Much of the talk about the Internet of Things (IoT) focuses on the “things” themselves – wearables, sensors, iBeacons, and other network-connected machines. However, the greatest value for organizations comes from combining the data generated by these devices with other customer or operational data to uncover insights and establish predictive models. This is the incredible promise of IoT, but without the ability to link data from the smart, networked “things” with other business data, its value is limited.
Instead of isolated IoT projects that aren’t connected to the core corporate data infrastructure, there is an opportunity to use IoT data to create business value. Organizations can combine customer data from devices with other customer data to gain new insights and identify new signals of potential churn or propensity to purchase. These insights can lead to a deeper understanding of and greater responsiveness to customers. For example, retailers can combine data gathered from in-store iBeacons with customer transaction histories and store behavioral models to determine the best promotion to send or other actions to take, such as notifying the store staff that a VIP has arrived.
There are also massive opportunities to put IoT data to work in predictive models to improve maintenance scheduling or provide just-in-time service before a product fails. For example, many owners of electric vehicles receive remote diagnostics reports with information about faults or upcoming service needs based on interpretations of the data the vehicles generate.
The challenge that many organizations face is how to systematically understand the data flowing from “things” and combine it with other relevant enterprise data to create value. IoT data usually consists of custom log files, is sometimes misnamed, and appears unstructured. In fact, IoT data has structure, but it isn’t in a traditional relational or other standard format. The log file structures and included data points vary from manufacturer to manufacturer, model to model, software version to software version or even company to company. For example, the structures of the terabytes of data coming from an individual jet engine differs by airline as well as by manufacturer and model. As another example, farming automation and industrial machine suppliers, all with many different models and variants, have hundreds or thousands of different log file formats coming from their products used by customers across the globe. This introduces significant complexity in any attempt to interpret the data and then format, normalize, and combine it with other relevant data for analysis or operational systems. Without tools to help interpret and parse IoT data, the time and effort to put the IoT data to work is costly, time consuming, and prone to manual error.
In order to fully realize the value of combining IoT data and other enterprise data in an agile way, organizations must leverage modern data management tools to accelerate and automate the processes. Using tools to profile data, intelligently discover its structure, and then automatically parse and combine it with other relevant enterprise data on an ongoing basis makes the IoT data more accessible. This creates an infrastructure for ongoing expansion and evolution in the use of IoT data to understand customers and create predictive models for operational systems.
Pioneers in the IoT space are creating innovative solutions for IoT data that combine interactive, visual models of the data with machine learning, in turn speeding up the ability to derive business value from IoT. Once the data model for a particular device is understood, mapped, and prepared, data transformation and delivery to consuming systems can be automated for production use.
Join Informatica in the IoT data revolution and give us feedback on our new machine learning-powered system for intelligent structure discovery of IoT and other machine data here. And, for advice about big data that will help you keep your project on track, download The Big, Big Data Workbook.
Caption: Machine learning-assisted Informatica Intelligent Structure Discovery