Follow Datanami:
August 16, 2016

Finding Hidden Value in IoT Analytics


Sensors on the Internet of Things (IoT) are already generating a ton of data, but turning that data into business value isn’t always easy. In particular, businesses must be creative in how they mine their IoT data, which can yield unexpected gems.

By the year 2020, there will be 25 billion devices on the IoT, according to Gartner. These devices will generate zettabytes worth of largely time-series data for a myriad of uses—from locomotives and jet turbines to fitness trackers and connected cars.

Much of the data will be transactional in nature. But a big chunk of it will be used for analytic projects, according to Dan Graham, an IoT analytics of enterprise systems for Teradata (NYSE: TDC).

“In the majority of use cases, the customer is going to get the biggest ROI out of the analytics of things, out of analyzing the data,” Graham tells Datanami. “That doesn’t mean that operational data doesn’t have value. But the real multipliers come when you’re seeing things in your business model or in your overall deployments where you say, ‘Aha, that needs to change.'”

Graham has worked on a number of IoT projects for customers of the data warehousing giant, and has identified certain deployment patterns and behaviors of companies that get the most out of their analytic investments. In many cases, businesses can reap greater rewards with IoT analytics if they open their eyes and look for innovative ways to maximize their use of the data.

For example, one company that makes machines used by cardboard manufactures found a particularly creative way to use Teradata analytic tools. The company has worked for years to optimize operations and to minimize the need to take a machine down for maintenance to replenish “consumables,” like belts, felts, and chemicals.

A cardboard machine manufacturing company was able to boost revenues by $35 million from a simple report (Gemenacom/Shutterstock)

Following an IoT analytics engagement, a simple BI report enabled a cardboard machine manufacturing company to boost revenues by $35 million  (Gemenacom/Shutterstock)

“It’s like a Dreamliner,” Graham says. “You don’t really want it to stop flying. They want it to run 24 hours a day, week after week.”

The company used Teradata software to analyze sensor data to optimize the machine, which was all well and good. But the biggest value came elsewhere.

It turns out that many of the company’s customers bought consumables from third-party vendors. “The Chinese knock-offs were killing them,” Graham says. “But what they discovered was their consumables lasted longer. They were actually better consumables.  The belts and the felts would last a month longer than the knockoffs.”

By generating simple reports that showed the superiority of the company’s consumables compared to the lower-cost Chinese goods, the company was able to boost sales by $35 million.

This pattern has played out over and over, according to Graham. “It’s the bigger-picture use cases that go beyond that, and in many cases open new revenue streams,” he says. “We see that in most of our manufacturing customers. They start out trying to build a narrow use case, which is usually condition-based maintenance, just trying to make the machine run better, or doing root-cause analysis for engineers to design better machines.

“That’s now morphed,” he says. “Most of them who have their sea legs under them have moved over to identifying new revenue opportunities. ‘I can sell this data. I can sell the analytic results.'”

Identifying new business opportunities through IoT analytics isn’t easy, particularly when deadlines and budget constraints require project leaders to stay laser-focused on accomplishing their set goal. The big data industry is rife with stories of analytic ventures gone awry, of gold-seekers who struck out by venturing too far outside the bounds of what’s relevant to the business.

Identifying Profitable Data Niches

There’s a wealth of information hidden in IoT data, but the trick will be extracting it in a way that’s beneficial to the company. In most cases, that will require the services of a professional data scientist or analyst who’s comfortable working with advanced algorithms that can transform data from a mass of bits and bytes into actual information that can be acted upon.

“You don’t need a full data scientist,” says Graham, who has 40 years of experience in the analytics and BI industry. “You might just be looking for a good data miner [who’s proficient in tools] from SAS or Fuzzy Logix. They don’t have to be in the business of finding discoveries. They have to in the business of transforming data, which are neighbors but they’re not the same.”

This approach is playing out in Volvo’s effort to harness IoT data and predictive analytics to build a “death proof” car by 2020. The Swedish carmaker is using Teradata’s software, running across thousands of servers according to this Wall Street Journal article, to analyze petabytes of data in pursuit of its goal.

Volvo is using Teradata software in its quest to build a death-proof car by 2020. (Image source: Volvo)

Volvo is using Teradata software in its quest to build a death-proof car by 2020. (Image source: Volvo)

“They went from the short-term [goal of] ‘Lets make the machine more reliable’ to ‘Let’s really up our game so much on safety so much that nobody can touch us,'” Graham says. “That was the bigger payback, because safety in cars of course is, as soon as you get over the age of 25, the insurance companies know you care about safety.”

While data discovery is a big part of analytics, finding new revenue streams hidden in the data isn’t simple. One simply can’t script the discovery of game-changing insights. But there are approaches one can take that will increase one’s odds of making a genuinely impactful discovery in one’s IoT data.

Having an organized and well-run analytics project, complete with governance and security, is important. Having the algorithms and—more importantly—the skills to use them is another important item. In most cases, the data will need to be housed in a data warehouse or a lake that facilitates mixing, normalizing, and correlating data from different sensors and sources.

The goal is ultimately to put a single stream of data into the context of the overall business, Graham says. “The overall business context is usually integrated in a data warehouse,” he says. “It’s hard to do  it any other way. In fact, it’s difficult to do that in a data warehouse. It’s a difficult and rewarding effort.”

But at the end of the day, data will ultimately be a key differentiator between you and your competitors. Whoever can capture and mine IoT data streams in the fastest, the most efficient, and most creative way will obtain the upper hand.

“The real message is, while you’re doing these narrow use cases, look for the bigger picture” Graham says. “Look for ways to inject completely new use cases and completely new revenue streams.”

Related Items:

IoT May Fail Without the Design of Things

IoT Seen Driving AI Adoption

The Data of Things: How Edge Analytics and IoT Go Hand In Hand