A Year in the Life of the Internet of Things
I have 149 connected devices, using 17 different systems, and at least five wireless networks in my home. I can monitor and control lighting, energy consumption, weather, irrigation, exercise (me and our dogs), driving habits, doors and locks, safety, security, music, automobiles, and sleep.
Why? Because I decided that 2015 was to be the year of the Internet of Things (IoT). I already had DropCams, a Fitbit, some Philips Hue lights, internet-connected solar panels, and some internet-connected thermometers. I was just scratching the surface. Last year I decided to jump in with both feet and see how far I could get using my home, myself, and my ever-suffering spouse as the test subjects.
Beyond the personal interest, I also have some professional reasons at stake. I’m Jon Pittman,the vice president of corporate strategy at Autodesk, and part of my job is to watch for and interpret trends affecting our customers and our company. We serve people who design and make the world’s places, things, and media. Sensors and connected devices embedded in our buildings and products impact how things are designed and made. As places and things become more connected with embedded computation, media may well play a role in helping us interact with them, as well.
Second, I was fascinated to see just how far I might be able to take it; the geek in me was curious. Lastly, I wanted to minimize energy and water use. It was difficult to even know what I was using much less control it.
One year and three months in, I’ve learned a few things:
Many consumer IoT products are incomplete. Consumer IoT is hot right now, and a plethora of startups are targeting consumers by rushing devices to market. Unfortunately, some of these devices are ill-conceived, and many are just not ready for prime time. The dominant methodology in product development right now is something called Minimum Viable Product (MVP). Minimum Viable Product is the idea that a developer should target the most minimal product possible to ensure that they are actually meeting a customer need before fleshing out the product. The intent is laudable but many startups mis-apply MVP to create minimum but not necessarily viable products. The effect is to push out products that need to be debugged and completed — unfortunately often at the expense of the consumers time.
An IoT system is only useful if it performs consistently and reliably.
As an example, many IoT systems use a hub to connect and control the devices. The particular system I use is largely cloud-based. The hub locally monitors devices for events, sends those events to the cloud, and initiates actions in response to events. Events can be things like doors opening or closing; a person or vehicle arriving or leaving; a water leak; temperature reaching a specified threshold; a particular time; or the sun rising and setting. Actions are things like turning lights on or off; unlocking or locking doors; or sending a notification to my smartphone.
All these events and actions are managed by a cloud server. Unfortunately, the company that makes my hub does not have very robust server software. It sometimes randomly misses events and fails to trigger actions. This means that doors don’t lock or lights don’t go off when they are supposed to — thus the system cannot be trusted to perform.
My wife has spoken about this to a builder friend who does high-end homes, and he will not use any of the consumer IoT systems out there, instead using robust, proven home automation systems — for this very reason. Failing to turn lights off or lock doors may seem like a minor issue to the IoT developer, but, in fact, are mission-critical to the goals of home automation users.
If anecdotal evidence from a high-end home builder isn’t enough, consider the data from the Consumer Electronics Association, which tracks the return rate of consumer electronics. In 2014, 22% of buyers returned some type of consumer electronics device or accessory. The most common reason for return is that the device did not work as expected. This number is higher for products in early stages of their lifecycle (e.g. IoT devices) and lower for mature products (e.g. flat-screen TVs).
An IoT system is only useful if it performs consistently and reliably. What my system developer calls “platform stability” is not really an option, it is a necessity. Without it, the I cannot trust the system. Yet the developer treats is as if it is optional.
The neediness of my devices takes my time and money — neither things that I have in abundance — and is antithesis of why I purchased these products in the first place.
Things are often needy. I expect the things in my system to be self-sufficient. Unfortunately, many are not. I have 149 connected devices, which means on any given weekend I end up with a list of devices that need attention. Some are relatively simple fixes such as replacing batteries. Other tasks are more complicated and annoying, networking failures and failed software updates.
The batteries are relatively straightforward fixes, except for two things: each device seems to use different batteries, and it is not always clear when batteries are dead. Sometimes the devices just stop working or function abnormally and in ways that are difficult to detect. How about a clear and simple message saying the batteries need to be replaced and a standardization on battery type?
Networking problems are more pernicious. They are hard to diagnose and hard to fix. Some of my devices are really networking robust — they seem to maintain a network connection no matter what. Others are finicky. Fixing network connectivity issues is a black art and — again — each device has its own set of processes and quirks. It’s not as easy as just unplugging it for 10 seconds and hoping all is well when you plug it back in.
