Companies striving to become data-driven, first focus mostly on the actual processing of data and the creation of smart machine learning models. However, to gain value out of these models, the outcome needs to be made available for everyday use. After all – data and models cannot just be made available to anyone as is.
In your data-driven journey, an important step is to embed the data insights into the organization. There are several ways of doing this. One is to smartly visualize the insights from data and models into dashboards, and to embed looking at these dashboards and taking actions upon them into business processes. Another is to directly act upon the insights, for instance by creating an app that generates alerts when human interference is needed. A third way is using the data to directly control business processes or machines, for example using it to directly tune machines in a factory. In those cases, software is needed to make data-driven insights actionable.
And for any of the cases, granting the right people access to the right information is paramount. In the backend, we need to ensure proper authorization, authentication, and we need to create API’s that allow for interoperable use of data-driven insights between different systems.
A case in which an API is used to act upon data insights is our work for TNO, an independent research organization. In their automotive branch they use sensors to collect all kinds of data to enable testing of software for self-driving cars. Algorithms are used to split and classify this sensor data to create scenarios and simulation data for those scenarios to enable the testing. One such scenario could be overtaking a car on an 80 km/h road in the Netherlands when it rains.
All these steps take place within our data factory, which continuously ingests data from sensors, uses different data science techniques to classify the data, and enables the creation and storage of testing scenario’s, and leads to a large, dynamic database containing a multitude of different scenarios. Merely collecting these, however, is of no use. Therefore, we created an API together with TNO that makes the scenario’s accessible to OEM’s, who create autonomous driving software. This allows these OEM’s to retrieve the scenarios they need from the database to test their software. For instance, an OEM requests a scenario in Europe, for a specific road surface, in which a car is being cut off.
The API is an example of embedding results of data science models into business processes. It allows the OEM’s to remotely access the scenario’s and to test their software components on them. An important step, as the testing cycle for autonomous driving software is only becoming longer. This way, manufacturers are able to optimize their software in the early stages of the testing cycle – saving valuable time – before taking it out on the road.
"An important thing to keep in mind is the step to transform insights into practice."
Another example of embedding data-driven insights into daily practice would be our robotic office dog named Eddy (@I). He collects all kinds of data with sensors that act as his eyes, ears, and other senses. While training him for use in our internal processes, we do also need to instruct our team on how to work with Eddy. For instance, if we want Eddy to direct guests to meeting rooms, it needs to be in his code how to work with our receptionist and our guests. Software is used to make sure the receptionist receives a text message every time a guest has been successfully dropped off at the correct meeting room. Again – software is crucial to integrating him as part of our day to day office life.
All in all, the journey towards becoming data-driven is not linear. And we get that. An important thing to keep in mind is the step to transform insights into practice. Whether you start small or go big – the factory line of thought helps you in streamlining this process. Using software as a means to make insights actionable.