IT case studybub-9
DATA & ANALYTICS CASE STUDY SERIES
R E P R I N T N U M B E R 5 7 3 8 0
GE’s Big Bet on Data and Analytics Seeking opportunities in the Internet of Things, GE expands into industrial analytics. By Laura Winig
Made possible by exclusive
Global Insights case study sponsor:
C A S E S T U D Y G E ’ S B I G B E T O N D A T A A N D A N A L Y T I C S
Copyright © MIT, 2016. All rights reserved.
Get more on data & analytics from MIT Sloan Management Review:
Read the case study online at http://sloanreview.mit.edu/case-study/GE
Visit our site at http://sloanreview.mit.edu/data-analytics
Get the free data & analytics newsletter at http://sloanreview.mit.edu/enews-analytics
Contact us to get permission to distribute or copy this report at [email protected] or 617-253-7170
LAURA WINIG is a contributing editor to MIT Sloan Management Review.
GE’S BIG BET ON DATA AND ANALYTICS • MIT SLOAN MANAGEMENT REVIEW 1
CONTENTS CASE STUDY FEBRUARY 2016
3 / Introduction
4 / The Software Behind GE’s Industrial Internet
6 / A New Approach to Oil and Gas
13 / Racing to Lead the Industrial Internet
15 / Commentary: GE’s Calculated Bet on Analytics
GE’S BIG BET ON DATA AND ANALYTICS • MIT SLOAN MANAGEMENT REVIEW 3
GE’s Big Bet on Data and Analytics
I n September 2015, multinational conglomerate General Electric (GE) launched an ad campaign featuring a recent college graduate, Owen, excitedly breaking the news to his parents and friends that he has just landed a computer programming job — with GE. Owen tries to tell them that he will be writing code to help machines communicate, but they’re puzzled; after all, GE isn’t exactly known for its software. In one ad, his friends feign excitement, while in another, his father implies Owen may not be macho enough to work
at the storied industrial manufacturing company.
The campaign was designed to recruit Millennials to join GE as Industrial Internet developers and remind them — using GE’s new watchwords, “The digital company. That’s also an industrial company.” — of GE’s massive digital transformation effort. GE has bet big on the Industrial Internet — the convergence of industrial machines, data, and the Internet (also referred to as the Internet of Things) — committing $1 billion to put sensors on gas turbines, jet engines, and other machines; connect them to the cloud; and analyze the resulting flow of data to identify ways to improve machine productivity and reliability. “One billion dollars represents a big swing for GE,” says Matthias Heilmann, CEO of GE Oil & Gas Digital Solutions business. “It signals this is real, this is our future.”
While many software companies like SAP, Oracle, and Microsoft have traditionally been focused on providing technology for the back office, GE is leading the development of a new breed of operational technology (OT) that literally sits on top of industrial machinery. Long known as the
C A S E S T U D Y
If software experts truly knew what Jeff Immelt and GE Digital were doing, there’s no other software company on the planet where they would rather be.
–Bill Ruh, CEO of GE Digital and CDO for GE
4 MIT SLOAN MANAGEMENT REVIEW • GE’S BIG BET ON DATA AND ANALYTICS
C A S E S T U D Y G E ’ S B I G B E T O N D A T A A N D A N A L Y T I C S
technology that controls and monitors machines, OT now goes beyond these functions by connecting machines via the cloud and using data analytics to help predict breakdowns and assess the machines’ overall health. GE executives say they are redefining industrial automation by extracting lessons from the IT revolution and customizing them for rugged heavy-industrial environments.
One such environment is the oil and gas industry, where GE sees a $1 billion opportunity for its OT software. In an industry where a single unproductive day on a platform can cost a liquified natural gas (LNG) facility as much as $25 million, the holy grail for oil and gas is minimizing “unplanned downtime” — time that equipment is unable to operate due to a malfunction. Ashley Haynes-Gaspar, software and services general manager at GE Oil & Gas, notes that refining operations are typically tightly run but are in hard-to-access, remote locations. Increasing uptime is critical — particularly with oil prices at their lowest in six years. “An average midsize LNG facility sees five down days a year. That’s $125 million to $150 million. For an offshore platform, it can be $7 million per day, including oil deferrals, and these assets are never down for a single day. They have got to figure out how to drive productivity in their existing assets,” she says, “especially now that they are facing declining revenues from lower energy prices.”
