Formerly known as Wikibon
Close this search box.

Harvesting Value at the Edge

IoT, though a useful application of available technology, and well-defined at the hardware and network levels, the heart of IoT, that part that yields the real value, is edge analytics. Making sense of the sensor data before it comes home. We already know how to stream massive amounts of data into data lakes for later analysis, but analytics at the edge begs a different approach.

The Internet of Things (IoT) is a proliferating set of interconnected devices that have IP-addressable sensors embedded in or attached to them, providing those devices with capabilities to intelligently sense and respond to their surrounding environment. As deployment of IP-enabled devices proliferate, IoT will expand until it becomes a digital layer of web-connected objects around the globe. As this new web of objects radiates outward, organizations are faced with opportunities to create new value-add services.

The challenge facing any organization is how to tap these opportunities. Some are betting on placing and owning the devices and selling the data being generated (e.g., Planet Labs launching 100s of small satellites into orbit and selling planet imagery). Others are exploiting IoT to translate product sales into utilization services (e.g., GE Aviation’s use of aircraft engine telemetry to provide services to improve aircraft maintenance and fuel efficiency). Others are utilizing IoT technologies to sustain, extend, or modernize traditional SCADA-based infrastructure businesses. However, across the full range of IoT opportunities, three “things” are becoming clear:

  • It’s not about the hardware and architecture. The value in IoT is not in hardware or networks, it’s in analytics, what you learn. Moreover, the physical architecture is almost incidental. It doesn’t matter what devices you use, where the location intelligence resides, or if you use the cloud or not.
  • IoT changes operations, not just costs. At this point, it is pointless for any organization to invest in IoT unless their strategy is to change their business, not just take out cost.
  • Existing governance and security regimes will be inadequate. <blah>

It’s Not About the Hardware and Architecture

There are dozens of reference architectures for IoT, and they do matter in terms of cost, complexity, performance, and fitness to work. However, we see too many in the community falling into the trap that captures so many during industry transformations: putting the technology first. IoT is a marketer’s dream – but potentially an implementer’s nightmare. Why? Because it is so big, so broad, so potentially transformative that any product, service, or company can stake a position in terms of IoT. The effort to implement IoT, on the other hand, requires clear objectives, careful planning, and disciplined implementation.

IoT decisions will touch upon a broad range of strategy, people, process, technology, and investment issues. Different industries and applications will feature different specializations and priorities. But one thing that will be common across all IoT applications: analytics will play a crucial role in operating and driving derivative value from IoT resources. As a result, your IoT architecture will be strongly influenced by choices regarding where and how to process edge analytics.

For example, commercial jets are loaded with sensors and the volume of their telemetry is huge. The engines alone generate 10 terabytes of data every 30 minutes. A cross-country flight would generate 240 terabytes of data. At roughly 87,000 commercial flights per day in US, that would generate a staggering 10 million terabytes of data per day. Assuming everything proceeds normally, it is not economical to store all this data. Successfully applying IoT in this way – and, we believe, all ways – requires decisions regarding what to do with raw data from the sensors.

These choices will be shaped by questions of data movement costs (including latency and bandwidth) and data security/governance concerns (see Figure 1). Giving devices control of the type and volumes of data to transmit requires hard choices such as:

  • Sending filtered data only if it is required.
  • Only transmitting abnormal values.
  • Compressing the data without using unfounded/untested algorithms.

Figure 1. Edge Data Is Not Like Other Data

IoT Changes Operations, Not Just Costs

IoT analytics will generate great benefits to organizations, users, even mankind, but don’t get into edge analytics unless you are ready to change the way an operation works. Merely looking to reduce costs will fail because the cost and effort to install instrumentation, conducting telemetry, perform complex analytics processing, etc., likely will exceed any operational savings.

For example, the advent of consumer mobile devices has catalyzed enormous interest in using IoT to manage the growth of costs in healthcare. A cost-based focus suggests enormous returns: lower patient service costs, higher profits. However, data shows that most people lose interest in wearable devices with a month. Moreover, older people, who have the most health problems, have a difficult time mastering smart devices. Furthermore, medical providers can’t seem to get the streaming telemetry into their proprietary EHR system. The biggest hindrance of all is the FDA where putting a mobile app into the market using a smartphone can be determined to me a medical device, requiring hundreds of thousands (or even millions) of dollars of clinical trials (and years). It is gratifying to speculate how IoT will save enormous sums in healthcare, but reality intrudes.

Another seemingly good example for IoT is agriculture, but the details prove otherwise. A quick calculation is that there are 2.3 billion acres of arable land in the US. Some existing sensors are about $1,500 per acre. For a 200 acres’ field, the sensor cost alone would be $30,000. This would not include all of the other infrastructure needed for the sensors to transmit data, storage, and probably proprietary software to evaluate the data and drive a “smart” irrigation process.  That is just too pricey for most farmers. Currently, corn yields are about 170 bushels/acre at a price of $3. That’s about $102,000 for the 200 acres. The projected income per acre to raise corn, for land, fertilizer and fixed + variable equipment, is about $47.50/acre, or about $9,500 for 200 acres. In addition, farmers surveyed say they have less than 5% of each day to be concerned with data, and the sensor systems overwhelm then. With an average farm size in the US at around 500 acres, it’s plain to see that IoT will be a niche application except for very large farms.

Existing Governance and Security Regimes Will Be Inadequate

Because IoT ultimately changes operations, and not just makes existing operations less costly, IoT and edge analytics presents a whole new class of security and governance issues. Governance is not just a restraint on using data, a means for a committee to lock it down and reduce risk. Governance can be a foundation for delivering modern data functions, including:

  • Self-service access for more users, with varied interests.
  • Discovery of useful/usable data assets by those audiences.

Governance still must also support stewardship and the need for compliance and risk

mitigation has not gone away. Businesses should shift their focus of governance to include more enablement, to preserve and create value out of (disorganized) data from a wide variety of sources. This is different than the data warehouse era which was limited to predictable, defined, data coming in, with only limited/pre-defined access given to the organization.

Edge analytics should promote the free flow usage of data. The magnitude is forcing a rethink of governance. IoT also is growing alongside a more common abstraction of compute and storage, changing access patterns further, not a monolithic application, but many use cases at once on the same repository

Action Item. Do not start on IoT unless you plan on changing how something works – a new business for example. Give thought to analytics at the edge. These devices have very limited computing power but with small amounts of data and pre-built algorithms can be powerful and economical analytical engines. Machine Learning algorithm development is a slow and data-intensive problem, best suited for the cloud, but applied ML algorithms are compact and can function very well at the edge.

Article Categories

Book A Briefing

Fill out the form , and our team will be in touch shortly.
Skip to content