Transportation

Studies examine biometric kiosk adoption, predictive vending maintenance

Two academic studies examine what it takes to make unattended systems more useful to passengers and more reliable for operators.

Image: Adobe stock

July 31, 2026 by Richard Slawsky — Editor, Connect Media

Summarize withChatGPTGrokPerplexityClaude

Self-service technology can deliver faster transactions and leaner operations, but neither benefit is automatic. In this week's research roundup, we look at two studies that examine what it takes to make unattended systems more useful to passengers and more reliable for operators.

The first paper examines passenger acceptance of biometric airport check-in kiosks in Thailand. The second proposes an Internet of Things and machine learning framework intended to predict vending machine failures before they interrupt service.

Together, the studies point to a common lesson: Advanced technology creates value only when users trust it and operators can translate its data into better decisions.

Biometric kiosk adoption depends on more than speed

Written by Phutawan Ho Wongyai, Thanh Ngo, Hanjun Wu and Kan Wai Hong Tsui, "Passengers' acceptance of biometric check-in kiosks: The case of Thai airports" was published in Tourism and Hospitality Research in 2025. The study examined how usefulness, ease of use, trust, privacy concerns and perceived risk affect travelers' willingness to use biometric check-in kiosks.

The researchers surveyed passengers at Suvarnabhumi Airport and Don Mueang International Airport between March and April 2024. After incomplete responses and outliers were removed, the analysis included 577 surveys. The sample included Thai and international passengers, although it skewed relatively young and well educated. More than 93% of respondents said they already used some form of biometric technology in daily life.

The results suggest a traveler's general attitude toward a biometric kiosk is one of the strongest links to intended use. Perceived usefulness and ease of use did not directly predict adoption. Instead, they helped form a positive attitude, which then increased the intention to use the kiosk. Passengers were also more willing to use the technology when they believed they had the knowledge, resources and ability needed to complete the process.

Trust and privacy mattered as well, but primarily because they shaped that overall attitude. Greater trust improved passenger perceptions of the kiosks, while privacy concerns weakened them. Perceived risk, however, did not have a statistically significant effect on attitude or intended use. Social pressure — whether people important to the passenger thought the technology should be used — also was not significant.

For kiosk deployers, the practical message is that speed alone may not be enough to encourage adoption. Airports and airlines should show passengers how the kiosks work, explain what benefits they provide and clearly communicate how biometric data is collected, stored and protected. Reliable performance is also important because every failure can undermine trust in a system that handles sensitive personal information.

Physical design and placement matter, too. At the Thai airports examined, biometric check-in kiosks and self-service bag drops were separate, requiring passengers with checked luggage to join two queues. The researchers suggested pairing the machines so travelers can move through the process more easily. A technically capable kiosk may still seem inefficient if the surrounding passenger journey remains fragmented.

The authors cautioned that the cross-sectional survey measured intentions rather than actual kiosk use over time. The results also came from two airports in one country, and the sample's familiarity with biometrics may limit how broadly the findings apply. Even so, the study offers useful evidence that privacy, trust, interface design and process integration should be treated as core deployment issues rather than secondary considerations.

Predictive maintenance model targets vending downtime

The second paper, "Predictive Maintenance Optimization for Smart Vending Machines Using IoT and Machine Learning," was written by Md. Nisharul Hasan, who is affiliated with Lamar University in Texas, and submitted to arXiv, an online repository for research preprints, in June 2025. Unlike the airport study, it was not presented as a peer-reviewed journal article, and its findings came from simulations rather than a live commercial vending fleet.

The proposed system equips vending machines with low-cost sensors that monitor cabinet and motor temperature, vibration, electrical current and usage. A microcontroller processes the readings and sends updates to a cloud platform, where machine learning models classify the machine's condition and estimate whether a failure is developing. A dashboard then provides service alerts, confidence levels, recommended actions and priority rankings for technicians.

Using synthetic sensor data, Hasan compared Random Forest and Long Short-Term Memory models. Random Forest delivered the stronger results, achieving 94.2% accuracy and 95.6% recall, although those results still require validation with data from operating vending machines.

In a separate six-month simulation involving 20 machines, the proposed system reduced unplanned downtime by 32% and unnecessary technician dispatches by 27% compared with scheduled preventive maintenance. Mean time between failures increased from 21.3 to 28.4 days, while mean time to repair declined from 2.4 to 1.6 hours. The paper estimated that existing machines could be retrofitted for less than $50 per unit using commodity sensors and open-source software.

Those figures are promising, but they should be interpreted as modeled outcomes rather than proven field results. Real machines would introduce additional complications, including inconsistent sensor calibration, wireless connectivity, different component configurations and incomplete maintenance records.

For operators, the strongest idea may be the shift from servicing machines according to a fixed calendar to servicing them according to condition and risk. Better fault warnings could help route technicians more efficiently, reduce emergency calls and allow parts to be replaced closer to the end of their useful lives. The same data platform could eventually support inventory monitoring and refill scheduling, creating a more unified view of machine health and product availability.

Looking ahead

Both papers illustrate the difference between installing technology and making it operationally successful. Airport kiosks need intuitive interfaces, visible privacy safeguards and a journey that does not simply move the passenger from one line to another. Predictive maintenance needs dependable sensors, representative training data and integration with real dispatch and inventory systems.

The kiosk study provides evidence from actual travelers but measures stated intentions. The vending paper offers detailed performance projections but still requires validation in the field.

For deployers, the next step is the same in both cases: Test the technology under real operating conditions, measure what users and technicians actually do and refine the system around those results.

About Richard Slawsky

---