The Internet of Things: Overcoming Data Difficulties to Tame It
With the advent of low cost sensors, ubiquitous connectivity, and massive amounts of data, the Internet of Things (IoT) is set to change the world. We’ve all heard the estimates of the billions of dollars and items that will come into play by 2050. However, that’s only the tip of the iceberg. While solving the problems associated with “things” will be critical to unlocking the true potential of IoT, most importantly, overcoming data challenges.
This ranges from extracting data from terminals, machines and remote platforms to interpreting the data to boost productivity and increase performance. Whether it is for a connected home, a portable terminal or an industrial solution, there is often a lag between collecting the new data and presenting the analyzed information in such a way that it can be understood and explored in great detail.
There are three key elements in overcoming these obstacles and taming the IoT:
When it comes to data analytics, every question we ask ourselves about data requires its own graph and visual perspective. This is especially true for the explosion of data coming from the sensors that form the basis of IoT. Unfortunately, most IoT applications come with “one size fits all” views. They answer a set of predetermined questions, deemed worthy of being answered by a small group of “experts”, whether they are the health experts behind Fitbit or the engineers who created the Predix platform. GE.
To fully harness the potential of IoT, tools must be much more flexible and must allow users to shape and adapt data in different ways, depending on their needs or those of their organization. Interactivity, exploration in detail and sharing are fundamental to make IoT data useful, without requiring a huge project around this data. Ideally, users will be able to have informal and in-depth conversations with their data, while also exploring other data to discover all kinds of changes. They can sometimes even reveal previously unknown trends.
For example, you might have an IoT application that analyzes historical data of the activity of a failed engine, gas turbine, or locomotive, and determines the conditions that are causing the malfunctions. as well as how often they are likely to occur. But how do you know which parts are the most fragile? Which factories made them? And what is the date of manufacture? Or which vendors caused the most problems? Interactivity and the possibility of sharing information are fundamental to finding the answers to these questions.
To get answers, interactive data analytics are not enough – IoT data must also be associated with additional context.
Let’s start with a real-life example: You want to combine your Fitbit data to possibly find a link between your exercise program and your sleep cycles. You ask yourself the following questions:
How does my daily physical activity influence my sleep cycles?
Is my performance better when I sleep a lot?
Fitbit’s native dashboards only allow you to analyze exercise data in isolation. However, if you export the data, you can combine that information with other information, such as tracking your physical activity and food intake, body measurements, and sleep cycles. Exporting data is not necessarily the ideal method, but sometimes it is the only way to broaden the scope of the analysis.
Now imagine that you are merging disparate data into actionable insights for your business. Sensors built into aircraft engines can help determine when maintenance is needed. This would anticipate possible failures and save billions of dollars. In addition, integrating data from these sensors with other information can also reveal the savings made against planned budgets by product and region, for example.
The export of data (knowing that this is not the ideal method), brings us to a last important point: we live in a world where it is more and more utopian to have “perfect data”. Your data, however organized it may be, is likely to be stored in a source to which you do not have access. They may also not include certain key elements that are necessary to answer your questions, or be formatted in such a way that their in-depth analysis becomes complex. IoT applications suffer from the same drawbacks, especially when there is no consensus on standards and protocols for supporting device interoperability.
However, rather than let incomplete or shoddy data cripple our business, we need to use what we have and iterate until we find the right solutions. As you iterations, you learn to distinguish between “acceptable” data and poor quality data. Acceptable data is usually sufficient to answer most, if not all, questions. In addition, understanding the gaps in some data improves the process for collecting and processing it. This will help you troubleshoot issues with your data collection and integration processes. Ultimately, this will help us all tame IoT faster.
Edouard Beaucourt, Director France, French-speaking Switzerland and North Africa, has joined Board in 2013 as Commercial Director for Large Enterprises. Previously, he was the Territory Sales Manager for the Professional Analysis Tools sector at IBM. He also worked in the sales department of Clarity Systems, Microsoft and Hyperion Solutions.