Software updates are insidious, too. Sometimes they fail — at best requiring human intervention (which I regard as a waste of my time) or worse, “bricking” the device, requiring me to send it back to the manufacturer for replacement. Worse, do I have partially successful software updates which leave my things in a weird limbo state? This reduces trust in the system.
Ultimately, the neediness of my devices takes my time and money — neither things that I have in abundance — and is antithesis of why I purchased these products in the first place.
IoT creates a complex ripple effect. My IoT system is a complex network of device and connections. My house is a network of networks; a complex tangle of devices and connections all competing for the wireless networks I’ve set up. I have at least five wireless networks in my house. I say “at least” because sometimes it is unclear how devices communicate with me and with each other. Given this murkiness, sometimes things happen and it is not clear why.
This Christmas, for example, I added a Sonos™ wireless music system to my house. I connected the Sonos system to my hub (an allegedly supported function) and suddenly my hub began losing network connectivity every few hours. I figured it was some kind of obscure radio interference problem coming from Sonos. It turns out it was a limitation in the hub programming and had nothing to do with radio signals. Who knew?
With such complexity, unintendend and unanticipated things happen creating a ripple effect. This phenomenon is well known among IoT early adopters. The blogs describe these as impediments to “spousal acceptance!”
The biggest challenge is in visualizing and controlling the IoT network. The biggest issue I have is understanding the network and the connections amongst all my devices. While each set has a controller — usually a web or smartphone app — there is nothing that lets me see and control the whole system. Each system focuses on its own devices and their state — and the connections are often tenuous, invisible, and undocumented.
I know how the system works because I assembled it and maintain it. Even so, I have had to build a large spreadsheet to keep track of all of the things and connections. In industrial IoT parlance, a “digital twin” is a digital representation of the physical thing that lets you understand and manipulate the physical thing. Ironically, I have had to build a digital twin of my IoT networks just so I can see and understand them. Duh!
After the initial delight and excitement, the reality of unfinished products, needy devices, and the ripple effect set in — leading to dismay, disgust, and resignation.
The Internet of Things is about networks. There is a need for something to visualize not just the things but the network connections and interactions between them. There may be the kernel of a product idea there.
IoT is as much about the connections and the data as the things themselves. Ultimately the things in the Internet of Things don’t do much by themselves. They sense something — such as presence, motion, temperature, open/closed states — or they do something — lock a door, turn on a light, send a notification. The interesting things happen when we connect the things. The connections are far more powerful than the things themselves, yet the developers focus almost exclusively on the things — not the connections.
Further, the data that the things produce is fascinating. I can look at a stream of data produced in my house — and the things are constantly chattering. What are they talking about? They are reporting temperature changes, humidity changes, motions detected, arrivals and departures, doors opening and closing, and a myriad of other things. Only a small portion of this data actually triggers an action. But it is constantly going on — I presume some of the people who make these things are capturing the data and will find a use for it. Clearly, there is very little cost to collect the data and the data must, at some point, have some utility. Ultimately, the Internet of Things need to focus less on the things and more on the “internet” part.
Why does this matter? It could be dismissed as the grumblings of an early adopter. Fair enough, but I think there are two important issues that it highlights. First — in consumer IoT, developers are adopting an industrial attitude — of delivering half-completed products and expecting customers to complete them. That technology approach harkens back to the 1980s — remember PCs of that era? Nowadays, consumer expectations are for things that just work — and work well. The industry may be retarding its growth by not accepting this state of affairs.
Accenture just released a report saying that estimates of consumers migration to IoT are overblown. They cite cost and concerns about security and privacy as impediments to consumer IoT adoption. They are probably right but I think an even bigger factor is value. While consumer IoT is costly, the real issue is that there is just not enough value for what a consumer has to pay. The tasks consumer IoT does are mostly convenience tasks. The ones that are more mission critical — like energy savings and security have to work. The kinds of system failures I describe above reduce or eliminate the utility and value of an IoT system. By failing to address robustness and reliability, consumer IoT manufacturers are limiting their market opportunity.
More importantly, consumer IoT is a harbinger of industrial IoT. Consumer IoT is estimated to be as great as a $1.7T business — but it pales in comparison to projected industrial and business uses of the Internet of Things, which might be as big as $10.4T. If developers tackle Industrial IoT with the same cavalier attitude they approach the consumer IoT with — we are destined to reach an IoT Apocalypse.