Improving the productivity of existing assets by even a single percentage point can generate significant benefits in the oil and gas sector (and in other sectors). “The average recovery rate of an oil well is 35%, meaning 65% of a well’s potential draw is left in the earth because available technology makes it too expensive,” explains Haynes-Gaspar. “If we can help raise that 35% to 36%, the world’s output will increase by 80 billion barrels — the equivalent of three years of global supply. The economic implications are huge.”
GE executives believe software, data, and analytics will be central to the company’s ability to differen- tiate itself within the oil and gas industry. “I think the race is on from a competition perspective,” says Haynes-Gaspar, “and everybody understands the size of the Industrial Internet prize.”
In September 2015, GE projected its revenue from software products would reach $15 billion by 2020 — three times its 2015 bookings. While software sales today are derived largely from traditional measurement and control offerings, GE expects that by 2020, most software revenue will come from its Predix1 software, a cloud-based platform for creating Industrial Internet applications.
GE has long had the ability to collect machine data: Sensors have been riding on GE machines for years. But these pre-Internet of Things (IoT) sensors were used to conduct real-time operational performance monitoring, such as displaying a pressure reading on a machine, not to collect data. Indeed, a technician would often take a reading from a machine to check its performance and then discard the data.
GE researched companies that were producing high- quality data analytics quickly and inexpensively,
Since the early 2000s, British oil and gas company BP plc has used its own software to monitor conditions in its oil wells. Recently, how- ever, BP decided it didn’t want to be in the software business. GE, on the other hand, does. In July 2015, the two companies announced that by the end of the year, BP would outfit 650 of its thousands of oil wells with GE sensors as part of a pilot to test Predix. Each well will be outfitted with 20 to 30 sensors to measure pressure, temperature, and the like and will transmit 500,000 data points to the Predix cloud every 15 seconds. BP hopes to use the data to predict well flows and the useful life of each well and ultimately to gain a fleetwide perspec- tive on its oil fields’ performance.i
PREDIX AT BP
The Software Behind GE’s Industrial Internet
GE’S BIG BET ON DATA AND ANALYTICS • MIT SLOAN MANAGEMENT REVIEW 5
but it wasn’t the traditional IT companies that were excelling; it was the consumer-facing Internet giants. GE drew lessons from these companies around speed and cost, though the scale and data were different. Indeed, the sheer volume of data that GE hoped to collect — 50 million data variables from 10 million sensors installed on its machines — would be many times more than most social and retail sites could ever generate. “Machines generate time-series data, which is very different than social or transactional data. We had to optimize for the kinds of analytics that would help us understand the behavior of machines,” says Bill Ruh, GE’s chief digital officer.
To handle these massive data sets, GE needed a new platform for connecting, securing, and analyzing data. They began developing their solution in 2012, a cloud-based software platform named Predix that could provide machine operators and maintenance engineers with real-time information to schedule maintenance checks, improve machine efficiency, and reduce downtime. Initially developed for GE, not only would this data inform their own product development activities, but it would also lower costs in its service agreements. “When we agree to provide service for a customer’s machine, it often comes with a performance guarantee,” explains Kate Johnson, vice president and chief commercial officer, GE Digital. “Proactive identification of potential issues that also take the cost out of shop visits helps the customer and helps GE.”
It didn’t take long for GE engineers to realize that they could find interesting and unique patterns in the data. They thought the patterns of sensor data could be used to provide an early — albeit weak — signal of future performance problems and better predict when its machines should be scheduled for maintenance. In early 2013, GE began to use Predix to analyze data across its fleet of machines. By analyzing what differentiated one machine’s performance from another — what made one more efficient, for example — GE could more tightly hone its operational parameters. “We’re moving from physics-based modeling, where you create maintenance manuals based on generic operating
models, to combining it with very high-performance analytics,” says Ruh. When GE combined the physics modeling and the data modeling, it found that, in Ruh’s words, it could “do what no one’s ever done in the world before for industry.”
For example, in the last few years, GE started to notice that some of its jet aircraft engines were beginning to require more frequent unscheduled maintenance. “If you only look at an engine’s operating parameters, it just tells you there’s a problem,” says Ruh. But by pulling in massive amounts of data and using fleet analytics, GE was able to cluster engine data by operating environment. The company learned that the hot and harsh environments in places like the Middle East and China clogged engines, causing them to heat up and lose efficiency, thus driving the need for more maintenance. GE learned that if it washed the engines more frequently, they stayed much healthier. “We’re increasing the lifetime of the engine, which now requires less maintenance, and we think we can save a customer an average of $7 million of jet airplane fuel annually because the engine’s more efficient,” Ruh explains. “And all of that was done because we could use data across every GE engine across the world and cluster fleet data.” Johnson credits Predix directly with improving the productivity of these engines, as this would not have been possible without a robust data and analytics platform.
