Amazon Fresh
Cashierless checkout via computer vision
In 2024, Amazon pulled its Just Walk Out cashierless checkout technology from its US Amazon Fresh supermarkets in favor of Dash Carts, after reports of reliance on human annotators; the technology remains in Amazon Go stores and with third parties.
Objective
Remove the checkout in-store: the customer takes their items and leaves, the system detects what they carry out and bills it automatically, to reduce friction and the cost of checkout.
The deployment
Just Walk Out equipped Amazon Fresh stores with ceiling cameras and sensors to automatically bill what the customer carries out, with no checkout. In 2024, Amazon pulls this technology from its US Amazon Fresh stores and replaces it with Dash Carts, carts that scan items. Reports, including from The Information, say that a large share of transactions in fact went through human annotators in India who validated the video. Amazon disputes the characterization of a dependence on human reviewers.
Results Proof C
Withdrawal reported by established press (Retail Dive, based on The Information) naming Amazon and its official spokesperson response. No isolated financial disclosure and no court decision.
How it works
Documented architectureThe stack in detail
- plateforme Amazon Just Walk Out In-house cashierless checkout technology, pulled from US Amazon Fresh stores in 2024 but kept in Amazon Go stores and licensed to third parties via AWS.
- llm Vision par ordinateur et fusion de capteurs In-house models that detect items picked off the shelf from the camera and sensor feeds; architecture not published.
- infra Cameras plafond et capteurs de rayon Physical instrumentation per store, the main cost driver of the supermarket format.
- outil Annotation humaine Teams, notably in India, that train the model and validate low-confidence visits; Amazon disputes a structural dependence.
- outil Dash Cart Onboard scanning cart that replaced Just Walk Out in the US Amazon Fresh stores, a semi-automated compromise.
Post-mortem
GraveyardWhat happened sourced
Amazon deploys Just Walk Out in its Amazon Fresh stores starting in 2020. In 2024, the company pulls this technology from its US Amazon Fresh stores and replaces it with Dash Carts, carts that scan items during the shopping trip. The Information reports that a large share of sales relied on annotators in India validating the video. Amazon responds that the technology does not depend on human reviewers, that these teams mainly train the model and validate only a small minority of visits. Just Walk Out remains in Amazon Go stores and with third-party customers.
Reason for failure sourced
On the supermarket format, the economics did not hold up: an environment dense in items and customers required costly human supervision to reach billing reliability, without the expected economies of scale. The Dash Cart, which hands scanning back to the customer, offered a simpler cost/reliability trade-off for this format.
Cost sourced
Cost not publicly disclosed. Multi-year investment in cameras, sensors, and per-store infrastructure, plus the cost of human annotation, for a format where the technology was ultimately replaced. Reputational cost tied to the gap between the image of full automation and the reliance on annotators.
Warning signs inferred
Inferred: a high per-store cost (ceiling cameras and sensors), a persistent need for human annotation to reach reliability, and a supermarket format expansion slower than announced. The dependence on human supervision in complex scenes indicated that the automation was not yet autonomous at this level of density.
Lessons in hindsight inferred
Inferred: measure the full cost of an automation, human supervision included, and weigh it against the real format. A technology that works in a small convenience store (Amazon Go) does not transpose mechanically to a dense supermarket. And communicating an automation as total when it partly relies on humans creates a reputational risk when the gap becomes visible.
Inferred: yes, the frictionless checkout pattern remains valid, but its economics depend heavily on the format. Amazon itself keeps it in smaller stores and licenses it to third parties. The failure is a targeted retreat on the supermarket, not a condemnation of computer vision checkout. The Dash Cart shows that a semi-automated compromise can fit a given format better.
How your customers perceive this type of use
Sourced studiesLe pricing algorithmique est le terrain le plus inflammable : 68% des consommateurs disent se sentir leses quand les marques utilisent le pricing dynamique et 80% jugent plus dignes de confiance les marques aux prix constants (Gartner, 2024). L'equite percue varie selon le secteur : le pricing dynamique n'est juge juste que par 33% a 40% des repondants selon qu'il s'agit de concerts ou de cinemas (YouGov, 17 marches). Le prix personnalise par les donnees individuelles est le plus rejete : 47% des Americains s'y opposent fermement (Consumer Reports, 2024).
Acceptance conditions
- La constance des prix comme signal de confiance : 80% jugent plus fiables les marques aux prix stables (Gartner 2024)
- Le secteur conditionne l'equite percue : le pricing dynamique est mieux tolere pour les cinemas (40% le jugent juste) que pour les concerts (33%) (YouGov 2024)
Red lines
- Le pricing dynamique percu comme abus : 68% se sentent leses (Gartner 2024)
- Le prix individualise a partir des donnees personnelles : 47% d'opposition ferme (Consumer Reports 2024)
- Les frais caches et hausses imprevues, vecus par 79% des consommateurs sur un an et associes a la perte de confiance (Gartner 2024)
Sources: Gartner 2024 · YouGov 2024 · Consumer Reports 2024
How to replicate
Inference, not sourcedData prerequisites
- multi-camera video feed
- weight/shelf sensors
- catalog and planogram
Org prerequisites
- an annotation and supervision team
- an economic model validated per format
Possible stack
- computer vision + sensor fusion
- scanning cart (semi-automated compromise)
- GDPR compliance on in-store capture
First step: Cost out the full cost per format, human supervision included, on a pilot store, before committing to instrumentation at scale.
Sources
- S1 Amazon removes Just Walk Out tech from US Amazon Fresh stores Established press archive pending
- S2 Amazon hits out at report that Just Walk Out relies on human reviewers watching from afar Secondary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.