ARI Fleet Management manages 1.2 million things with wheels across North America and Europe, from telephone company trucks to corporate vehicles to railroad maintenance trucks.
The telematics sensors on its rolling stock of vehicles capture data between every three and 30 seconds. "Every two weeks, we get the equivalent of all the data we've accumulated in the last twenty years," says Bill Powell, director of enterprise architecture for the Mt. Laurel, N.J.-based firm.
It's a carousel of information: ARI can tell from its gyroscopic sensors if drivers are jackrabbiting from stops or slamming on their brakes; it can tell from engine sensors that they're letting the engines idle too long.
One of the most intriguing and granular pieces in all these terabytes of data is the one that compares where a gas credit card was used, based on the geocode of the vendor, and where the vehicle was at the time. If the differential is more than 20 feet, an ARI audit can show that someone was fueling an unauthorized vehicle.
As that example shows, the internet of things (IoT) isn't just about sensors and data, it's about using data in context. That makes it an interdisciplinary challenge for IT executives, one that encompasses information technology, operations and business processes.
IT practitioners with any tenure would be forgiven for expressing skepticism about IOT's maturity. For years they've been reading about sensors that would allow vending machines to report that they were full (and thus didn't need to be replenished), or about RFID-enabled smart supermarket shelves that could help with inventory control -- interesting developments, but not the kind of progress that has the potential to transform a business from the ground up.
That progress may finally be underway. In its May 2016 forecast, market research firm IDC predicted that spending on IoT will grow from $692.6 billion in 2015 to $1.46 trillion in 2020, with a compound annual growth rate of 16.1%. And the installed base of IoT endpoints (that is, the "things" or sensors) will grow from 12.1 billion in 2015 to more than 30 billion in 2020.
Matthew Littlefield, president and principal analyst of LNS Research, who has conducted an IoT survey for the past two years with the Manufacturing Enterprise Solutions Association, says IoT is increasingly on the radar for manufacturing companies. In 2015, 44% of survey respondents said they "do not understand or know about IoT"; in 2016, that number dropped to 19%.
Four lessons from early IoT adopters
Computerworld spoke to a number of early IoT adopters, entities that have gotten their hands dirty in everything from manufacturing and logistics to smart cities and agriculture. Almost all report bumps along the way, but also say they have either achieved or anticipate significant payoffs from their investment. With IoT at last becoming a force in the enterprise, here are four lessons to heed.
Lesson No. 1: Prepare for a deluge of data
Bill Powell, ARI Fleet Management
ARI's Powell acknowledges that the technology his company used even as recently as five years ago was "very immature," limited in some cases to GPS data. "Originally, it was just bread crumbs. We could see on a map a vehicle's movements for the previous five days." Now that data is collected every 30 seconds. "The amount of data has become more like a fire hydrant," he says.
In Powell's opinion, data is perhaps the biggest issue that CIOs have to deal with. "Don't underestimate the volume of information," he warns. "People think, 'If I collect the info, I'll have an epiphany of business value.' That's not the case. You may be able to do that with lower volumes of info, but you'll drown if you try it with telematics. Don't collect data just because you can. Push the business logic [as close to the sensor] as you can."
Jon Dunsdon, CTO of Evendale, Ohio-based GE Aviation, concurs. His company has been monitoring data on jet engines for 20 years. "One of the challenges is that very few aircraft stream data continuously. You do the initial analysis onboard, and then when you land, you transmit the information," he explains. "Ten years ago, it might have been 3.2 kilobytes of information; today, hundreds of megabytes come in when the plane lands."
GE Aviation captures hundreds of megabytes of data anytime a plane lands.
At the same time it's collecting information from the aircraft, GE Aviation is also compiling other data to create additional value for its customers, which include United Airlines and Southwest Airlines.
"We look at things like scheduling data, weather data, airport curfews. If there's a storm in Chicago, how does that affect decisions? How can you reduce cost and disruption?" Dunsdon deploys a data lake to accommodate the volume of analytics and increase performance and recommends fellow IT executives consider the same.
BASF, the Ludwigshafen, Germany-based global chemical company, has a billion-euro machine known as a steam cracker that's the size of two football fields. Its purpose is to convert a petroleum fraction into raw chemicals, and the company uses sensors to calculate the steam cracker's downtime in the millions of euros per hour.
Wiebe van der Horst, BASF
Prior to deploying those sensors, BASF would get only get a few hours of warning before potential breakdowns, says Wiebe van der Horst, senior vice president of global process and enterprise architecture, but that has now changed. "We worked on artificial models using different algorithms, correlating the activity of the machine and its maintenance records," van der Horst explains. After using SAP Hana to analyze the data, BASF can see potential downtime a few weeks in advance. "That's really valuable to us," he says. "We're taking the results of this and applying it to other plants worldwide."
The project would not have been possible without the ability to combine unstructured and structured data. "It was not economically viable to do these kinds of calculations and analysis not so long ago," says van der Horst.
