When a camera closes, a cart opens: Amazon shifts focus from Just Walk Out to Dash Carts.
I find that there is almost always a lag between reality and the perception of reality, especially with big things like “the economy” or “business”. Things seem to actually bottom out about 2-3 months before people notice and start to comment on how things have bottomed out, and by the time the commentary achieves a critical mass, things have already started to turn around.
That’s how it feels to me right now. Retailers are pessimistic and cautious, but consumers are starting to show true signs of Spring – and yet, at a cautious enough rate not to spook inflation or, as a result, Jerome Powell and the Fed. Inflation continues its slide, employment is strong enough to not make consumers nervous, and wage gains are soft enough not to make economists nervous. Mall visits are up, retail space is getting leased, AI continues to garner a lot of hype (amid more examples of why the hype may not all be deserved), and retailers and tech companies continue to innovate (or pull the plug on innovation that isn’t working, as the case may be). Let’s dive in!
Retail Economic Indicators
I don’t know how this squares with downtown office buildings still struggling to get workers to return to office, but it turns out that retail occupancy levels overall hit a 5-year high in 2023 at 88.4%, according to the Datex Property Solution 2024 Market Outlook report. The report also notes that vacant spaces’ time to re-lease is getting shorter, an 18% decrease over 2022, and that 61.5% of new leases increased their rates in 2023 (vs the previous tenant). The CEO of Datex, Mark Sigal, noted that the growth in rental rates exceeded the growth in retail sales, setting up a potential squeeze on retail profitability.
Alongside these numbers, comes a report that mall visits are returning to pre-pandemic levels. According to Placer.ai, traffic is back up, though they note that malls are different than how they were before the pandemic, with a greater reliance on entertainment, restaurants, and events to draw people in. This is really just the continuation of a trend that was already in place before the pandemic, but still notable and good news: the pandemic did not kill the mall.
In the UK, March brought good news, with shop price inflation reaching the lowest level in two years – down to 1.3% in March from 2.5% in February. Inflation is the lowest it has been since January 2022. Food inflation is higher, but it too is dropping – from 5.0% in February to 3.7% in March.
And it was “jobs week” for US economic indicators. The Bureau of Labor Statistics published both its Job Openings and Labor Turnover report, covering February, and also the report on non-farm payrolls for March. As of the last day of February, not much had changed from the month before – which is all good news. Job openings (8.8 million), number and rates of new hires (5.8 million and 3.7%, respectively), total separations (5.8 million) as well as quits (voluntary separations – 3.5 million) were all relatively unchanged from January. The non-farm payroll report came in hotter than expected, but not enough to create anticipation that it would derail any plans for rate cuts later this year. Analysts were expecting to see 216,000 added jobs, but it turned out to be more like 303,000. But unemployment remained unchanged at 3.8%, and wages rose only 0.3% compared to 4.1% on average over the last 12 months. ADP rounded out the week with its Unemployment Report, showing that private employers added 184,000 jobs in March. This was the biggest jump they’d seen since last July, and it did come with wage gains.
Retail Tech & Research Data
A WalletHub survey found that consumers are feeling “credit card fee fatigue” – 85% of respondents to a consumer survey in the US said they feel like they are being “nickel-and-dimed” when they are asked to pay an extra fee to process a credit card payment. Half say they will not use their credit card if they have to pay a fee (though I would take that with a big grain of salt – I don’t think we see this behavior play out in real life). Retailers can potentially find some take-aways in these two stats: 48% said retailers are not transparent when charging fees for credit card transactions, and 59% said they believe it’s unfair when merchants pass their fee costs onto customers. It appears that retailers need to make their case to consumers for making credit card fees explicit.
AI & Retail
Co-op in the UK reported that it lost GBP70 million to shoplifting in 2023. Matt Hood, the managing director for the food business, elaborated on the numbers behind that number, which were pretty astonishing. The incidents of “shoplifting, abuse, violence, and anti-social behavior” increased by 44% in 2023. This works out to an average of 1,000 incidents per day. Every day, four colleagues were attacked and a further 116 abused. He also said that Co-op would remain committed to self-checkout (SCO) despite the rise, because the convenience for shoppers outweighed the drawbacks. However, the company also seeks to mitigate the drawbacks by investing in AI linked to CCTV cameras to monitor what items are put into the basket at self-checkout. I think, one, those numbers are depressing. But two, more tech is not necessarily the answer here. Monitoring SCO just means more confrontation when theft is detected. Deterrence – in the form lots of smiling, friendly people – would reduce the need for confrontation, and thus abuse.
If retailers can’t figure out how to best staff SCO, then governments will do it for them. California is already contemplating a bill that will require, among other things, a minimum of one worker for every two “cashierless” registers. I guess more than that and you’re basically back to a staffed register. It’s a proposal for now.
More directly in the AI space, forget about not enough energy to power AI, what if there isn’t even enough data to train the next generation of large language models (LLMs)? AI companies are apparently so data-hungry they’re starting to consider whether to use “synthetic data” – AI-generated source material. This gets us over to the question of generation loss pretty quickly, which researchers say could cause “crippling malfunctions.” My favorite quote from the article: “Feeding a model text that is itself generated by AI is considered the computer-science version of inbreeding.”
Algolia, a search and discovery platform, surveyed 1,000 US consumers on the role AI could play in fashion, and 44% reacted positively, saying that they believe AI can enhance their fashion sense. 1 in 4 would trust an AI bot trained on their retail purchases and preferences to pick out an outfit for them – and would trust it more than their friends (though this was higher for younger generations). 59% think wider adoption of AI by online retailers will create better shopping experiences.
