Data analytics as the next step to IoT
By Miguel González
If you have found yourself in front of a lot of tables, graphs and raw data, collected by sensors or machines, surely you have asked yourself:
What is all this data for? Can I get valuable information from them? 🤔
In this article I will help you answer these questions and understand: why analytics is the necessary next step in data collection for decision making.
Context: The IoT Today
With the great growth that the IoT has had in recent years, companies have begun to collect a lot of information about their processes, thanks to cheaper sensors and space in the cloud every day.
Including several things, from times of entry and exit of people in the workplace, through electrical consumption of machines in the industry, GPS tracking of trucks, temperature of refrigerators and even water flow in an irrigation equipment in the middle of the field.
The applications are endless, but the end is the same. Have an objective view of the processes based on the data obtained, to make decisions that improve the performance of the company.
The problem: The gap between data and decisions
Although data collection is necessary, in most cases it is not enough to obtain an objective overview when making certain decisions.
It takes another step, a fundamental link between data and decisions is needed. Something that shapes the data, that translates it into clear and relevant information for the decision you want to make.
This next step is: Data analytics 📊
What is data analytics?
The idea is to go beyond the raw numbers.
It is giving context to filter, group, process, cross, categorize and even predict future behavior based on the past.
But always with the aim of obtaining useful and valuable information, which otherwise could not be known for sure.
Data analytics explained in an example
Let’s look at an example to make the concept clearer.
Suppose we have an industrial plant with 10 equal machines, which should always be running and that we measure their electricity consumption every 2 minutes.
As a certain period of time passes, we are going to have many consumption values for each machine.
Perhaps, they are quite similar, and even in certain periods there will be no consumption because the machine broke or stopped.
Now suppose we wanted to plan the maintenance of each machine.
Consumption data by itself, even if we graph it, would not tell us much, we would even tend to think that it is of no use to us
However, with analytics, knowing that our goal is to plan maintenance, we could process consumption data to obtain value for that purpose.
For example, we could calculate for each machine:
- What percentage of time was it not working
- Each time it stops, how long does it take to work again?
- How many times do you stop for a problem
- How often does it stop on average
- The time it was operating outside its nominal limits.
We could even go one step further and categorize the data from all the machines to get more interesting results.
For example, grouping them by brand, we would see which brand spends less time downtime and requires less maintenance.
If we add to this that, based on consumption, we can know how much money each brand spends per month on energy, and predict how much the next one will spend, we could, for example, know which one to choose in a next purchase taking into account maintenance and consumption.
Answering the two initial questions: What is all this data good for? Can I get valuable information from them?
They are always useful, even to make decisions that a priori would seem to be unrelated to the data we have, as in the case of the example: data on the electrical consumption of the machine for maintenance planning.
However, it is important for the data to be useful, to be clear about what decision we want to make, or what conclusion we want to reach, and then through analytics, use it in the best possible way with that objective.
At the end of the day, data will have value if there is a need for which it is useful.
That is why analytics is so important, because through it we can see the data interpreted in the way that is most useful for that particular need.
About the author: Miguel González
Full-Stack IoT Engineer
Miguel is a Telecommunications Engineer and Master in Software and Network Systems Engineering.
At Nettra he specializes in developing and improving IoT solutions, as well as supporting their implementation and subsequent support.