That same year, GE executives began to think there could be a market opportunity for Predix, much as Amazon.com Inc. created a market for its cloud- computing platform, Amazon Web Services Inc. “We realized that there were three developing markets for cloud platforms — consumer, enterprise, and industrial. Industrial was essentially being treated as an extension of enterprise, which we knew wouldn’t work. There were no credible cloud-based platforms for industrial being developed, and we saw that as a potential opportunity for growth,” says Ruh. Why now? GE executives say the economics of amassing, storing, and running analytics on large lakes of data — pools of customer data that combine maintenance and repair data with time-series performance information — have dropped dramatically in the last 10 years, making the market viable.
6 MIT SLOAN MANAGEMENT REVIEW • GE’S BIG BET ON DATA AND ANALYTICS
C A S E S T U D Y G E ’ S B I G B E T O N D A T A A N D A N A L Y T I C S
The driving force behind taking Predix to market was the scope of the opportunity: GE determined that the market for a platform and applications in the industrial segment could reach $225 billion by 2020.2
GE spent a year evaluating the market, all the while using Predix 1.0 to further develop its offerings and collect internal and external feedback. The company built a team to develop the commercial version, Predix 2.0, and in October 2015 made the platform directly available to channel and technology partners as well as customers who could use the platform to build their own set of analytics. “We feel we’ve got it right for ourselves, and now we’re taking it out to customers and partners,” says Ruh.
GE has enjoyed success as a physical infrastructure provider, known worldwide as the company that “brings good things to life.” But that corporate iden- tity is beginning to shift. “When we think about the future, digitizing our customers’ businesses requires a technology shift, a business model shift, and a skill shift,” says Ruh, noting that GE and its custom-
Two million miles of transmission pipe span the world, moving liquid oil or gas from its point of extraction to refining, processing, or market. In the U.S., 55% of that pipeline was installed before 1970. Pipeline spills, though rare, cause significant economic and environmental damage — as well as nightly news visi- bility — and pipeline operators are struggling to figure out where their next rupture is going to be. “Any given pipe- line may have 500 pinhole leaks, but the question is, which leak is going to be- come a big issue?” says Haynes-Gaspar.
Pipeline operators don’t have a reliable way to gauge pipeline integrity. “They have multiple sources of data but no way to integrate it into one location so they can see and understand the risk in their pipeline,” Haynes-Gaspar says.
In 2013, GE researched the challenges facing oil and gas industry verticals and identified pipeline risk management as a significant market opportunity. GE subsequently began developing a
pipeline-management software suite to provide operators with a single interface for accessing, managing, and integrating critical data for the safe management of pipelines. The first application in this suite was a risk assessment tool to monitor aging infrastructure.
In July 2015, GE deployed the new risk- assessment solution, which combines internal and external factors to provide an accurate, real-time, visual representation of where risk exists in the pipeline. “Pipeline operators previously have not had access to an integrated risk-management tool that considers the impact of external risk factors, such as flooding. Using the risk assessment tool, operators can see how recent events impact their risk and make real-time decisions on where to deploy field service crews first to assess impacted areas along the pipeline,” says Haynes-Gaspar.
The risk-assessment tool, with visualization and analytics running on
Predix, has a straightforward interface based on familiar consumer applications. “We are taking some consumer-facing design methodologies and applying them to the industry, because that’s how people expect to interact,” explains Brad Smith, business leader for the GE Oil & Gas Intelligent Pipeline business unit, who says GE used Google Maps APIs (application program interface) within the application. “We need to make it very easy to use, and that’s the only way we’re going to get the organizational adoption to enable the knowledge transfer of domain expertise that’s leaving the industry,” says Smith.
GE is also pulling data from weather systems and dig-reporting services to give a more comprehensive view of the pipeline network. “We’re bringing dif- ferent data sets together and enabling information sharing across organiza- tions to help drive better decision making and management for safer out- comes,” says Haynes-Gaspar.
A New Approach to Oil and Gas
GE’S BIG BET ON DATA AND ANALYTICS • MIT SLOAN MANAGEMENT REVIEW 7
ers will need to optimize all three to be successful. Indeed, in the oil and gas sector, where customers have been struggling to improve productivity amid declining revenues, GE is using Predix to transform what it’s selling, how it’s selling, and who is invited to the negotiating table.