Takeaway: Understand where data is coming from, and determine how you're going to analyze it.
Lesson No. 2: Be open to partnering with operations
IoT incorporates multiple technologies -- wireless networks, sensors, analytics, security -- that encompass not just IT but also operations, particularly factories, equipment and logistics. That means IoT projects inherently require an alignment of business and IT.
BASF's van der Horst says, "In the past, operations technology was done by someone else. Today, every piece of production equipment is designed by our engineering group -- it's not something you buy off-the-shelf. With IoT, to get the data from the equipment, you have to work shoulder-to-shoulder with engineering."
"There has to be collaboration between technology and operations to make IoT work. That's where you unlock the business value," agrees Chris Howell, head of business systems at Gatwick Airport outside London. In addition, he says, you need people who can handle a deluge of data, whether internal data scientists or an external partner who understands big data and analytics.
Howell uses tools from Splunk and Amadeus to keep track of where planes are on the field. The goal: operational efficiency. Although it's deployed a number of IoT-based systems, including self-service check-in and baggage tagging, in the last two years, Gatwick's crown jewel is a laser-guided system on the field. It takes transponder data from the plane and tracks where it is on the ground, how long it takes to get to the gate and when it leaves again.
This helps Gatwick calculate how long planes really spend at the gate. Using the data, Gatwick has been able to increase slots on the runway from 53 to 55 an hour, which Howell says is a world record for a single-runway airport. Other benefits: there are no longer discrepancies in gate pull-backs caused by an airline employee listing one time and an airport employee listing another. Finally, "on a practical level, we're saving four person-hours per day because we no longer have to take data out of a spreadsheet," says Howell.
Takeway: Determine how and when to combine operations and information technologies for maximum data insight.
Lesson No. 3: Be prepared to work with multiple vendors
Bob Bennett, chief innovation officer for the City of Kansas City, is working with Sprint, Cisco and other vendors on a $20 million pilot project (of which the city spent $3.7 million) along a 3.2-mile downtown streetcar line.
"I need to know where people are flowing through the cityscape. I need to know where concentrations of people are located dynamically, so I can react with police resources or utility resources," says Bennett. "If there's a surge in water usage without a corresponding number of people, that may indicate a water leak."
The project combines data from street sensors, streetlight cameras and even mobile phones, flowing into 328 access points along the route and as much as five blocks on either side of the streetcar tracks. In accommodating both technical and operational data, he cites the need to understand from a high level to a highly granular level.
Bennett recommends designating a chief data officer, because you need to establish controls at a macro level to avoid having conflicting data that causes algorithm malfunctions. He also needs to know "who to talk to to make sure Google's application for dynamic spectrum sharing is working."
Takeaway: In orchestrating the many moving pieces of an IoT rollout, make sure you know who plays what part.
Lesson No. 4: Watch for pitfalls
Some early IoT adopters have reported reliability issues with either sensors or vendors or both, and others have struggled to reconcile competing protocols.
Ken Albert is founder of Shelburne Vineyard in Shelburne, Vt. A former electrical engineer for IBM, he took advantage of a prototype sensor system to monitor temperature, humidity, soil temperature and leaf wetness, all of which would help his staff apply fungicides more efficiently. But then the company providing the sensors was acquired, and the new owner began to redesign its sensors and stopped supporting the ones in Shelburne's 17-acre vineyard.
Even before that, however, "the system was less reliable than I thought it would be," in part due to wireless connectivity issues. Albert acknowledges that exposure to the environment probably contributed to the downtime, but that's to be expected in agriculture. A new system would cost $2,000, but Albert probably won't go with the current company.
While practitioners like Albert are out in the field (literally), industry watchers like Dale Pfau are behind a computer screen, lamenting the proliferation of communications protocols that are fragmenting the nascent market.
Pfau, who has been offering strategic advisory services for wireless technology and IoT companies for more than 20 years, cites multiple subsets of Wi-Fi, such as 802.11ad (a.k.a Wi-Gig) and 802.11ah (sub-1GHz); Bluetooth Mesh; Zigbee (some versions of which aren't backward compatible); Z-Wave; and LoRa (for long-range); as well as Sigfox, a proprietary protocol for IoT.
"It'll be a while before it all settles out," says Pfau, currently managing director of Stonecroft Capital.
He also worries about security. "A lot of these products are double-enabled with Wi-Fi and Bluetooth," he says "It's possible to breach networks via a Bluetooth signal and then get into the Wi-Fi network."
Takeaway: Be prepared for setbacks in an immature market, and try to select a protocol that has long-term industry support and a sound security footprint.
Despite the challenges, GE Aviation's Dunsdon believes that companies haven't even begun to scratch the surface of IoT's payoff. "There is much opportunity to change the industrial space. There are areas we thought were already optimized, but now they're going through the same revolutionary changes that banks went through when they shifted from paper checks to electronic money transfers," Dunsdon says. "The industrial industries are seeing massive changes."