OperationsFramework / modelModelAccessible

Sensor data

How can sensor data improve people, teams, or organisational effectiveness?

AccessibleOperationalOrganisation3 min read

Contents

A sensor is a device, usually electronic, that takes a physical quality such as temperature or light, measures it and converts it into information or data that can then be analysed for insights.

A sensor converts a physical, chemical or biological condition—such as temperature, light, motion or pressure—into a signal that can be recorded and analysed. Sensor data can reveal conditions continuously or at a scale that manual observation cannot match.

When to use it

Use sensors when a decision depends on a condition that can be measured reliably and the benefit justifies deployment, maintenance, security and data governance.

Smartphones combine location, acceleration, orientation, proximity, ambient light and short-range communication. Vehicles use proximity and other sensing for assistance; environmental networks monitor water, soil and weather; industrial systems detect vibration, heat, pressure and flow.

Start with the decision rather than the device. A sensor is useful only when its accuracy, location, sampling rate and response process match the phenomenon.

Origins

Sensors evolved from mechanical measuring instruments, electrical transducers and twentieth-century control systems. Advances in microelectronics, wireless communication and computing made sensors smaller and cheaper, while the Internet of Things connected devices and services at large scale. No single invention accounts for the category.

What it is

A sensor system includes the sensing element, calibration, signal conversion, timestamp, communication, storage and interpretation. Context—where the sensor was placed, how it was configured and what changed around it—is part of the data.

Connected devices can coordinate actions, such as adjusting building systems when occupancy changes or alerting maintenance when vibration indicates possible failure. Automation can improve safety, efficiency and response, but it can also create cyber-physical risk when false data trigger the wrong action.

Why it matters

Sensors extend observation across time, distance and volume. They can detect deterioration early, improve energy use, support precision agriculture and measure process conditions.

They do not remove human responsibility. People must decide what to measure, validate performance, interpret anomalies and manage consequences. Automated monitoring can shift work rather than eliminate it.

How to use it

Define the decision, physical variable, acceptable error, response time and operating environment. Select a sensor whose range, resolution, accuracy and durability fit. Establish calibration, maintenance and failure detection before deployment.

Pilot placement and sampling. Compare readings with a trusted reference and test false positives, missing data, drift, latency and environmental interference. Design the action workflow: who receives an alert, what evidence they see and what safe fallback applies.

Secure devices, communications and update mechanisms. Minimise collection, restrict retention and access, and conduct privacy and safety assessment where people can be identified or inferred.

Possible data sources

Common sensor classes include:

Temperature – detects thermal conditions and change.

Light – measures illumination for control or environmental observation.

Pressure – monitors fluid, gas or mechanical pressure.

Moisture – detects water content or humidity.

Level – measures the amount of material in a vessel or environment.

Movement – detects motion, acceleration or vibration.

Proximity – estimates nearness or presence.

Each produces different error patterns, units and calibration requirements.

How difficult or costly is it to collect?

Hardware cost has fallen, but total cost includes installation, connectivity, calibration, power, maintenance, data engineering, security, review and eventual disposal.

Sensors can collect consistent high-frequency data, yet accuracy is not automatic. Cheap, poorly placed or drifting devices can produce a large quantity of misleading evidence.

Practical example

A small fashion retailer wanted to understand window interest, store entry and purchase. A footfall system combined aggregated pass-by and entry counts with transactions, allowing the company to compare displays, offers and conversion.

The original implementation observed mobile-device signals. Such tracking can identify or single out people depending on the method, so a current deployment should prefer the least intrusive technology, display clear notice, avoid persistent identifiers and complete legal and ethical review.

The analysis also challenged an earlier location study: one store had too little passing traffic to support its economics. Management closed it and redirected resources. Before attributing the result to the sensor, the company should validate counting accuracy, seasonality, weather and whether footfall caused or merely accompanied sales.

Top practical tip

Write the decision rule before collecting data, then calibrate the full system in the operating environment. Measure data quality, device health and response performance as carefully as the physical variable.

Top pitfall

Do not collect identifiable movement, location, health or behavioural data simply because a sensor can. Minimise by design, explain the purpose, secure the device and provide meaningful rights and safeguards.