Year after year, Internet of Things (IoT) continues to be in the top 5 technologies that companies are making significant investments in. As I noted in my initial article, Role of Data in Trending technologies, Gartner, Forrester, et al. have touted IoT as one of the main technologies to watch. Given this background, this article delves into the key role that data plays in extracting business value from IoT investments. Even though you might be well aware of what IoT is, let me introduce IoT in one sentence as IoT refers to devices other than computers that are connected to the Internet and can send and receive data without human assistance.
Types of IOT
IoT can be categorized in many different ways but for the purpose of this article, I’d like to distinguish between consumer applications of IoT versus industrial use of IoT, which has its own term called IIOT. Consumer side of IoT, like turning off lights remotely or automatically ordering out of stock staple food items like milk, etc. from a IoT enabled fridge may get more attention, but the Industrial IoT or IIoT is where more significant investments are happening. Bain predicts B2B IoT segments will generate more than $300B annually by 2020, including about $85B in the industrial sector. Some examples are sensors on oil rig equipment, transportation fleets, HVAC systems, and machinery on the factory floor. The Industrial IoT connects critical machines and sensors in industries such as aerospace and defense, healthcare, and energy— where failure often results in life-threatening or other emergency situations.
Let me sum up the importance of IIOT by reproducing this sentence from World Economic Forum’s report on IoT. “The Industrial Internet [of Things] will transform many industries, including manufacturing, oil and gas, agriculture, mining, transportation and healthcare. Collectively, these account for nearly two-thirds of the world economy.” — World Economic Forum, Industrial Internet of Things Report
Unique characteristics of IoT data
The data generated in IoT appliances and devices has some unique characteristics that need to be handled differently than data generated in conventional corporate transactions such as sales operations, customer operations, product operations, etc. So let’s first look at these unique IoT characteristics.
One of the most obvious characteristic of IIoT data is its high frequency. High frequency of data. Examples of this high frequency of data is from vehicle sensors (multiple messages per second), data from video cameras (continuous streaming), or data from multiple sensors on oil rigs, and so on.
A second unique characteristic of IIoT data is the large volumes of data that these devices generate. The challenge associated with these large volumes is both with respect to storing (or only storing relevant data) and processing the volumes of data. It can be overwhelming for organizations to deal with these large quantities of data.
The third differentiating characteristic of IoT data is that the data generated is not structured as it can be images, videos, random text which is very different from a structured data generated in a traditional corporate transaction. Identifying the various formats, processing them, and analyzing the data can be a gargantuan task without the appropriate tools.
By now, you have realized the common thread between IoT data and what is called big data with respect to their 3 Vs— i.e. the Velocity of data, Volumes of data, and Variety of data that is being generated by IIoT devices and sensors. The challenges are plenty— such as how to store/what to store, how to decipher noise vs. relevant data, how to parse for relevant data, and most importantly how to analyze this data to deliver some meaningful insights.
This is where the big data technologies such as Hadoop clusters, NoSQL databases, in-memory data grids, real-time data integration tools, and stream processing technologies come to rescue to store/manage data for later use, do real-time analysis, and produce visual insights. Once the data is managed using these technologies. predictive analytics, data mining, and big data analytics tools can be used for downstream analysis.
IOT Data Analytics
In this section, I’ll discuss data analytics as it relates to IIoT data. Data Analytics, as you very well know, is a process used to analyze data sets to extract meaningful conclusions. These conclusions, such as trends, patterns, and statistics, are intended to help business organizations in effective decision-making. In the past, I have written about different types of data analytics including descriptive analytics, diagnostic analytics, predictive analytics, and prescriptive analytics. IOT data analytics are also intended to arrive at one or many of those business objectives but how you get there differs from data analytics for traditional corporate transactions.
Given that data generated in IOT is unique and distinct from data in conventional corporate operations, the data analytics resulting from this data also needs to be thought of and dealt in different ways other than traditional ones. As mentioned above, big data technologies applied to IoT offer automated mechanisms for processing/managing machine or sensor data into Hadoop clusters and other big data platforms for analysis. The type of IoT data we deal with lends itself to few different types of analytics as mentioned below:
- Time Series Analytics: When IOT data is time-based data like in machine log data or health monitoring data, time-series analytics will identify trends and patterns.
- Streaming Analytics: Real-time data streams from sensors can be dealt with streaming analytics to detect anomalies resulting in emergency situations requiring immediate attention and are typically used in fleet maintenance, machine traffic analysis, financial transactions. Streaming analytics can also sometimes be referred to as stream processing or in-motion data analytics.
- Spatial Analytics: Location-based IOT applications dealing with geographic data from sensors (for example parking meter sensors) that rely on spatial analytics and is used to identify geographic patterns and their relationship to physical objects.
Data Governance for IOT
So far, I wrote about the differences for IOT data and IOT data analytics compared to conventional corporate data, but one common theme is the importance of data governance. From my consulting engagements in these areas, I realized that data governance is even more important for IOT data management. Primary reason is that the data processing can quickly get out of control without a set of procedures, standards, and data quality.
The global IoT market will grow from $157B in 2016 to $457B by 2020, attaining a Compound Annual Growth Rate (CAGR) of 28.5% as per this 2017 roundup from Forbes. According to the same article, Bain predicts B2B IoT segments will generate more than $300B annually by 2020, including about $85B in the industrial sector. The market is real and the data generated from IoT devices and sensors is unique and needs to be processed using big data technologies. The unique characteristics of IoT data drive us towards a unique way of dealing with data analytics in IoT space. But one common thread to traditional markets is the importance of data governance in managing IoT data.