Part of:Drill into IoT data analytics architecture and strategy
Focus the role of data analytics in IoT on actionable insights
Organizations might find it easier to use the limitations of their IoT data analysis project to guide its creation, such as connectivity or battery power.
The real value of IoT rests with data and, more specifically, from analyzing the right data at the right time to drive the desired action.
Organizations must design an IoT deployment that considers multiple complex factors that affect what data is collected, how it's transmitted, when and where it's analyzed, and where and how much of it is stored.
"In a world with limits, you have to consider what you want to do, how much it will cost and whether it's even possible," said Alexander Wyglinski, a professor of both electrical and computer engineering at Worcester (Mass.) Polytechnic Institute and a senior member of IEEE. "You have to know what you're trying to achieve with your IoT environment, your capabilities, what the limits are and what are the restrictions."
Wyglinski cited an IoT deployment proposed by students as a case in point. The students developed a plan to use the technology to manage crop irrigation on the desert's edge in Senegal. The plan addressed the challenges of transmitting data in a remote area; how to power endpoint devices; how much analysis to perform at the edge, given the available power supply; how to power those devices; and whether there was enough power to handle the desired analysis in the field.
"With IoT, there's a lot of engineering decisions," Wyglinski said.
Use case is key
Executives should start IoT plans by determining what they want to achieve so they know what type of data they must collect and the role of data analytics in their IoT deployment, said Geoff Mulligan, founder of the consultancy Skylight Digital and an IEEE member.
"IoT is often this cool buzzword, like cloud, and so many people think just by deploying IoT, they'll solve their problems, but IoT projects often fail because they don't turn the data into anything actionable," he said.
Data analytics in IoT deployments fall into three different categories:
Predictive analytics, where organizations analyze data to understand not only what's happening at any given time, but predict what will likely happen in the future.
Prescriptive analytics, where data from a combination of sources can be analyzed to understand how variables affect outcomes.
When enterprise leaders determine what they want to achieve -- whether to understand what's happening, engage in predictive analytics or use data to prescribe actions -- and determine the necessary data, teams can then engineer an IoT deployment to deliver on those objectives.
Without focus, organizations end up collecting data swamps with very little turned into actionable information and pay for the cost of transmitting, storing and protecting it without gaining anything in return, Mulligan said.
Place analysis into the environment
Organizations have choices for when, where and how the needed data is analyzed.
"Data analysis fits in every segment of IoT data pipelines and recommends appropriate action at every stage," said Sachin Vyas, vice president of data, AI and automation products at Larsen & Toubro Infotech consultancy. "Specifically, most of the statistical and analytical tools used in data management solutions clean IoT data before storing it for further analysis. Scalable data storage manages the ever-increasing influx of data, data visualization capabilities spot trends and take action, and robust reporting engines deliver actionable intelligence."
You have to know what you're trying to achieve with your IoT environment, your capabilities, what the limits are and what are the restrictions.
Alexander WyglinskiProfessor, Worcester Polytechnic Institute
Engineers can opt for analytics capabilities embedded in endpoint devices, edge devices and a centralized location. Complex deployments could require analytics at all three points.
On the other hand, many deployments have limitations that eliminate one or two of the placement options. Multiple factors -- from cost to connectivity to battery power -- affect where, when and how often analysis occurs.
"The complexity involved in generating true value from IoT data is a function of multiple things, including the number of devices, frequency of data, type of data and response time of outcome," added Vyas, co-author of How to Compete in the Age of Artificial Intelligence: Implementing a Collaborative Human-Machine Strategy for Your Business.
Executives must also consider the importance of the data and the protection levels it requires.
Location matters, too, experts added. For example, take the students' planned IoT deployment in remote Senegal, where limited power and data options exist. However, geographically isolated areas aren't the only places where executives face limitations. Executives planning IoT deployments to monitor the structural health of public infrastructure such as bridges will have to consider the cost and complexity of maintaining the endpoint devices embedded with analytics capabilities that draw a lot of battery power.
"You don't want to have to change those batteries every six months," Wyglinski said.
