BT

Analytics, Machine Learning, and the Internet of Things

| Posted by Seth Earley Follow 0 Followers on Jun 07, 2015. Estimated reading time: 12 minutes |

This article first appeared in IEEE IT Professional magazine. IEEE IT Professional offers solid, peer-reviewed information about today's strategic technology issues. To meet the challenges of running reliable, flexible enterprises, IT managers and technical leads rely on IT Pro for state-of-the-art solutions.

 

Our increasingly connected world, combined with low-cost sensors and distributed intelligence, will have a transformative impact on industry, producing more data than humans will be able to process. Will businesses be able to adapt and evolve quickly enough to maintain their place in the competitive landscape? How will humans make sense of and benefit from these new sources of information and intelligence embedded in our environment?

Exploiting Evolving Technology

Organizations will need to get their internal data houses in order so they can leverage new sources and streams of data. Smart connected devices will also remove humans from the loop in some cases, so devices will be making their own decisions and self-adjusting or course correcting and repairing themselves as needed. In other cases, collections of devices will act as systems that can be optimized in new ways, and systems of systems will share data and behave as an ecosystem of data and devices. Machine learning - a term that describes numerous approaches to deriving meaning from data - will have to be part of the equation, but so will traditional business and data analysis techniques as organizations prepare for the Internet of Things (IoT).

The IoT, or as some prefer to call it, the ''Internet of Everything'', has been on an increasing growth trajectory that Gartner projects will reach 26 billion units by 2020, with the value of IoT products and services reaching US $300 billion1.GE, a long time player in the industrial Internet - which comprises the mechanisms and applications for monitoring and optimizing the performance of industrial equipment (including jet engines, locomotives, power turbines, and manufacturing processes) - estimates that the industrial Internet will add $10 to $15 trillion (yes, trillion) to the global gross domestic product over the next 20 years2.

Of course, there is an enormous amount of hype in the marketplace around new and emergingtechnologies. In fact, Gartner’s infamous ''hype cycle'' report has the IoT at the ''peak of inflated expectations'' (big data has already entered the ''trough of disillusionment''3).  Yet regardless of entrepreneurs' breathless excitement or journalists' enthusiastic vision of the future, there are a number of challenges that organizations must wrestle with to exploit this evolution in technology.

The Challenges

Organizations will need to focus on

  • understanding the relative maturity of enterprise capabilities in the realms of product technology and IT;
  • understanding the types of IoT functionality that can be incorporated and where new capabilities will impact customer value;
  • understanding the role of machine learning and predictive analytics models; and
  • rethinking business models and value chains based on how quickly the market is changing and the relative agility of competitors.

Let's consider each of these challenges in more detail.

Understanding Product and IT Maturity

This factor can be considered in two dimensions. How mature is the product portfolio? Is it a traditional class of product with slower changes and gradual evolution, or is it a faster moving, more complex technology ecosystem? Mining equipment is technologically complex but has longer equipment life cycles and relatively slower evolution than scientific research instrumentation. However, this doesn’t mean that the instrumentation firm is better equipped to extend its IoT offerings into system optimization. Another factor needs to be considered - that of IT process maturity. Each type of organization would benefit from IoT enablement; however, the models for that evolution will vary.

Consider the dimension of the level of IT maturity. For example, the scientific research equipment supplier might be technologically advanced but not have strong IT architectures, processes, and governance. The mining equipment manufacturer might be very mature in internal IT processes. The implication for the scientific instrumentation firm might be that IoT will allow for functionality updates of field instrumentation, but the firm might not want to attempt to optimize a laboratory information ecosystem consisting of multiple classes of equipment. (It is certainly possible that a lack of maturity in IT as a cost center wouldn’t translate into a lack of maturity of IT in a profit center; however, many organizations build on existing foundational IT capabilities when developing or extending IT services offerings.)The mining equipment example is discussed in a recent Harvard Business Review article on the IoT: Joy Global is a mining equipment manufacturer that offers monitoring, maintenance, and optimization of a fleet of equipment from multiple vendors by leveraging its expertise across various systems and processes related to mining operations4.

Understanding IoT Capabilities

The next idea to consider is what capabilities to leverage in smart connected products. According to the same Harvard Business Review article, there are four types of IoT capabilities4:

  • monitoring - sensors provide data about the operating environment and product usage and performance;
  • control - product functions can be controlled and personalized;
  • optimization - feedback loops from monitoring and control allow for improved efficiency, better performance, preventative maintenance, and diagnostics and repair; and
  • autonomy - monitoring, control, and optimization allow for independent operation, coordination with other systems, interaction with the environment, personalization, replenishment, and self-diagnosis and repair.

