments lead to individualized pricing based on an individual's past shopping behavior, or was it just general, what you call A/B testing that randomly selected people?

Derek Kravitz: When we spoke to them about this, we went back and forth over, I think, four months of questions and answers with Instacart. They described this as randomized by both geography, where in the country you are, and then the product category: cereal, eggs, or milk, et cetera. We didn't find anything that suggested that they were so individualized or personalized down to the individual shopper that we could tell from our study, but we looked at patents. We looked at Instacart, bought an AI company several years ago, Eversight, and they filed hundreds of patents over the years.

These patents make clear that they can use demographic data, your age, your gender, how many people are in your household, your income, all these data points, in order to inform their algorithm and segment you into what they call subpopulations. Now, how that's being used, if that's being used solely for promotions and discounts, as opposed to these core base price experiments, that's an open question. That's one that regulators, state attorneys general that we've spoken to, and the FTC are interested in.

Brian Lehrer: Listeners, any questions or, for that matter, stories, if maybe you figured out that you've been subject to algorithmic pricing, not just necessarily by Instacart at grocery stores, but in any context, 212-433-WNYC, 212-433-9692. Call or text for Derek Kravitz, reporter and deputy editor at Consumer Reports, whose story from earlier this month on this, as he's been describing, has actually had an effect on Instagram in particular, pulling some of this back.

By the way, congratulations on that, Derek. I always think that's the highest compliment we can pay to anyone in journalism. Your story actually made a difference and changed something.

Derek Kravitz: I appreciate that. Really, I think it's a lot about timing with some of these projects and stories. Truism is-- Our big stories or our favorite stories don't necessarily get the attention or impact we're hoping for, and then sometimes things just happen in terms of timing and luck where people respond. I do think this hit at a crucial time for Americans amid rising grocery prices, the fastest increase since the late 1970s, and then the affordability debate going on nationwide. I think that's what's driving it.

Brian Lehrer: Doesn't this already go on at the individualized level? I feel like I've been hearing for years by now people saying, "Oh, I see the price change on—name your shopping site—when I go back and look at it a second time. Maybe it goes up because they know I'm interested in this sort of thing from my past behavior, or maybe it goes down because they know that I've looked a few times and I didn't buy." This already goes on a lot, doesn't it?

Derek Kravitz: Yes. Really, other industries have been using this for years. Think about airlines or concert tickets, Ticketmaster, or hotel reservations. Dynamic pricing, the term basically rapid shifts in supply and demand, and the cost of an airline ticket increasing a few days before you want to fly somewhere. That's been in practice. We've seen it less with essential items like groceries and other industries.

It's only in the last few years has this really become a thing because the tech and again these third-party intermediaries, these pricing algorithms, companies like Eversight or Revionics coming onto the scene, they're offering a product that really appeals to retailers, big companies, because again, it increases both sales, revenue, and profit margins. It's a no-brainer.

Uber has been ahead of the curve in terms of algorithmic pricing for years. There's been many books written about that, including one of our study co authors wrote a book about Uber, and they really set the standard in terms of how this personal data can be used to influence the price that you see at any given moment. Yes, it's pretty much everywhere.

Brian Lehrer: Listener asks in a text, "Can your guest, Derek, suggest a way to stop Instacart from using my personal data?" I guess I would generalize that question to stop any company.

Derek Kravitz: Generalizing. One thing that we put at the end of our story were some tips for folks to consider when they're shopping. This applies not just to groceries, but really any retailer. Studies have shown that buying in bulk actually does save you money in the long term. The per-unit aspect of that really works in your favor. Also, that point, per unit, looking at the per unit price from one product to an alternative or a competitor, that really is the great equalizer. That's why so many states have passed per-unit pricing laws that mandate that that disclosure be on the physical price tags in stores, because it really does give you a lot of information to work with.

Then, shopping in person. I think for a lot of communities that we found, that's not necessarily an option in every case, especially those with disabilities or those in rural areas or food deserts. We find that a lot of SNAP EBT benefit holders use Instacart for various reasons. If it's possible to shop in person, we do recommend it.

Brian Lehrer: Matt in Brooklyn, you're on WNYC with Derek from Consumer Reports. Hi there.

Matt: Hey. Thanks for taking my call. It's funny that what Derek had just mentioned is what I had told your screener about. I think it's worth pointing out that oftentimes, these prices, not just individualized prices, but the stores themselves who participate on Instacart or Uber Eats, whether it be food delivery for restaurants or grocery delivery, they have to hike up their prices of the individualized items. You may go to a restaurant's website or the grocery store's website and see one price that they're listed, and then when you go to buy it through Instacart or Uber Eats, it's 20% to 30% more, perhaps.

Your guest mentioned briefly-- I heard him talk about how a lot of people with SNAP and EBT benefits are utilizing these. My understanding is we're trying to get the most bang for our buck out of SNAP and money, the most efficient, and it just seems really sad that people, at the end of the day, end up spending more of their benefits by using these sites. One more thing I want to throw out there really quickly is I saw a praise of Instacart and Uber Eats offering when the SNAP was shut down, saying about, "Oh, we're giving 50% of off, 50% off, $50 off," and all this kind of stuff. When you actually add up all the fees and the price hikes from the business side, I think people will pay more.

Brian Lehrer: Thank you, Matt. Do you want to say anything else about targeting low-income consumers in general? People using SNAP benefits in particular?

