IoT initiatives are fundamentally about solving problems and improving business outcomes. These are goals that are achieved by leveraging data from connected devices to analyze “thing” behavior and derive actionable business intelligence. The desired outcome depends on unique organizational needs, and examples could include improving safety of fleet drivers, implementing preventative maintenance for factory equipment, or more accurately tracking real-time location of donated organs in transit. At minimum, data collected via IoT sensors and other devices provide operational metrics, however when analyzed and paired with other data sources it can provide valuable insights into crucial business decisions. It is critical for organizations to have a formalized strategy as to how this data will be used before scaling out their IoT solutions.
Defining the Business Problem
To achieve desired outcomes from IoT initiatives, it is important to define the business problem you aim to solve at the outset of a project. This will not only drive the “IoT solution” at the front end (choice of device, choice of network, choice of application or cloud platform), but also the analytics strategy that sits behind it. Think of a fleet tracking and optimization use case – in this scenario, the data collected from fleet vehicles can range from simple location tracking, to metrics on engine start and stop times, to hours-of-duty for drivers. The downstream data analytics can enable vehicle efficiency metrics, accurate driver payments, as well as automated reporting for compliance with government-mandated policies. Identifying the intended analytics needs upfront will empower you to select a device that is capable of collecting your desired data points, while also guiding the strategy for integrating IoT data sets with the right set of non-IoT data sets to provide the desired outcome. In the fleet example, one would combine the engine ignition times with the vehicle model and weight to calculate the optimum start and stop times, enabling better fuel efficiencies and reduced fuel costs. It’s important to note that every expected outcome should be expressed in financial terms.
Defining the Datasets
The types of “sensor data” points that IoT devices collect will define the types of data analytics that an IoT solution will deliver. With that in mind, the next step is to define which data points will be collected, understanding that sensor data can vary greatly depending on the device you choose. Going back to the fleet example, there are basic devices that are only capable of providing position data (latitude, longitude) and there are also more advanced devices that can provide engine, braking, and acceleration data. The next data sets are internal from your organization or from external sources, and these data sets combine with the sensor data to provide the right outcome. For example, combining position data with traffic information and studying it for a period can enable a fleet company to create automated routing rules, which in turn can lead to significant fuel savings. Combining your desired sensor data sets with other internal or external data sets enables greater analytics capabilities and visibility into business performance.
Another type of data which can be useful – especially as an IoT solution scales to thousands of deployed devices – is meta data. Meta data is typically obtained from the connectivity network and can lead to identification of anomalies in data collection, helping immensely to ensure the security of your solution. For example – imagine a national ambulance management company’s IoT network is compromised. Analyzing data patterns at the network level can enable early detection of data traffic that is not “normal” so appropriate action can be taken to disconnect the solution or proactively resolve the issue.
Defining the Analytics Strategy
An IoT solution will typically have multiple layers of analytics to be performed, and the key to doing this successfully is to make sure your data is clean. The first step hence is to ensure that the data coming in from the sensors is clean and free of noise, which is usually achieved while the data is streaming. Streaming algorithms not only clean the data, but can also perform some of the computes as the data comes in. Taking the same fleet management example, the data coming in from connected devices provides the position in terms of latitude and longitude. This can be then used to compute geo-location, run-times, and driver hours. Combining this with a geo-fence information, one can also compute alerts around the same in real-time and all this is done as stream analytics. Next, the data is stored in a data lake and combined with other internal or external data sets to create the analytics solution for the business outcomes expected. A typical analytical solution will use a combination of a clustering, classification, or regression techniques to form an algorithm. Another consideration is which of these techniques will also need historical and training data models. Getting the analytical solution well defined will really unleash the power of the collected data and enable richer outcomes.
In conclusion, one should adopt a simple structure of defining the business outcomes upfront, defining the data collection, and defining the analytics strategy, to create an effective analytics solution that uses your IoT data to maximize returns on IoT investments.
To learn more about navigating the complexity of IoT, visit the KORE Reporting & Analytics page for more information.