I’ve been wondering for some time how long it will be before Sigfox ditches their own proprietary protocol and adopts a different standard, probably going for NB-IoT. Whilst many may question why they’d do it, making the assumption that the Sigfox protocol is their crown jewels, I’d question that assumption. Last week, at the Sigfox World Expo in Prague, they made an announcement that suggests that that day may not be too far away. I firmly believe that it has to happen, because Sigfox’s business model is looking a lot like Uber’s, which means going for global domination by any means possible. The means to that end is probably not based on their own proprietary protocol.
Sigfox started life because the mobile network operators took their eye off the ball. Despite the fact that the operators were gung-ho about the opportunities from the Internet of Things, they were rapidly refarming their 2G spectrum which supported GPRS – the cellular service which has been the workhorse for M2M applications and the potential service for the early years of the IoT, leaving their customers with no viable solution. They’d become so obsessed with the data requirements of a new generation of smartphone users and the voracious data appetite of mobile video, that they’d forgotten to design a replacement. Instead, they assumed that IoT users could stomach the additional cost and power consumption of 4G modems. That was so fallacious, it makes you wonder whether they actually understood what the IoT is.
Sigfox saw the opportunity to step in. They had developed a simple communications protocol, running in the license-free ISM bands at 868MHz and 915MHz. It uses very low cost, standard chips, has a usable range of 5 – 10 km and a battery life which can support years of low data-rate transmission. At first glance, it looked like the answer to an IoT maiden’s dream. The only problem they had was how to deploy it. Unlike the roll-out of cellular connectivity, where you could start with the areas of greatest use – typically capital cities, the IoT customer base is much more diverse. Cities come into it, particularly if they’re trying to be smart, but there are plenty of applications in agriculture and transport which need much wider coverage. Building their own network would have taken too long and cost too much, so they persuaded mobile operators to partner with them and install Sigfox’s gateways on their existing towers, providing a fairly rapid coverage, as the map below shows. It may have a lot of uncovered territory, but it’s impressive for a few years’ work.
You can see the short term appeal for network operators, as it fills a hole in their IoT capability and keeps them in the game. But it raises the question of what part is left for Sigfox? With the mobile operators supplying the network, what is Sigfox’s role? This is where we start to see the Uber analogy.
To understand why, you need to look at the value stack of the IoT. I’ll cover this in more detail in a future article, but the diagram below shows the basic blocks.
You start at the bottom with hardware – the sensors, which generate your data. As a general rule, the cheaper, the better, which is a plus for Sigfox. You then have the annual data charge for connectivity, which is the operator’s bread and butter. Sigfox will get a cut on both of these. What is interesting is what comes above. The major new item is IoT infrastructure, which could also be described as the Device Life Cycle management. For IoT applications this includes security, updates, monitoring the state of the network, provisioning, cloud ownership, data validation and data storage. It’s everything that you need to ensure that your network of deployed sensors stays up and working and that the data you collect is robust and reliable. Above that you have the data analytics which generate sector-specific insight and value, which is why companies put in the time and effort to deploy IoT applications. On top of all of that you have the overheads of managing the whole process.
For most of the history of connected devices – through M2M and the foothills of IoT, the bottom two of these have been the most difficult. The companies who made money in the M2M era achieved that by providing pricey, custom hardware or eye-wateringly expensive mobile data contracts. That’s why so much of the IoT debate is about owning the hardware and connectivity, as analysts and incumbents still think that these will be profitable areas. However, as volumes go up, economies of scale will make them progressively cheaper until both are commoditised. At that point the value moves up to the IoT infrastructure and applications portion of the value chain. Applications will remain largely the purview of individual industry sectors, leaving the IoT infrastructure piece as the opportunity for a few, specialised companies to extract value from everyone using the IoT.
Owning that IoT infrastructure piece is where Sigfox would like to be. Their endgame is to be the company that can take a slice of every piece of data which travels over the IoT. Effectively they want to become a global IoT operator / aggregator, using network providers to handle the nuts and bolts of connectivity, while they cream off the margins coming from a higher-level play. They’re not the only company after this – most of the network operators think it’s their prize, but history has shown it’s too orthogonal to their standard business model and expertise. It’s a bold and disruptive vision, which is why Sigfox has managed to accumulate over $300 million of investment. Whereas transforming applications to be a service is what differentiates the IoT from M2M and is the top down view of most companies which invest in the IoT, Sigfox would like to own the bottom-up approach of providing the IoT as a Service.
This is where we start to see the Uber analogy. Any company acquiring that level of investment is signing a Faustian pact. Once a company has received over a quarter of a billion dollars, the VC’s providing that money want to see a unicorn, not just a French filly with an antenna on its head. That means an exit. It may either be an acquisition or an IPO, either of which will value the company in the multiple billions. In Sigfox’s case, they’ve set their sights on an IPO. Whichever path they choose, for the exit to be credible, they need to show usage. For Sigfox, that probably means being able to demonstrate enough connections to satisfy new investors that the IoT has really taken off. Which means they need to connect something in the order of 100 million devices.