These levels of capability allow for redefined supply chains and reconfigured value chains. Rather than considering products as having fixed functionality, we need to view them as more flexible and adaptable. When products are intelligent and connected to the Internet, they become variable and have the ability to change as the user’s needs change. Software manufacturers have recognized this for years. Now, physical objects become vehicles or containers for softwaredriven functionality. These levels of capability require increasingly sophisticated data analytics approaches - from collecting and applying data to allowing algorithms themselves to apply data and learn while doing so.

So, the first level of capability - monitoring - becomes a real-time mechanism to better understand field performance and user needs and offer new capabilities. This means that the boundaries of an organization’s traditional products and services are blurred and extended. Consider field equipment that was traditionally maintained by a contract field service firm, not by the manufacturer. With intelligence and monitoring, equipment can inform the manufacturer of needed service ahead of a breakdown. Routine maintenance can become part of the manufacturer’s offering, with complex repairs still being handled by a specialist contractor if the margins and logistics make sense for the organization. This disintermediation can extend to distribution chains as well. Equipment can automatically call for a replenishment of supplies, removing distribution costs and inventory from the supply chain.

Control is a more sophisticated application built on top of Rathermonitoring. We can monitor equipment operations and then extend the boundaries of human intervention by controlling multiple pieces of equipment or multiple systems. Consider the role of humans in running a system or machine, where most of the functions are automated. Humans guide the operation and look for edge conditions, anomalies, and exceptions that weren’t anticipated (or cost-effective) during the system design. Then, they use their judgment to make a change, correction, or adjustment. The human doesn’t need to be with the equipment and might not need to be monitoring in real time (depending on the process). Monitoring is simply taking in the data and processing it (something must be done with the data at some point). Control is applying that data in real (or near real) time to the operation of the equipment or device. The strategic decision that organizations need to make is whether and when to make more control capabilities part of the product offering and whether to offer that as a service or to allow the customer to have that capability.

The third level of capability - optimization - can extend to the performance of an individual object, a fleet of objects, or an ecosystem of objects across multiple manufacturers and technologies. The strategic decision about whether to extend offerings to this realm hinges on the level of knowledge and sophistication around the value chain and the boundaries of the processes. The mining example illustrates the advantages that Joy Global might have over a vendor with a more limited view of the process ecosystem. A truck manufacturer, for example, would be poorly positioned to optimize complex mining equipment but would benefit from optimizing its fleet of trucks (and potentially a fleet of other manufacturer’s trucks) if the industry dynamics made business sense.

Extending optimization to independent operation requires an extension of capabilities to allow for less constrained interaction with the environment and with other systems. Autonomy requires greater intelligence around algorithms that can deal with unplanned situations - those situations for which programmers and system engineers didn’t explicitly design. Autonomous operations require incorporating adaptable machine learning approaches for dealing with novel situations into the core algorithms used for monitoring, control, and optimization.

Understanding Analytics and Machine Learning

In November 2014, Mike Kuniavsky of Xerox PARC gave an IDTechEx presentation, ''The User Experience of Predictive Analytics in the Internet of Things'', in which he suggested that virtually all functionality resides (or will soon reside) in the cloud. Data and functionality can be accessed from any location and through multiple devices. Specialized devices provide context in which the user accesses the data.

A fitness bracelet can access data about the user’s physical health via an iPhone or laptop in the specific context of exercise. In this case, the fitness bracelet acts as an IoT sensor as well as provides a means for accessing and consuming data. The device also subsumes other devices (such as a pedometer) through software functionality. The data provided by the device can offer additional insights about the consumer’s usage and preferences, which can be leveraged when updating functionality and developing new features. If aggregated across a population of users and combined with other datasets, new insights can shed light on epidemiological data, activity levels across populations, lifestyles, and demographic data. This information has value to marketers, healthcare providers, insurance companies, and government agencies. (Of course, we must account for privacy considerations and data usage permissions.)

Machine learning algorithms can be used to make predictions based on these data patterns. For example, in a Mayo Clinic study, activity data was correlated with recovery rates for cardiac patients5.

