A long time ago we wrote about how messy the Internet of Things (IoT) is. And while we are big believers in IoT being a huge thing someday, it is still a mess when it comes to actual components. In that original post, we pointed out that the industry is incredibly fragmented and as a result something called an “IoT device” can mean many different things. The sheer number of possible use cases means any such device could have a whole array of logic, sensors/actuators and connectivity options, making for an overwhelming set of combinatorics.

This is true for each of those subsets. Here, we want to take a look at networking options in particular, both as a way to explore that market and to provide insight into the broader problem.

There are basically five different ways to connect a device wirelessly, as most IoT devices will be – Bluetooth, Wi-Fi, cellular, Zigbee, LoRa as well as a few more niche methods. The choice of which to use depends entirely on the “use case” of the system – that combination of location, purpose and end user.

Those five technologies can be divided into long-range and short-range solutions – cellular and LoRa can operate over miles between the device and their connection point to the Internet. By contrast, the other three can really only operate over distance measured in tens or hundreds of meters. Another consideration is data rates. Cellular bandwidth is measured in MB/sec and Wi-Fi in GB/sec, while the other three are measured in kb/second. Not all, in fact most, IoT use cases work perfectly well with those lower data rates. Finally, there is the cost of the system. Setting up a cellular network is very expensive. By contrast, anyone can set up a Wi-Fi network, with the others somewhere in between.

Air Interface	Range	Data Rate	Operations Complexity	Device Cost	Network Cost
Cellular	Miles	High	High	High	High
Wi-Fi	Meters	High	Low	Low	Low
Bluetooth	Meters	Low	Mid	Low	Low
Zigbee	Meters	Low	Low	Mid	Mid
LoRa	Miles	Low	High	Mid	Mid

So how does that play out in the real world?

Often when people talk about IoT they are referring to in-home use, which we actually think of as Home Networking, distinct from IoT. In this use case, there is a reasonable assumption that the home has a working Wi-Fi network. At the other extreme are highly spread out deployments – like an oil field that covers thousands of acres. This dictates use of one of the long range systems. Typically, these deployments are vast sensor networks, which translates into very low data rates, making LoRa a viable option.

In between those two extremes are tens of thousands of other possibilities. A prototypical example is a fleet of delivery trucks. Owners want to keep track of the track of the trucks (so add GPS), but they may also want other types of data – for instance tracking the drivers’ speeds or the temperature of the trucks’ contents. This tends to favor cellular – long distance plus larger amounts of data. But this an enterprise deployment which means they also have to consider back-up solutions which likely means adding LoRa or possibly Wi-FI. That second system does not need to be as reliable or used often, but will be important when it is needed.

Another important consideration in all this is who will operate the network. You manage your home Wi-Fi network. How is that going? Imagine the headaches involved in getting coverage all over your house plus initial set up plus fixing it when it goes down and multiply all that by orders of magnitude. Bluetooth is in many ways a perfect technology for many IoT use cases, but there are very few bluetooth network operators today, making scaling deployments hard. LoRa is a bit better in this regard, for the most part there are no national LoRa networks anywhere in the world, instead just a sea of independently operated LoRa networks. Cellular networks can rely on the operators to manage all these problems, but that makes it hard to integrate into other enterprise IT systems (hence a key selling point of private networks).

Finally, we have to consider the cost of all these things. Aside from the cost of setting up the network there is also device cost to consider. Bluetooth is very cheap, below a $1 per chip, cellular is probably the most expensive, but LoRa devices can be expensive too as this ecosystem lacks scale volumes. Zigbee is similar, it is ideally suited for some very specific applications – in-home lighting, warehouse and industrial settings. But these are all fairly small markets, which means the cost of Zigbee systems can be higher than Wi-Fi.

We wish that we could say there is a solution to all this, but there is not. In fact, these endless permutations should be seen as a feature of IoT, they provide immense flexibility for everyone to customize individual solutions. The fact that many are disappointed by this speaks more to the funding markets than anything else. If one company could unify all of this it would be immensely valuable. Instead, we are left with dozens of vendors providing tailored solutions. In the future we will take a look at companies like Silicon Labs and Samsarra who are building some solid platforms here. None of this is easy to scale, but still good businesses potentially.