Internet of things.

What the internet of things is

The internet of things is a network of physical devices connected to the internet that collect data and exchange it. It includes everything: smart thermostats and fitness bands, industrial sensors, cars, till equipment, meters, cameras.

Such a system is made of three parts. Devices and sensors take readings. Communication channels - wireless networks, short-range links, mobile internet - transmit them. Analysis tools process what is received and turn it into conclusions.

The main difference from familiar data sources is that readings are taken automatically and continuously. Not a person filling in a form once a day, but a device transmitting its state every minute.

Let us look at what this changes in data collection and analysis, where it benefits a small company and where it remains an expensive toy.

What changes in data collection

Connected devices expand collection capabilities in three directions at once.

Automation. Data enters the system without manual input. This removes not only labour but also errors: forgotten entries, typos, delays.

New sources. Metrics appear that simply did not exist before. How many people entered the premises and how many stayed longer than a minute, at what temperature goods were stored, how often equipment was used.

Volume. Instead of a snapshot once a month - a continuous stream. This lets you see not the outcome but the process: exactly where and at what moment something went wrong.

The flip side is obvious: the volume of data brings no benefit by itself. A company that collects readings but does not look at them is simply paying for storage.

What changes in analysis

A continuous stream of data gives three new possibilities.

Real-time reaction. A deviation is visible immediately, not in last month's report. A refrigerator out of its temperature range, a queue at the till, idle equipment - all of these become a reason for immediate action.

More accurate forecasts. Models built on detailed data predict demand and load better than estimates based on aggregated figures.

Personalisation. Knowing exactly how a client uses the product allows offers tied to their real behaviour, not to a segment. How this works in marketing is covered in our piece on personalisation.

All three possibilities require one condition: the data must land in the same system where you already count money and clients. Otherwise it will remain a separate screen nobody looks at.

Where it already works for small companies

Talk about the internet of things often sounds like talk about factories. In practice small business has several clear applications.

• Retail. Sensors at the entrance count visitors, and you get real conversion from passers-by to buyers - rather than guesses about how many people came in.
• Catering and food. Monitoring storage temperature removes the risk of spoiled goods and complaints from inspectors.
• Services with equipment. Car washes, laundrettes, gyms: you see real utilisation by the hour, which means you can change schedules and prices.
• Equipment rental and maintenance. Usage readings allow maintenance to be planned before a breakdown, not after.
• Logistics. Tracking movements gives honest delivery times instead of promised ones.

What all the examples share is that the data lands where decisions are made. A visitor counter is useful not by itself, but together with revenue for the same hour.

Where it is not needed yet

It is worth naming the flip side honestly too. Implementation is not always justified.

If the business works entirely online, physical sensors will add nothing: web analytics already collects all the data you need.

If the company has no basic record-keeping in place - enquiries get lost, advertising spend is not reconciled with revenue - a new data source will not improve the situation. It makes sense to first put in order what you already have, and here a customer records system helps.

If there is nobody who will regularly look at the readings and make decisions based on them, the equipment will pay off only in the supplier's report.

And finally: cost. Sensors, connectivity, data storage and processing are a recurring expense. Before implementing, it is useful to work out which decision it will enable and how much that decision is worth in money.

Data protection: a mandatory part

Connected devices are additional entry points into your network, and they need to be treated accordingly.

The basic minimum: change factory passwords, put devices on a separate network segment, update software regularly, restrict access to control panels. Skipping any one of these turns useful equipment into a vulnerability.

Personal data is a separate question. If devices collect information about people, it must be handled under the same rules as any client database: a clear purpose for collection, consent, restricted access. The link between data protection and client trust is covered in a separate piece, and the ethical side of working with large data sets - in our article on data ethics.

The order of implementation

1. Formulate a question you currently have no answer to. Not “let's install sensors” but “I don't know how many people come in and how many of them buy”.
2. Check whether the answer can be obtained more cheaply - from till data, from the accounting system, by observation.
3. Choose the minimum set of equipment for that question, not a full kit “to grow into”.
4. Link the readings to the system where you already look at sales and clients.
5. Appoint someone responsible who looks at the data at a set frequency.
6. After a couple of months, assess: have the decisions you make changed?

The last point is the main one. If the decisions have not changed, the data is being collected for nothing, and it is more honest to wind down the experiment than to expand it. How to link any collected data to money is described in our piece on data analysis, and we take on non-standard tasks of this kind as part of our special projects.