And then Adobe weighed in with a survey of their own, and found that 58% of respondents believe that GenAI has already improved their online shopping experience. I’m not sure how GenAI was defined in this context, because I don’t know that 58% of shoppers could tell you if something they interacted with was GenAI or not (although, I guess Amazon tells you the product review summary they generate is GenAI, so in that case, maybe 58% of shoppers have encountered it after all). 52% surveyed said they are likely to use GenAI tools to help with clothing purchases, and 71% said they think using GenAI to try on products virtually could assure them in their online shopping experiences. The top use cases that survey respondents liked probably fall in the exact reverse order of how many retailers provide them today: top of the list was automating product filters based on consumer needs, followed by creating customized items, summarizing product reviews, and then last place in the list was customer service chatbots.
Personally, I like the summarized product reviews, but I do find myself wondering how legitimate they are, given the long-standing issues with genuine reviews. Also, don’t use those for dog food!
Bernard Marr wrote up comments about AI that IKEA’s chief digital officer, Parag Parekh, shared with him on his podcast. They’ve set up an AI task force, and one thing I appreciate about his comments is that he makes a point to distinguish between GenAI – which powers an experimental personalized design assistant that helps shoppers erase the existing furniture and décor in a room, and then assists them in repopulating it with IKEA, based on how they answer questions about preference and style. They’re also using GenAI via Microsoft Copilot to adapt photo shoots to be seasonal – adding a Christmas tree behind the sofa, for example. But then, on the optimization side, they’re optimizing space in delivery vehicles and using drones to monitor warehouse inventory. Not all AI is GenAI.
Finally, just because last week I highlighted the poor auto dealer chatbot that was (easily) duped into creating car offers that included things like a specially prepared meal from the celebrity chef of your choice, here’s a new example of GenAI in the wild going wrong. New York City is experimenting with an AI chatbot to help advise the public about starting and operating a business in the city. The problem is, as Colin Fraser points out in his excellent essay (I linked to it last week too), GenAI is not designed to be accurate. It is designed to approximate accuracy. And when the difference between the two means breaking the law, that’s not good. A chatbot with high legal requirements is not even the right application of GenAI, and chatbot technology – based on natural language processing, yes – has already tackled this problem, requiring more up-front effort but also responding more accurately. As just an example of how illegal the chatbot has been, it advised landlords that it’s OK to discriminate based on income sources and also that it’s OK for restaurant owners to take a share of server tips.
Retail Winners and Losers
This week, I feel like this section should just be rebranded “Amazon” because the news was all about Amazon. They debuted a spring sale near the end of March and for a first run it was probably OK – the event had discounts as high as 60% off, but according to Earnest Insights, created a seasonal lift of 3.1% vs. the average of the trailing 4 weeks. Compared to the October sale, which generated lift of 23% by the same measure, it seems there may be a long way to go to get consumer engagement.
Of course, the big news was the near-death of Just Walk Out. It’s still around in a few locations, but Amazon pulled it out of nearly all of its Fresh stores. Much was made in the press over the expense of the service – it requires a lot of tech – and also questions about the reliability of the tech, with reports that Amazon had to employ human-run error checks on as many as 700 out of every 1,000 orders – this in spite of the tech investment. The best commentary centered on just how often what looks like tech turns out to be “just a guy”.
Amazon is replacing the tech with Dash Carts – smart carts that consumers can scan and bag as they go through the store. I’ve had a front-row seat to attempts at this technology as far back as 2005. I wish them luck. However, it’s fair to call out (as Oliver Banks does) that Amazon has never been afraid to innovate – or to pull the plug if it’s just not working. And there’s a big side benefit to this kind of experimentation. When you’re constantly innovating, then sometimes companies try to skip on their own innovation efforts by being a fast follower. So I wonder how Trader Joe’s feels to have to announce around the same time that Amazon is pulling the plug, that they are opening their own Amazon Go-type store called Pronto, complete with some sort of cashierless tech.
And finally, on the Amazon front, The Interline published a great article on the necessity behind – and risks – of having to leverage your infrastructure by providing it to other businesses. The subjects of the article are Shein vs. Amazon. On the surface, what the two are doing is very similar, but their focus areas are very different, with Shein far more on the sourcing side of the supply chain, getting into distribution, versus Amazon more focused on distribution up to storefront.
Just so that I don’t have to make this section entirely about Amazon, I will note that Levi’s got attention with their latest quarterly results, mostly because they reported that nearly 50% of their revenue comes from DTC – a combination of eCommerce and owned stores. That’s 25% higher than it was 2 years ago. While the company notes that they expect a slowdown in discretionary spending that will cap their full year revenue growth at 1-3%, they were also careful to talk about the importance of their wholesale partnerships – it appears they’ve learned a lesson watching Nike go too aggressively at cutting down wholesale relationships and finding that they needed to retrench.
The Bottom Line
It’s okay to innovate and fail. And even when someone like me makes snide comments about the next thing to try, companies should never fall prey to “oh we tried that already and it didn’t work.” The days of winning at retail by just showing up to a location with some products are long gone. Retailers have to balance all kinds of things – tech vs. real estate investments, omnichannel vs. focusing on any single channel, innovation vs. optimization. That’s not easy, but it will definitely be what separates the winners from the rest.
