Almost two years ago I wrote about the masterful re-branding exercise that we now call the Internet of Things (IoT). We used to call it Machine to Machine (M2M) communications. Fifteen years ago, everyone was very excited about M2M, a new day of connected vending machines, refrigerators and industrial sensors was just on the horizon. Unfortunately, it stayed on the horizon for that whole time. A few years ago, new players entered the market boosting new life to the marketing effort, and we now call the same idea IoT.

I point out all this history for what should be an obvious note of caution. The promise of connected devices has been with us for a long time. It is complicated stuff, touching on a dozen product areas and an even larger number of industries. I do not mean to imply that we should ignore IoT, but I do think it is important to understand that there is no magic technology or silver bullet that will make us wake up one morning and say IoT has arrived.

That being said, I do think we are seeing an important development in this field, and it was on display at CES this year. For the first time, we are starting to see real connected products. This is not so surprising, but there is now an implicit understanding that IoT will look very different in different industries. The idea that there will be one common operating system (OS) or platform underpinning IoT has moved off the radar. The auto industry has its range of LTE-enabled vehicles, with in-car Wi-Fi hotspots. The home remodeling market has several ecosystems now (more on this in a moment) which connect locks, thermostats and sprinkler system. The vending machine vendors have their own innovations. The health care industry is beginning to take active interest in wearables. And while not at CES, I imagine the industrial machinery market is moving ahead as well. What is common to all of these is the fact that there is nothing common to any industry.

If we are honest about it, it was never reasonable to assume that any one company or platform would be able to cover all of these.  This does not mean that there is not some major platform play out there, but for the foreseeable future we can all celebrate diversity, and the IoT ‘industry’ (movement?) will still progress.

An IoT Taxonomy

At this point, I want to take a step back and look IoT from a very high level. This is actually a note that I have been working on for a while.

What does the Internet of Things actually mean? For most people, IoT means that there will be billions of machines connected to the web. Today, most such connected devices require some form of human interaction to generate web traffic. For instance, a human clicking on a web link in a browser, or playing a game on an iPhone and getting served an ad. In the IoT, machines will ‘talk’ to machines. At its most abstract, the IoT is just a catch-all term for having billions of Internet connected devices.

Dig a layer deeper, and important distinctions start to emerge. First, the most common vision of IoT is the Internet connected vending machine. It sends regular reports to headquarters with alerts when certain flavors are low or the machine needs maintenance. The average vending machine in the US generates something like $25 a month, or $300 a year but loses 15% of potential sales because Diet Coke is sold out constantly. So a device that alerts the owner to restock Diet Coke when it gets low, generates a meaningful return for its owner. You can almost hear the salesman saying “It pays for itself.” Other examples would be pressure sensors on gas wells, or light bulbs on airport runways. But at its most basic, the IoT will connect machines to the Internet and generate some form of business value.

Increasingly, there is a second strain of thought merging into IoT. The idea of wearable sensors and the Quantified Self, has become very popular in the past year or two. The new found enthusiasm was built largely on the expectations around the Apple watch, but has grown beyond to include all kinds of new fitness monitors, and more broadly around wireless health. This points to an IoT of sensors that monitor and ‘help’ humans in some way.  Again, lots of complexity, we will come back to that.

Beyond these basics, there will also likely be a control layer of devices, things that facilitate web traffic. Here I am thinking about systems like Artemis’ pCell which employ some phenomenally difficult math to provide Internet connectivity. In fact, probably the first systems that tap into the potential for IoT are the monitoring systems used in data centers, powered by the likes of Splunk. These are machines (servers) talking to other machines about their ‘health’. This layer is going to blur a bit with the ‘network’ itself, but I think merits attention.

Finally, I would add one more layer that is something like an inverse of the others. So far, all of these devices we have talked about are essentially monitors or sensors, but I think there will also be a category of devices that are ‘actuators’, which take their instructions from the network and then do something in the real world. In some cases, these devices will be the same ones as above, like the monitor on a pressure valve that can adjust the machinery if needed. But I think there will be a much more robust bestiary of actuators. At the apex will be ‘robots’ but this word has overly  anthropomorphic connotations (“Danger Will Robinson! Big words being used loosely. Danger!”), so I want to point out that most of these devices will be much simpler contraptions, drones or microbots or some term not yet invented or just ordinary light switches.

A new Taxonomy

There was a meme in the blogs last year, with reports that the founder of Nest disliked his products being labeled as part of the IoT. Which got me to thinking that IoT has passed some barrier on the hype curve. The press, marketers and analysts have been labeling label everything as IoT. We have reached the point where we need better terms to describe this emerging industry or technology or fable or whatever it turns out to be. And that seems to have begun at CES this year.

So the first divide is sensor versus actuator or input versus output devices. Whenever you see a new connected device ask yourself whether it is a sensor, actuator, or both. Actuators will prove to be one of the trickiest parts of IoT because they will likely need mechanical elements, and they also will likely require a high degree of retrofitting.

A second divide is the expected longevity of the device – permanent versus disposable. Electronics are still expensive enough that this may not seem important, but in the not too distant future I think there will be a large number of essentially disposable or limited-use devices. For instance, we know one company that is working on one-time temperature sensors. When temperatures cross a certain threshold (i.e. a container filled with produce that gets too hot) it sends a one-time alarm and then permanently changes color. For various reasons, this device can be built at a much lower cost point than a longer-lifed, reusable device, and this price cut greatly expands the potential customer base, essentially making the project feasible.

