Industries / Retail, hospitality and personal services / Retail and e-commerce / Vending
Remote monitoring
This is the work of having machines report what they sell, what they hold, how cold they are and when they fail, so that people can act without walking up to them.
5
Milestones achieved, from 1982 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1982
Coke machine on the network
Carnegie Mellon University's computer science department ran its own Coke machine, and people who walked to it often found it empty or freshly loaded with warm bottles. In 1982 John Zsarnay wired the machine's six "empty" lights to a serial port, Mike Kazar wrote a server that tracked when each column was last emptied, and Ivor Durham made it answer the network command "finger coke". People could then check from other computers, and in time from anywhere on the internet, which columns held cold bottles.
Credit
Mike Kazar, David Nichols, John Zsarnay and Ivor Durham, Carnegie Mellon University
1999
Heat pricing backfires
In October 1999 Coca-Cola's chairman, M. Douglas Ivester, told the Brazilian magazine Veja that a cold drink is worth more at a summer final and that a new machine "will simply make this process automatic", raising the price as the temperature rose. The New York Times reported that the machine used a temperature sensor and a computer chip, and PepsiCo said such machines "exploit consumers who live in warm climates". Coca-Cola then said none were in use and "We have no plan to put them in the marketplace."
LessonA sensor that lets a machine change its price can look to buyers like a machine built to take advantage of them.
Credit
M. Douglas Ivester, Coca-Cola
2010
Machines guess the buyer
In August 2010 JR East Water Business, part of the railway company JR East, put a "next generation" drink machine at Shinagawa station in Tokyo, with 500 more planned over two years. A camera estimated the sex and rough age of the person in front of it, and the machine used that, the time of day and the temperature to choose drinks to recommend on a 47-inch touch screen. The company was still promoting these Acure machines to tourists at Akihabara station in 2015.
Credit
JR East Water Business
2025
Most machines report in
Berg Insight, an analyst firm, estimates that 8.1 million vending machines worldwide were connected to a network in 2025: about 3.3 million outside North America and Europe, chiefly in China and Japan, 2.6 million in North America and 2.2 million in Europe. It names cashless payment as the main reason for adding a connection, because payment devices and touch screens now arrive with telemetry, the sending of machine data, built in. It counts Cantaloupe and Nayax as the largest suppliers, with about 900,000 connected machines each.
Credit
Berg Insight (estimate)
2025
Face analysis ruled a breach
In February 2024 a student at the University of Waterloo in Ontario posted online an error message, "Invenda.Vending.FacialRecognition.App.exe", from a campus snack machine, and River Stanley wrote it up in the campus paper mathNEWS. The university had the 29 machines, made by Invenda and supplied by Adaria Vending Services, removed by March 2024. In June 2025 Ontario's Information and Privacy Commissioner found that their cameras had captured identifiable faces and estimated each buyer's age and gender, logged with the time and the item bought, and that this was a privacy breach made without notice. The university had not known about the face detection, because its procurement did not include a privacy impact assessment.
LessonA machine can collect more than the site owner agreed to, and only a check made before signing would have caught it.
Credit
Information and Privacy Commissioner of Ontario; River Stanley
Watch for
Machine data picks sites
Coca-Cola Bottlers Japan, which runs about 650,000 machines and booked an impairment loss of about 90 billion yen for 2025, mainly on its vending business, says it will use data from its machines and its own phone app to decide where to place machines, what to stock and when to restock.
Achieved whenthe company or an outside analyst reports sales per machine and visits per machine on the changed routes against earlier years.
Claimed by
Coca-Cola Bottlers Japan, as reported by Jiji Press, April 2026
Open
01
Which machines have cameras
A public list of camera-equipped vending machines and kiosks in Ontario's public buildings, such as universities, colleges, hospitals and city halls, with the operator, the machine's maker, whether a camera is present, what data the contract lets the machine send, whether a privacy impact assessment was done, and the documents behind each answer.
The briefCredit · unclaimed02
A public DEX test set
An open collection of real DEX and EVA DTS audit files read from at least 30 models of vending machine and payment device, with serial numbers and locations removed, and a checker that lists where each file departs from the EVA DTS standard.
The briefCredit · unclaimed
Open 01 · the brief
Which machines have cameras
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public list of camera-equipped vending machines and kiosks in Ontario's public buildings, such as universities, colleges, hospitals and city halls, with the operator, the machine's maker, whether a camera is present, what data the contract lets the machine send, whether a privacy impact assessment was done, and the documents behind each answer.
Why it is still open
The commissioner examined one university. Invenda told CBC in 2024 that about 100 of its machines had been shipped to Canada and that it was not clear how many were installed, and no public list of camera-equipped machines in public buildings was found.
Built with what exists
requests under Ontario's Freedom of Information and Protection of Privacy Act for vending contracts and privacy impact assessments; photos of machine fronts and labels in public areas, with no people in frame; an AI model that answers fixed questions about each contract, such as whether it mentions cameras or data sent to a cloud service (see the Note "A thousand questions, one pass"), checked by two people; a public spreadsheet.
First step
Send access requests this month to five Ontario universities for their current vending contracts and any privacy impact assessments.
Done looks like
40 Ontario public bodies, each with a filled row and its source documents attached; at least 90% agreement between the AI model and two people on 50 sampled contract questions; the list published under an open license.
Open 02 · the brief
A public DEX test set
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open collection of real DEX and EVA DTS audit files read from at least 30 models of vending machine and payment device, with serial numbers and locations removed, and a checker that lists where each file departs from the EVA DTS standard.
Why it is still open
Open readers exist but test against little real data: dex-parse, an EVA DTS reader on GitHub, ships two sample files, and sigalabs' project simulates a machine instead of publishing real files. The European Vending Association offers the standard itself as a download, not a set of real files.
Built with what exists
an open ESP32 board that reads DEX through the machine's MDB connection (nodestark's mdb-esp32-cashless on GitHub, MIT license) or a standard DEX handheld cable; the existing open readers; Python; Zenodo for publishing.
First step
Ask three operators or used-machine dealers to read DEX from machines they own, and write the script that strips serial numbers and locations.
Done looks like
at least 300 files from 30 models and five makers, published under an open license; the checker run over all of them; a table of the ten most common departures; at least one reader project outside the authors adds the set to its tests.