Delivering New Services to a Conservative Market
GE entered the oil and gas industry in 1994 through the acquisition of Italy-based Nuovo Pignone, a manufacturer of turbo machinery, compressors, pumps, static equipment, and metering systems. Just over 20 years later, GE’s oil and gas subsidiary has
become a roughly $20 billion business, ranging from oil and gas drilling equipment and subsea systems to turbo machinery solutions and downstream pro- cessing. The company considers itself a “full-stream” provider, operating across the entire oil and gas value chain: upstream exploration and production, midstream transportation (via pipeline, oil tanker, or the like) and storage of crude petroleum products, and downstream refining of petroleum crude oil and processing of natural gas.
GE’s customers in the oil and gas market are known for being conservative, a necessity in the highly dynamic and sometimes harsh environments in which they work. “They wait for their peers to try
For instance, the potential impact of weather on risk management can be significant for operators with pipeline in areas prone to seismic activity, water- ways, and wash outs. “Checking weather patterns along thousands of miles of pipe for rain or flood zones, and integrating that with other complex pipeline data sets, can be a challenging, manual exercise,” said Haynes-Gaspar. “What we’re doing is putting all the data in one place, so operators have easier access to information and so they can prioritize and address areas with the greatest potential impact.”
Other data sources include seismic, GIS, and geospatial information (i.e., proximity of the pipeline to high- consequence areas like hospitals, schools, and transit depots) and inspec- tion data that identifies structural weaknesses. This data gets repre- sented through visualizations with calculated insights for enterprise risk using current operator programs and accepted engineering practices. Spe-
cific factors of active risk areas are taken into consideration, highlighting recent operational changes over the network. Missing or incomplete data is identified and represented with color coding on a dashboard display.
GE expects a customer to initially bene- fit from having all of its data integrated. The next step is moving from looking at the active, current risk to a what-if cal- culation tool that will allow pipeline operators to run hypothetical scenarios; for example, if they want to adjust op- erating pressures or address particular areas of corrosive pipe. The visualiza- tion provides a color-coded “view” of how those actions affect their pipeline risk. “This will enable users to examine data in different ways, quickly locate areas of interest, and evaluate threats or remediation measures, allowing for data-driven decisions [and] safer deci- sion making,” explains Smith.
The risk-assessment functionality also will help operators address growing
concerns from the public and regula- tory agencies. Current and pending regulations include process, documen- tation, data management and integration, risk, communication, and continuous improvement regulations. “Our management tool will make the path to compliance and safety manage- ment more efficient and easier,” says Haynes-Gaspar.
“We think these capabilities are going to change the nature of safety and pro- ductivity on these pipelines,” says Ruh.
GE went live with its risk-assessment solution in July 2015, with plans to tar- get just about every major natural gas and liquid pipeline operator in North America and globally.
8 MIT SLOAN MANAGEMENT REVIEW • GE’S BIG BET ON DATA AND ANALYTICS
C A S E S T U D Y G E ’ S B I G B E T O N D A T A A N D A N A L Y T I C S
something new, and if it’s successful and they see there’s value in it without added risk, they jump all over it,” says Steve Schmid, GE senior product manager. As a result, GE runs pilots — lots of pilots. “I tell customers about a new product and a pilot we’re doing with a highly respected operator in their industry. They say, ‘Fantastic, keep us up to speed on your progress. Once you’ve got a product released, we want to know the results, and then we’ll be interested in entertaining moving forward with a proposal.’ But it’s difficult to take that first leap,” says Schmid. “However, those that do so are the first to gain an incredible competitive advantage in the market — and others are soon a fast follow.”
Jeff Monk, GE’s global key account executive for North America, adds: “In general, oil and gas com- panies like to keep things close to their chests. They don’t see a lot of value in publicizing the details of their operations — whether those be big wins or big saves. If a customer can save $100 million as a result of data analytics, that’s great, but they will be concerned about publicizing that because they be- lieve stakeholders will think they might not have avoided those costs. Oil and gas companies have to come to terms with the market’s need for them to be more transparent, and the added value that data and analytics provide is relatively new, so that issue is somewhat mitigated.”
New Service Value Propositions
GE believes Predix can help the oil and gas industry address four of its most pressing challenges: improving asset productivity; creating a real-time picture of the status of an entire operation; stemming the costly loss of tacit knowledge from an aging workforce; and building an Industrial Internet platform that meets customer needs.