A third distinction will be the connectivity features of the device. Some devices will get by with just local connections (i.e. Bluetooth LE) that connect to a local point of control. Other devices will need to communicate with cellular phone networks, or possibly even satellite networks. Aside from the very large cost implications, this divide will be important as it will set the stage for the power dynamics in the industry. The wireless carriers are still very attached to the notion that every remote sensor is going to need a cellular connection, which has given rise to middleware management layers from companies like nPhase (now part of Verizon) and Jasper (closely tied to AT&T). There is some appeal to this because it frees the user from having to build their own wireless network. However, increasingly we are seeing many deployments that run on top of existing Wi-Fi networks (think of wireless speaker company Sonos).

I will update this coverage on the importance of cellular connectivity at a later date. The standards bodies have been doing a lot of work to help the networks cope with the amount of traffic involved. Dealing with such issues as whether or not each IoT sensor needs a ‘phone number’, a fact which is still true today.

The last divide I would consider is the size of a system. There will be industrial systems that have thousands of nodes (sensors or actuators) and home systems that will only have a few dozen, or personal ones that have only two or three. Big systems will need big management consoles, smaller ones may just need a good app.

What we talk about when we talk about IoT

So for me, CES demonstrated a very promising trend. The way that we talk about IoT is changing. We will probably still be barraged with talk about the Internet of Everything, and marketing around 50 billion connected devices (or is trillion?). But beyond that high level marketing, the real business of building ecosystems is beginning. It will not be one ‘industry’ but new products and features in many industries.

I think this was best on display at the Lowe’s booth. Lowe’s is giant hardware retailer, and I only stopped in their booth by accident, a friend of mine had just bought some locks for his home and saw a new model on display. Lowe’s was promoting its Iris ecosystem of connected devices. Beyond locks, this also included thermostats, sprinklers, windows, alarms and a whole range of other products you could expect to see on their shelves. I do not know much about Iris. It is a freemium service that sends sensor alerts for free and charges a monthly subscription of $10 if you want to apply more detailed rules to that (e.g. alerts when a window opens after 10pm). But they had a whole booth filled with partners. They are not relying on Nest or Apple or AT&T, but Schlage, Pella and other hardware suppliers. Traditional tech industry wisdom holds that eventually there will be one common platform that dominates. That is the economics of software. I think this may not happen in the home IoT segment. The market is just too big, with too many players. We could very well see multiple ecosystems thriving.

Conclusions

I want to close by getting a bit more specific. I have been avoiding doing stock picks in D2D for a while, but it seems to me there are some sector calls to make on this.

First, the component makers are all very hopeful about IoT. This is premised on the idea that there is going to be demand for so many devices, and semiconductor makers love (need) volume. Unfortunately, this is counter-balanced by the fact that IoT devices are likely going to be small and dumb. They will need very basic connectivity, largely Bluetooth and Wi-Fi. They may need a bit of memory, but likely the economics will limit this greatly. They will also need some kind of processor, but again, the amount of intelligence needed by IoT devices will be small. The real brains will rest in the cloud. A debate for a future day will be whether these devices need full-blown processors (ARM, Intel, etc.) or just micro-controllers. That is the difference between a chip being priced for dollars or cents (yes, I’m  oversimplifying). I suspect there will end up being a very wide range of options depending on the application, but the economics are not straightforward. A few years ago I noted that many outdoor electronic sign makers are building IoT modules that are essentially small Android devices. These need more memory and beefier processors which added to their cost, but the ability to tap into volume software programming environments more than offset the extra variable cost. I suspect this may continue. So it was not lost on me that Intel showed off a “PC stick”, a full computer that fits into and HDMI plug-in dongle. Bottom line, the component makers have some reason for optimism, but the real economics are still yet to be determined.

The next category is module makers. There is a very large number of companies building IoT modules right now. That is to say, they are designing printed circuit boards (PCB) with all the associated chips and cases. This is a good business right now because IoT volumes are so small, and almost entirely custom. Let me be clear, this will not last. This model cannot scale. Usually, I hedge my predictions. I do not want to claim accuracy I cannot back up, but in this case, I feel strongly. Maybe a few of these companies can develop a hit product, but this is not the path to IoT riches. True IoT features will get subsumed into other systems, and retrofit bundles will either get incorporated into wider system packages with software or just remain low-volume craftwork.

Above this level, there are a number of device makers coming up with interesting new things, gadgets that will look like robots or at least like Star Wars droids. These devices may be custom industry workhorses or consumer playthings. I think there is a lot of innovation possible here, and one of the reasons we are seeing so many hardware accelerators pop up. As far as I know, there are no public companies with anything interesting here yet, although that may change with some smart M&A. (I suppose Google falls into this camp, they are doing a lot of interesting things with Nest, but it is too soon to buy GOOG because of any of them.)

Finally, we get to software. I think we still need some software or SaaS solutions to plug different parts of the ecosystem together. But I no longer believe that this has to be the domain of Apple, Google and Xiaomi. They will have an important role to play on the consumer, but they are not guaranteed dominant spots anymore.

Clearly one beneficiary of IoT will be all the analytics companies. I mentioned Splunk earlier, as one company that is already processing Big amounts of IoT-like data. And I think that capturing and manipulating IoT datasets will be a big business. However, I think the heterogeneity of IoT will mean we need new classes of software, or maybe some kind of resurgence of middleware. I have an example in mind, but I have to check their stealth status and hope to profile them soon. Instead, my point is simply that there are going to be some interesting opportunities coming from a new dimension that should open the door for new kinds of companies to emerge.

So to conclude,  for once I do not have to close by saying that software will eat anything. There are a number of opportunities up and down the stack here. And most of it has yet to emerge.