This post is a contribution from Cheryl Wiebe, Partner of Advanced Analytics at Teradata West Region.
If you follow Gartner, Forrester and other industry analysts, you’re probably a reader of Gartner’s ‘Hype Cycles’ and Forrester’s ‘Waves’. According to Gartner, IoT is at the top of the Hype Cycle, and is forecast to deliver economic value of £1.2 trillion by 2020 ($1.7 trillion USD, for those of you who like to cover all the currency bases). Thus it should be obvious to all of us that the Internet of Things has taken the place of Big Data in the Hype Cycle.
I’d like to go further, using a few examples from industry, to discuss how important IoT really is, and how operational systems in the IoT are informing what I call the ‘Analytics of Things’. This trend is creating an emerging business model, which will lower the cost and risk of the IoT business model.
First, let’s level set. In my view, the Internet of Things is an expansion of the notion of the Internet of People –the Web – the Internet.
People were the first beneficiaries of the Internet. The first real use cases were the creation of communities, content sharing, and of course ecommerce. We also created lots of data with cat videos and baby pictures, emails and social posts.
Now, as we head into the 3rd age, the digital age. Millions of devices and machines are connected, and these things, and the tiny chips embedded in them, are able to communicate their status with each other, and with headquarters – whether that’s a central machine, a cloud computing instance, or a corporate headquarters IT department.
They do this through a series of aggregation points or networking devices called gateways, which aggregate and package up data to send onward. The networks themselves convey this sensor data, either to public clouds or private data centers, to enable people (and machines) to do analytics on the machine and sensor data.
What really worries companies is the prospect that this growing volume of data is MUCH bigger than when it was just the Internet of People and their devices. Sensor data is probably a couple of orders of magnitude bigger than web log data and consumer interaction data. The data produced by the IoT is doubling every two years, and this creates fear among businesses trying to budget to manage that data – storage, analytics, compute resources, and the complexity and execution and opportunity costs of that data – and the risks of what if I don’t get started fast enough? What kind of analytics will be done? At what speed and where will it be done?
A distinction I want to make is that there are really two subsystems here. The first is where the operations among these things occur and where the thinking about these operations is done. The ‘operations of things’ in the operational network where the devices live, where the gateways live at the edge – that’s where the actions and decisions are made that inform what businesses do with all that data. The second subsystem is where the thinking about these things and the analytics is done. If you, as I do, believe that thinking should precede acting, that’s a simple way of thinking about how analytics needs to power the operations of things. To be complete, there is also a third domain, where we close the loop and take action, orchestrate and execute those actions back into the operations of things.
To be ready for the Internet of Things and the Analytics of Things, we have to start doing a few things differently. We have to retain architectural flexibility. We may have to decide where and when to compress sensor data before passing it along to an aggregation point and upwards to the cloud or headquarters. We have to decide: will we store all this data in the cloud? Will we pass all the raw data to the cloud? It’s hard to say now, and it will depend somewhat on the operational characteristics (speeds and feeds) of each use case, but it’s important that we understand and make architectural decisions that don’t close the door on some of those options.
Some of the machine and sensor data is very redundant, some of it is clearly erroneous, and some of it indicates there’s a problem in the actual sensor device: a battery problem or an energy spike. Also, we’re converting the analog world to the electronic or digital world here so we may need some intelligent processing at the edge. It’s a very controversial issue as to how much intelligence will actually get pushed down from the headquarters to the edge, because these devices are increasingly intelligent on their own. Some are also talking about moving the data beyond the edge but not quite to the cloud – what we call ‘fog computing’.
On the internal user side, the marriage of machine data with important information from a company’s corporate data assets, such as order data, machine maintenance records, operator training data, and environmental data is what creates value and the ability to create complex predictive models. Having the ability to react in real time to this data is important; this is one space where VoltDB’s very fast SQL operational database operates.
On an externally focused business model, you can see how a manufacturer’s ability to collect its own sensor data, potentially multi-channel sensor data from its asset, like a truck, that’s operating in the customer’s context (their fleet, their construction/mining/drilling site, their factory), allows them to better underwrite the risk of failure of that asset, and potentially take on the ownership of the asset, essentially outsourcing and selling information and uptime rather than selling a capital asset. That’s a new business model that externally focused companies are after.
Regardless of where you stand – internally focused as a company on your own operations, or externally focused on how your product is operating in the field for your customers – data from the industrial Internet, from all those sensors and machines, is key to improving machine up time, improving yields, improving field service outcomes, improving the ability to manage risk – the risk of a time variability, the risk of orders not being met, and the risk of poor quality.
If you are on the PROVIDER side: provider of internet–connected machines and devices such as large equipment, auto, aerospace, large process automation equipment – your goal for IoT is to improve your product so that when those connected things are integrated and operated by the end-user, they operate more effectively. You can then provide a more complete service, predict the behavior of that device and can make sure it’s performing properly. You can lower the cost of warranty and create new revenue streams for those customers. Instead of new capital asset sales, you might be renting the product and providing a service based on number of hours of usage. The goal is to create new revenue streams and improved user experiences for your products.
On the other hand – those on the USER side: this is going to be probably the vast number of companies getting into IoT. They are the owners of campus’, factories, hospitals, oil exploration or mining sites, etc.; they are the buyers of those internet-connected machines and devices. They have a different type of task; integrating all of those different types of technologies into one coherent set of signals they can understand and operate effectively and efficiently. What we find is today these companies are really wondering how to get started with the IoT. That’s because their task can be a bewildering mixture of technologies and products from many vendors, with no clear standards for blending.
The industrial IoT is enabling our organizations to improve their operations or their products. It’s giving people and data analysts more real-time data on connected machines and devices, enabling them to use that data in real-time to make decisions and take actions to make the data more valuable, and to make the world a better, safer, more aware place. It’s where the value of all those machines and devices, connected for decades, taking to each other – will be most visible to us, and most beneficial to our companies.
Every industry will have many, many opportunities to capitalize on IoT’s new capabilities and new sources of data, whether it’s from smart meters to smart agriculture, or supply chain industries that are popping out that span traditional verticals. Analytics will fuel every one of these opportunities. Analytics is the heart of these opportunities and without analytics built on access to all of your data, including the sensor data from IoT, some of opportunities may pass you by.
It’s important that we all start understanding the use cases and the business cases, and perhaps most importantly, what are the business models we want to pursue?
To hear Cheryl speak with VoltDB’s Dennis Duckworth about the most impactful IoT use cases and business models, and key technology considerations to ensure effective analytics against IoT data, download their webinar today.