Amazon Go stores deployed incredibly advanced systems to tackle an ancient retail challenge, demonstrating the perils of heavy tech investment without a clear path to profitability.
<div class="media-with-label__label">
Credit: Scott Warman / Unsplash </div>
</figure>
</div>
</div>
</div>
</div>For several years, Amazon Go stores epitomized cutting-edge retail technology, featuring an extensive network of high-resolution digital cameras designed to visually track every customer and their interactions with each product.
These stores not only highlighted Amazon’s technological prowess but also introduced an almost eerily seamless and human-free experience to convenience shopping. Regrettably, this elaborate setup also served as a prime example of technology that failed to achieve profitability and lacked any viable path to do so. (This situation bears a strong resemblance to the current landscape of vendors pushing generative AI (genAI) and Agentic AI systems; the title for pursuing unattainable ROI has clearly been passed on.)
The operational concept was remarkably straightforward: shoppers gained entry by scanning a code from their Amazon app, after which the cameras and analytics systems took over, allowing the customer to simply pick items and exit.
Amazon persevered for as long as possible, but on January 27th, after nearly eight years, the company conceded and announced the closure of all its Amazon Go stores.
It’s important not to misinterpret this decision. Amazon did not abandon the underlying technology that powered these stores (now rebranded as “Just Walk Out”). Instead, the company recognized that the true value of this innovation lay in a different application.
The initial vision for Go was to provide a completely frictionless shopping experience. However, it turned out that customers primarily visited Go for its novelty, finding it no significantly better than their conventional shopping alternatives.
Nevertheless, Amazon discovered that the technology unequivocally offered significant speed advantages. Not only were Go stores unprofitable, as noted by retail IT leaders, but they also failed as loss leaders, generating insufficient revenue to justify the investment. Given their lack of success in a small convenience store format, scaling them up to larger stores would never be viable.
Amazon eventually realized that the technology’s true potential wasn’t in expanding to larger retail spaces – such as those the size of Costco, Walmart, or Target – but rather in shrinking its footprint significantly, to much smaller applications.
Once Amazon executives understood that speed was the singular, most compelling benefit, they began seeking environments where rapid transactions were critical. Consequently, they started licensing the technology to compact venues where efficiency directly translates into revenue.
According to Amazon, the technology is now “cutting cafeteria wait times from 25 minutes to just 3 minutes at BayCare’s St. Joseph’s Hospital” and “allowing sports fans at Scotiabank Arena to grab concessions in a mere 30 seconds.”
This represents Amazonian ingenuity at its finest, even if it took many years to fully materialize. Instead of incurring losses in their own physical stores, they pivoted to licensing the technology to others. This generates immediate profit, as it’s considerably difficult to lose money on a technology perfected almost a decade ago.
By targeting ultra-small spaces, they are entering markets where speed takes precedence over all other factors. Consider stadium concession stands: they typically have limited windows for selling hotdogs, popcorn, and beverages – primarily during halftime and before the game. Post-game sales are challenging because most attendees wish to avoid queues and head home promptly.
The Go technology dramatically reduces queue lengths, enabling merchants to sell more products during these brief selling opportunities. The only remaining bottleneck is the time it takes to process a customer’s order and hand over the items. The payment process had always been the primary source of delay.
Zak Stambor, a senior analyst specializing in retail for Emarketer, shared that he witnessed the technology operating “exceptionally well” at a small snack kiosk in a train station he frequents.
The individual transaction revenue is relatively minor. “If I’m only purchasing a soda or a snack, the profit margin isn’t substantial,” Stambor remarked. However, speed proves to be the crucial differentiator. When he has only a minute before his train departs, Stambor lacks the time to cross the street for a conventional purchase. Yet, with the elimination of the payment step, the entire transaction flows effortlessly. He grabs a snack and boards his train.
“Amazon has gained considerable knowledge from this undertaking,” Stambor observed.
This situation parallels how Apple refined its authentication processes to facilitate almost instantaneous ticket purchases in the New York subways via NFC on an Apple Watch.
A fascinating aspect of the NYC subway experience – which I use occasionally – is the superior speed of the Apple Watch compared to the iPhone. During a recent trip to NYC with my wife, I effortlessly passed through the turnstile using my Apple Watch, while my wife struggled repeatedly to gain entry with her iPhone.
Even more inconveniently, she had to hold her iPhone, which is far from ideal in a crowded subway station. The fact that I didn’t need to physically handle anything for the payment to process made the experience feel truly magical.
This underscores the core insight: Amazon, Apple, and other innovators have pinpointed the specific scenarios where speed is the paramount factor, thereby enabling truly transformative technology to become, well, truly transformative.
The key takeaway here is this: when an IT department advocates for a business case for a new technology (and financial departments express reservations), the solution may not lie in refining the technology itself. Instead, it might be found in adjusting where and how that technology is deployed.
