There will be 80 billion connected devices in 2025, says IDC, helping generate 180 trillion gigabytes of new data that year.
The amount of data coming from the Internet of Things (IoT) that is analyzed and used to change business processes will be as big in 2025 as the amount of all the data created in 2020, according to IDC.
We’ve learned in recent years that huge quantities of data are essential to the practical and productive application of new approaches to machine learning such as deep learning. More data leads to innovative analytics and better decisions. But more data could also be overwhelming and distracting, even misleading, slowing down decision-making.
If the data collected and transmitted by the IoT is not relevant, if it is not provided in a timely fashion, if the analysis does not suggest ways to improve a business process or a customer’s experience, then enterprises will not benefit from the IoT. If employees, managers, and senior executives in enterprises big and small, private and public, don’t have positive experiences with the abundance of data, they won’t take advantage of it.
Here are a few considerations to keep in mind when developing and implementing the What, Where, and Why of IoT analytics:
It all start with the data itself—what is the quality of the data collected by the IoT? Automating as much as possible the “cleaning” of the data to reduce costs, eliminate errors, integrate better with other internal data sources, and ensure high-quality analysis. IoT data, probably more than other types of data, comes in many formats, from many sources, and in a continuous stream. It is imperative to set up thresholds, benchmarks, and rules for filtering the data stream so only the significant signals are isolated and then stored and/or transmitted in timely and efficient manner (e.g., using data compression).
Once you get the data up to acceptable quality, the next consideration is Where to analyze the data? Driven by the rapid adoption of the IoT, we hear a lot nowadays about how computing is moving to the edge. But we are also witnessing a parallel move of data, analysis and infrastructure resources to the cloud.
There are a few options to consider: Moving the data to a centralized “data lake” to be combined with other data and prepared for analysis; doing the analysis on the IoT device and only moving the results of the analysis to a central database or dashboard; or creating a distributed architecture with well-defined division of labor between the edge (the IoT device) and the cloud. The right choice depends on the data collected, the type of analysis done on it, and where the organization is in its IT evolution, to name a few important factors. The most important factor, however, is the specific objective of the IoT project or program in question—Why is the data analyzed?
There is a wide range of possible objectives for IoT analytics which can be classified into three major categories: What just happened? what will happen? and what should happen?
In the first category, we find objectives such as continuous monitoring of a process or product or service to define a baseline pattern of activity or show variations due to weather conditions, seasons of the year, time of the day, etc. Another objective that falls into this category is providing an alert if there is a deviation from the norm or if a pre-defined threshold is surpassed or the data value falls below a certain benchmark. An additional level of analysis can also help assess the validity of the alert, to avoid frequent “false positives,” or alerts that turn out to be of no consequence.
The second category is all about the future—predicting positive or negative outcomes for an existing process or product based on the data. One such application, predictive maintenance, is a top reason for investing in IoT for many enterprises today. When a piece of equipment “calls home” with the message “I’m going to fail, come and fix me!” it helps improve customer satisfaction and optimize maintenance contracts. In the case of high-priced equipment or process, avoiding downtime contributes greatly to the bottom-line. Deploying predictive maintenance ensures a different type of interaction with the customer than when the customer reports a failure after the fact.
Finally, the objectives of collecting and analyzing IoT data fall into the category that is all about action—what should be done given the analysis? Recommending the next best action based on the analysis and based on previous experience help reduce complex data to one or more straightforward action choices that nontechnical executives or line workers can consider and act upon. In some cases, when the decision needs to be made in real time, the decision to follow the recommendation could be automated.
A recent IDC survey found out that the top perceived benefits of the IoT are increased productivity (24%), time-to-market (22.5%), and process automation (21.7%). By unlocking the value of IoT data with the right approach to the what, where, and why of analytics, enterprises will realize the promise of the Internet of Things.
A few tips about #IoT analytics from Gil Press. Take a look at the What, Where and Why of IoT: