Industries / Construction and property / Elevator and escalator services
Remote monitoring
Remote monitoring connects lifts and escalators to a service center, first by telephone line and now over the internet, so that faults and wear show up in data before, or without, a visit.
5
Milestones achieved, from 1984 to 2023.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1984
Lifts report by modem
Otis engineers Charles Whynacht and Robert E. Hall filed a US patent in July 1984, granted in November 1986, for a system in which a microprocessor in each building watches the door, brake, alarm-button and demand signals, counts runs and door operations, and sends alarm and performance messages by modem over telephone lines to a monitoring center. InformationWeek reported in 2019 that Otis's Remote Elevator Monitoring (REM), begun in the 1980s, collected such data once a day from more than 300,000 of its elevators.
Credit
Otis Elevator Company; Charles Whynacht and Robert E. Hall
2015
Maintenance predicted in the cloud
thyssenkrupp Elevator launched MAX in 2015, a service that sends lift data to Microsoft's Azure cloud to predict which parts need attention. The company called it "the industry's first predictive maintenance solution" and said it was connected to thousands of units in the United States, Germany and Spain by September 2016.
Credit
thyssenkrupp Elevator, now TK Elevator; Microsoft
2017
KONE connects to Watson
On 8 February 2017 KONE launched 24/7 Connected Services, which sends data from sensors on lifts and escalators to IBM's Watson IoT platform, a cloud service for internet-connected devices, to be monitored and analyzed in real time. KONE said it would be broadly available during 2017 and aimed to connect more than 1 million more units over the next few years. The launch gave no measured results.
Credit
KONE, with IBM
2018
Otis ONE goes live
Otis launched Otis ONE at the World Elevator Expo in Shanghai in May 2018, a platform that gives customers live equipment status and gives its service teams predictive warnings, built on data from more than 300,000 connected units. In its 2025 annual report to the US Securities and Exchange Commission, Otis said about 1.1 million of the roughly 2.5 million units it maintains were connected.
Credit
Otis Elevator Company
2023
Add-on sensors find faults
Hong Kong's Electrical and Mechanical Services Department, a government department, and Sebit Company Limited fitted add-on sensors, which need no change to the lift's own design, to four government lifts of different brands and ran them for a year. Their methods detected the incidents passengers reported, and also found faults missing from the maintenance records: in one lift the brake let go and caught again extra times after about 20,000 trips, and two lifts had door faults nobody had logged.
LessonSensors that any owner can fit, from any maker, can find faults that the maintenance log misses.
Credit
Jimmy K.K. Chan and colleagues, EMSD; Sebit
Watch for
AI agents cut callbacks
TK Elevator says AI agents in its Digital Operations Centers, which watch connected units and run remote resets, diagnostics and briefings for technicians, avoided about 20,000 unplanned service visits in a 2025 US pilot. It also says callbacks, the calls a customer makes when a unit breaks down between visits, fell by more than 40% per unit where the centers' interventions were carried out, and contract cancellations fell by 33%.
Achieved whena building owner, regulator or researcher publishes a year of callback rates for comparable units with and without the centers' interventions.
Claimed by
TK Elevator with Microsoft, April 2026
Cameras in the shaft
Otis says optional 360-degree cameras in the shaft of its Gen360 elevator let service teams see, adjust, diagnose and solve many problems remotely without stopping the elevator.
Achieved whenan outside party publishes how many faults on camera-equipped units were fixed without a visit, compared with units without cameras.
Claimed by
Otis, in its annual report filed February 2026
Open
01
Check the phone ride test
An outside check of how well a phone measures lift ride quality. Ride 30 public passenger lifts as an ordinary passenger, with two or three phones running a ride-quality app lying on the floor beside a reference instrument built to ISO 18738, the international method for measuring lift ride quality, and compare their readings of vibration, acceleration and sound on the same rides.
The briefCredit · unclaimed02
Run a door to failure
An open dataset that follows a lift door from new to worn out. Mount a surplus car-door operator and panels in a guarded frame on a workshop floor, with no car and no shaft and an interlock that cuts power when the guard opens, and cycle the door day and night until rollers, belts or guides wear out. Log motor current, vibration, sound and door timing on every cycle, and add some faults on purpose, such as a dirty sill, a slack belt or a worn roller.
The briefCredit · unclaimed
Open 01 · the brief
Check the phone ride test
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An outside check of how well a phone measures lift ride quality. Ride 30 public passenger lifts as an ordinary passenger, with two or three phones running a ride-quality app lying on the floor beside a reference instrument built to ISO 18738, the international method for measuring lift ride quality, and compare their readings of vibration, acceleration and sound on the same rides.
Why it is still open
DEKRA's LIFT Check app, updated in May 2025, reports ISO 18738 ride-quality figures from one ride with the phone on the car floor, but its listing cites no validation. Zhang and colleagues described smartphone monitoring of lift ride comfort in Mechanical Systems and Signal Processing in 2018. No published side-by-side test of phones against a reference instrument across many lifts was found.
Built with what exists
LIFT Check; two or three common phone models; a rented or borrowed ISO 18738 ride-quality analyzer as the reference; a spreadsheet or Python for the comparison.
First step
Borrow a reference instrument from a lift consultant or university and ride five lifts in public buildings, such as a library, a shopping center and a station, with the building manager's agreement.
Done looks like
30 lifts, three rides each, every phone against the reference; a published table of each measure with the mean error and the share of phone readings within 10% of the reference.
Open 02 · the brief
Run a door to failure
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open dataset that follows a lift door from new to worn out. Mount a surplus car-door operator and panels in a guarded frame on a workshop floor, with no car and no shaft and an interlock that cuts power when the guard opens, and cycle the door day and night until rollers, belts or guides wear out. Log motor current, vibration, sound and door timing on every cycle, and add some faults on purpose, such as a dirty sill, a slack belt or a worn roller.
Why it is still open
Public lift data come from working lifts and carry few confirmed faults. Huawei's 2020 Zenodo set holds door-bearing, humidity and vibration readings at 4 Hz from one building's evening peak. A July 2026 Zenodo set from Universidad Politécnica Salesiana holds 820 trips from one lift and says its flagged trips are not confirmed mechanical failures. A 2025 Greek motor-fault set is restricted. A 2025 Applied Sciences paper by Gęca and colleagues built a door-drive test bench with 11 seeded faults and reported about 97% recall, but no public copy of its data was found.
Built with what exists
A used door operator from a parts dealer; a programmable logic controller or a microcontroller board to command cycles; clamp-on current sensors, MEMS accelerometers and a microphone; for running the warning model on the motor drive itself, chips of the kind described in the Note "Maintenance inside the joint".
First step
Buy a used door operator, build the guarded frame, and log the first 10,000 cycles.
Done looks like
At least 1 million logged cycles and 10 labeled faults, released under an open license with a baseline model and, for each failure, how many cycles of warning it gave.