GE had spent years developing analytic applications to improve the productivity and reliability of its own equipment, with oversight from GE global monitoring centers. GE’s strategy is to deploy these
solutions and then expand to include non-GE plant equipment as part of the solution. GE’s work with RasGas Company Limited, one of the world’s foremost integrated LNG enterprises, is an example of that approach. RasGas’s LNG production facility in Ras Laffan, Qatar — where over 2,000 critical assets are installed — has the capacity to produce approximately 37 million tons of LNG per year.
GE’s equipment specialists and analytic scientists, who were already monitoring the GE turbine com- ponents of the RasGas production power trains, worked closely with RasGas operations experts to enhance LNG productivity overall, including the non-GE components. This close collaboration identified critical components, failure modes, and process challenges.
In an initial proof of concept, the team focused on three LNG trains. Their primary goal was to demonstrate that a suite of next-generation predictive analytics tools could enhance asset reliability and maintenance effectiveness while optimizing processes. In early results of the asset performance management (APM) solution implementation, the team identified areas of improvement to eliminate wastes in production on one of the trains, which will translate to a significant LNG production improvement. Today, GE and RasGas are working in conjunction toward a full-plant APM solution deployment at the Ras Laffan site.
GE wants to go beyond helping its customers manage the performance of individual GE machines to managing the data on all of the machines in a customer’s entire operation. For example, if an oil and gas customer has a problem with a turbo compressor, a heat exchanger upstream from that compressor may be the original cause of the problem. Analyzing data from the turbo compressor will thus only tell part of the story. “We’re selling equipment that sits side by side with competitors’ equipment. Our customer cares about running the whole plant, not just our turbine,” says Johnson. GE is in discussions with some customers about
GE’S BIG BET ON DATA AND ANALYTICS • MIT SLOAN MANAGEMENT REVIEW 9
managing sensor data from all of the machine assets in their operation.
Customers are asking GE to analyze non-GE equip- ment because those machines comprise about 80% of the equipment in their facilities. “They want GE to help keep the whole plant running. They’re saying, ‘It does me no good if my $10 million gas turbine runs at 98% reliability if I have a valve that fails and shuts my entire plant down,’” explains Dan Brennan, executive director for the Industrial In- ternet for GE Oil & Gas. Indeed, capital-intensive equipment such as gas turbines are already well- instrumented with sensors and data controls. “They’re well-protected to make sure they’re reli- able,” he says. The supporting, smaller-investment equipment often did not warrant the cost of in- strumentation and data gathering. But industry thinking has evolved as the cost of getting data from those less expensive assets has declined logarithmi- cally, opening up a whole new world of monitoring by looking at the system not as a collection of criti- cal equipment but as an ecosystem. For one pilot program with a major customer, GE is analyzing data on all their rotating and static equipment — regardless of the machines’ original equipment manufacturer (OEM). The first phase is on all 160 of the customer’s gas turbines across the world, even though only 40% of them are GE gas turbines. GE characterizes it as an “agnostic” solution.
How do the OEMs feel about having their equipment monitored by GE? Erik Lindhjem, executive product line leader for GE Oil & Gas, acknowledges that some competitors are more comfortable with the idea than others. “I think there’s an uneasiness for some of the GE OEM competitors about how the data’s collected and our reach into it. However, ultimately it’s the end user — our mutual customer — who drives the use of the technology and the data,” says Lindhjem. GE also points out that its competitors are quietly exploring the same strategy, albeit not by building it themselves but by partnering with other software providers. Siemens has entered into a partnership with SAP, while Solar Turbines, a Caterpillar company, has partnered with a startup, Uptake, to try to equalize the value proposition that GE is bringing to market.
Because Predix is an open platform, GE Oil & Gas Digital Solutions Business CEO Heilmann emphasizes that GE encourages developers to write applications to support their own needs. “I can see a day in the future where we would encourage other OEMs, whether it’s a pump or a fan or a valve manufacturer, to participate in the Predix ecosystem and write applications for their own equipment and use the data to improve the operation of their own equipment,” he says.
In today’s oil and gas industry, operators seldom share data and collaborate with one another. “Operators view that data as a source of competitive advantage,” says Brennan. Schmid agrees, noting that oil and gas operators differentiate themselves by the way they operate: “They can buy the same tools, the same type of equipment. And they have very few choices in suppliers. We’re not at a point where they’re going to start sharing how their operational excellence gets them a lower cost per barrel of oil than their competitors. Today, the only cross- operator data they are interested in is baselining against their competitors on how they perform, but not sharing how they actually drive up their efficiency over their competitors.”
Brennan says oil and gas companies are unlikely to ever share data around their exploration activities, but he could imagine a day when they might be willing to share operational data. “Maybe five or six years from now, we’ll begin to see companies more …