Derek Kravitz: Definitely. Brought up some good points there. One thing to think about with SNAP EBT, a lot of retailers, not just Instacart, really promote their product to that population, specifically offering, say, Instacart plus premium membership 50% off for the first year to SNAP EBT benefit holders. That might seem like a good thing. They're getting a little bit of a deal and less fees over the course of that year, but really, that's in order to retain and keep that customer shopping and to have them shop beyond that first year and keep paying.

To the other point about markups, yes, these are baked in. The markups are part of these type of e-commerce third-party delivery platforms like an Uber Eats or a Doordash or Shipt, or Instacart. What we found is actually even more granular, and that's why it takes 5,000 words to explain in a story, they're manipulating base prices, discounts, and promotions separate from just the fees. The fees are a whole nother ball of wax to consider. It's a lot to unpack.

Then we found fictitious or false reference pricing where they're changing the strikeout original price in order to make a savings appear larger or smaller to psychologically motivate you or influence you to buy that particular product because it seems like a good deal. This has been around for decades, but the tech is just so much better right now, and that's why it's a debate. That's why people are responding the way they are.

Brian Lehrer: Let me get a couple of callers on here with stories from actually different sites. Not Instacart and grocery shopping, but stories of being charged different things for the same thing. Nick in Port St. Lucie, Florida, you're on WNYC. Hi Nick.

Nick: Hey, guys. I'm usually Nick from Corey, New York, but I'm on vacation. I was trying to help my mother yesterday because we shared a Netflix account for a long time, and now they've implemented this new rule where you have to have separate accounts per household, or you can add on or whatever. My subscription is 13.99, and it's billed as that, and it stayed that, and then her subscription was starting to be billed at 18.99. I just was holding the two devices up, and I'm wondering what's different? What is this? I don't know if anyone knows more about that.

Brian Lehrer: There's a story. Hold that thought for a minute, Derek, because we're going to get one more in here. Julie in Manhattan, you're on WNYC. Hi, Julie.

Julie: Just happened to me on Amazon. I bought a bra for $29, and I thought, "Oh, nice bra. Finally fits." I went back on, and it was $58. I called my cousin, I said, "Listen, don't copy this URL. It might be connected to me. Just put in the details." She saw it a $38 marked down from $49.

Brian Lehrer: There you go. I don't know if it's a coincidence that 58 is exactly double what you paid for the first one, but Derek, you hear those stories.

Derek Kravitz: Yes, I do.

Brian Lehrer: They fit a pattern.

Derek Kravitz: They do. We hear a lot of these stories from various retailers. Amazon and Walmart are some of the leading companies that use algorithmic pricing at scale everywhere. Julie's story is definitely not unique. I do appreciate that approach, which is similar to what we did, which is look across the country or look at different users and see what price you get when you're logged in or not on a platform like Amazon. The algorithms can be a black box. To peek inside that black box requires usually hundreds or thousands, really, of volunteers to work in tandem together to try to figure out what they're seeing and what the differences are.

One of my colleagues, Alan, who I think used to work for this show, was great about recruiting people and getting them to come to our virtual sessions and shop in real time. Really, that can be, again, a great equalizer, trying to understand what we're seeing and why. Another thing that we found is consumers can ask for the data that companies hold about them. In at least 18 states in the US, including New York, you can request a company to provide the personal data, inferences, and things that they hold about you as a consumer through a state privacy law.

Usually, there's a portal for each company where you can request that information. That can be also empowering. You can see what a company holds about you, what inferences they make about you, and how that information then might influence the price that you see online

Brian Lehrer: You gave a shout-out to Alan Smith there, who was a great producer of The Brian Lehrer Show in years past. He's the one who got in touch and said, "Hey, I'm at Consumer Reports now, and we've got this really good investigative story." Good on him and good on you. Just to finish on Instacart, they say they're going to retire their algorithmic pricing experiments based on that news release yesterday, but they recently acquired an AI company, Eversight, just a couple of years ago.

You reported that Instacart has been transitioning from the way people may think of them, a company that primarily delivers groceries to one that "sells a range of digital tools to help grocery retailers and manufacturers increase sales. The CEO, Chris Rogers, told investors during a November earnings call that they're the leading technology and enablement partner for the grocery industry." To announce that they're ending these experiments, it sounds like a complete overhaul of the business model they just announced, or how would you interpret it? 30 seconds.

Derek Kravitz: I think it's an artful and very smart pivot from a business perspective. Their former CEO is now one of the top execs at OpenAI. During her tenure at Instacart, they really did transition to a tech company. Not an e-commerce delivery platform, but more of this giant tech product for grocery retailers and consumer packaged good companies. Kraft Heinz, Frito-Lay, all of these companies that want to sell you something.

Really, this is what the future might look like. The digital price tags, digital carts with screens on them that can measure how much product you put in a cart, and blast your ads to that screen. These are the things that regulators are thinking about: how do we regulate this? How do we play catch-up? What do we do in response to these new tech products?

Brian Lehrer: Derek Kravitz, reporter and deputy editor at Consumer Reports. Thanks so much for coming on and sharing your reporting.

Derek Kravitz: Thank you.

[music]

Copyright © 2025 New York Public Radio. All rights reserved. Visit our website terms of use at www.wnyc.org for further information.

New York Public Radio transcripts are created on a rush deadline, often by contractors. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of New York Public Radio’s programming is the audio record.