The same machine learning and predictive algorithms are the basis for a number of connected intelligent consumer devices. Nest thermostats are an example of a device that leverages data patterns to predict the preferred temperature in a specific room at a certain time of day. (Another control and optimization example is seen at an aggregated neighborhood level, here power utilities can shift energy loads at peak times by remotely adjusting - with the occupant’s permission - hundreds or thousands of Nest devices by a couple of degrees.) Other consumer devices include those that learn from voice patterns (such as Echo, a personal-assistanttype device from Amazon6) to those that learn from much more complex behavior and activity patterns (such as Jaguar’s Land Rover monitoring system, which ''relies on a complicated software which enables the car to study, predict, check, and remind the car’s occupants [to] help the driver auto-delegate his tasks and make him concentrate more on his driving.”7)

Optimization algorithms use machine learning mechanisms to leverage data from both sensors and intelligent devices that interact under dynamic conditions. These variable conditions can’t be precisely predicted beyond certain parameters. The algorithms will need to sense, respond, and adapt. For example, as cars take on more responsibilities from the driver, they will be interacting with more environmental sources of data (sensors, lights, other cars, and so on). Classes of applications in industrial automation, logistics and transportation, power grid and energy systems, traffic management, security systems, and other ''systems of systems'' will let machines communicate directly with other machines. Furthermore, such applications will help machines interpret dataflows based on algorithms that can evolve and adapt, so the machines can achieve the desired end states given certain operational parameters.

Rethinking the Business Model and Value Chain

Intelligent, connected devices require organizations to reexamine how and where they create value in the marketplace and how that value will be enhanced or diminished as the competitive environment and information ecosystem evolves. Analytics will help validate some decisions (for example, getting real-time usage data regarding changes to features or added services and functions); however, business models might be so vastly transformed by new entrants and value-chain structures that analytics based on the company’s traditional business models will no longer be relevant. Products or services might be based on data stream exhaust from legacy products rather than revenue from the products themselves. New business models might extend far beyond the product and into upstream suppliers or downstream consumers.

At the core, all of these possibilities require organizations to have foundational capabilities around their internal data hygiene and analytic infrastructure: data curation, ownership and quality standards, consistent enterprise architecture, cleanly integrated systems, automated data onboarding processes, and mature analytic expertise. Without the basics in place and well managed, it will be very difficult to rapidly react to and evolve new analytic and data management functions and abilities.

Because the IoT will be based on dataflows and sophisticated approaches for gaining insights from information and applying those insights to value creation through integration with enterprise knowledge, organizations that don’t have those abilities will be left behind in the marketplace or relegated to low-value, lowmargin commodities. Data has been called the new oil - and extending that metaphor means that data is refined into high-value products through the knowledge refinery of analytic capabilities. Organizations need to invest in building that infrastructure now so they are prepared in the coming years when supply chains and value creation will be transformed, disrupted, and upended. Information agility will be a required core competency.

References

1. ''Gartner Says the Internet of Things Installed Base Will Grow to 26 Billion Units By 2020'', Gartner, 12 Dec. 2013;
2. ''Analyze This: The Industrial Internet by the Numbers & Outcomes'', GE, 7 Oct. 2013;
3. ''Gartner’s 2014 Hype Cycle for Emerging Technologies Maps the Journey to Digital Business'', Gartner, 11 Aug. 2014;
4. M.E. Porter and J.E. Heppelmann, ''How Smart Connected Devices are Transforming Competition'', Harvard Business Rev., Nov. 2014, pp. 70–86.
5. D.J. Cook et al., ''Functional Recovery in the Elderly After Major Surgery: Assessment of Mobility Recovery Using Wireless Technology'', Annals of Thoracic Surgery, vol. 96, no. 3, 2013, pp. 1057–1061;
6. D. Etherington, ''Amazon Echo Is A $199 Connected Speaker Packing an Always-On Siri-Style Assistant'', Tech Crunch, 6 Nov. 2014;
7. M. Mendoza, ''Jaguar Land Rover Develops Self-Learning, Intelligent Car'', Tech Times, 17 July 2014;

About the Author

Seth Earley is CEO of Earley & Associates. He’s an expert in knowledge processes and customer experience management strategies. His interests include customer experience design, knowledge management, content management systems and strategy, and taxonomy development. Contact him at seth@earley.com.

This article first appeared in IEEE IT Professional magazine. IEEE IT Professional offers solid, peer-reviewed information about today's strategic technology issues. To meet the challenges of running reliable, flexible enterprises, IT managers and technical leads rely on IT Pro for state-of-the-art solutions.

Rate this Article

Adoption Stage
Style

Hello stranger!

You need to Register an InfoQ account or or login to post comments. But there's so much more behind being registered.

Get the most out of the InfoQ experience.

Tell us what you think

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Email me replies to any of my messages in this thread
Community comments

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Email me replies to any of my messages in this thread

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Email me replies to any of my messages in this thread

Discuss

Login to InfoQ to interact with what matters most to you.


Recover your password...

Follow

Follow your favorite topics and editors

Quick overview of most important highlights in the industry and on the site.

Like

More signal, less noise

Build your own feed by choosing topics you want to read about and editors you want to hear from.

Notifications

Stay up-to-date

Set up your notifications and don't miss out on content that matters to you

BT