Reaching that number is proving remarkably difficult for everyone. Despite the hype, the journey to scale the IoT up to the predicted billions of connected devices is very hard work. Sigfox hasn’t been helped by the enthusiasm shown by the maker community. Low prices and a simple protocol have attracted the Arduino and Raspberry Pi folk to connect all manner of things. They’re good for publicity, as the latest MKR FOX 1200 challenge has shown, but these are mostly hobbyists, generating a few million connections. There’s nothing wrong this fluffy end of connectivity, but it’s a time-consuming distraction on the path to the real IoT.
Sigfox needs to break through this maker barrier if they are going to launch a credible IPO. They need to engage major, global vendors who will ship millions. Driven, I suspect, by their investors, they have started to follow Uber’s approach of using their cash mountain to subsidise customer acquisition. Where Uber is using predatory pricing to subsidise rides, Sigfox is supporting cheap hardware modules and lower data tariffs. It’s the old strategy of under-cutting the incumbent industry players to drive them out of business, so that they can acquire monopoly status. For an excellent analysis of how Uber is playing that game, read Hubert Horan’s analysis of that company. Sigfox is not in the same league, but it’s essentially the same modus operandi. The problem is that it’s not getting the same level of traction.
There are multiple reasons for that. As I’ve said, the IoT is surprisingly hard, and the IoT industry hasn’t helped itself by scoring some own goals. One major one has been the fragmentation of incompatible connectivity options. Whilst that’s the opportunity which Sigfox was able to exploit with its LPWAN solution, they weren’t alone. Other players, most notably LoRa, have helped to confuse the market; a confusion which is exacerbated by the acceleration of new solutions from the mobile industry, who are none too happy to see this unwelcome competition. We now have the initial trials of NB-IoT and LTE-M, which are further fragmenting the market. The deployment of all of these tends to be driven by national concerns, with the result that there is no global option. That doesn’t bother the maker community, who typically only ship to tech-savvy early adopters, who they expect to be aware of whether or not the product will work. But that’s a million miles away from the concerns of real industry who want to be able to ship a single SKU of product that connects anywhere in the world. Whilst connectivity remains fragmented, they’ll put off production plans, because they know from bitter experience that the customer support overheads of multiple, incompatible connectivity options across a global market would kill any project.
As NB-IOT and LTE-M have begun to mature, or at least approach early adolescence, it has had the effect of raising renewed questions about whether Sigfox’s simple protocol is sufficiently scalable. There have always been concerns about scalability and Quality of Service within the constrained 868MHz band, but a bigger concern may be the protocol’s essentially one way nature. Although most IoT data is one way, as it’s data sent from the sensor, to use it for insight it frequently needs to be collected as time series data. To ensure the validity of time series data you need to timestamp it, and without a responsive two-way connection to synchronise clocks with an NTP server, there’s a real risk that data integrity is impaired. For me that’s a mortal blow to Sigfox’s ongoing ability to pick up hundreds of millions of connections. That’s before you hit the related issues of an inability to deploy firmware upgrades and device management.
That’s what made the announcement of a combined Sigfox and cellular modem so interesting. It supports Sigfox’s protocol, as well as NB-IoT, LTE-M and EC-GSM. There’s no detail on cost and it’s certainly not going to match the $0.20 disposable chipset which Sigfox also announced at the Expo, but it provides a global solution and may be just what’s needed to persuade larger companies that they can begin to contemplate manufacture at scale. Sigfox’s press release is a little self-serving, suggesting this was all their idea. The reality is that GCT’s chip – the GDM7243i, is one of the first of the next generation of multi-radio chips which all of the IoT silicon providers are developing, combining NB-IoT, LTE-M, Bluetooth Low Energy and a sub-GHz radio. Once the radios are integrated in silicon, what it supports depends on which software stacks are available. It could just as easily run Z-Wave or any other sub-GHz protocol (and probably will in the course of time). But Sigfox got in first with spotting the potential, so good for them. Incidentally, Semtech’s ownership of the IP for their LoRa radio may put them at a significant disadvantage is capturing the benefits of chips like the GDM7243i, as it won’t be supported by these multi-purpose chips. That could be a major threat for LoRa as other silicon vendors launch multi-band-multi radio application processor chips.
So how does a combined Sigfox / cellular modem change the game? The hardware won’t be as cheap, but the price will fall – that is just a numbers game. GCT are out there first, but ARM has acquired companies with expertise in the radio technology and protocol stacks, so expect a lot more offerings. Meanwhile a new tranche of wireless chip specialists, including Nordic Semiconductor and Hisilicon are hard on the heels of GCT. The industry is already talking sub $5 for these chips in volume. If it provides the security for manufacturers that their product will connect anywhere in the world, that becomes interesting for any product that has a retail value above $100, which gets you into the billions of chips territory.
The more interesting question is how tariffs will be arranged for these multi-protocol devices? If you start with a Sigfox tariff of a few dollars a month, but need to move to NB-IoT because of a lack of coverage, would the network operator support the same data tariff? If so, it’s a win for Sigfox, who gets to own the customer. It would also mean that Sigfox would have effectively gained the power to set cellular IoT tariffs, emasculating the network operators.
The flip side, is that by including cellular connectivity, network operators, who today see Sigfox as a necessary evil to get to market, could simply switch consumers to their service by offering them a better tariff. Either way, it commoditises the IoT, which is great for customers.
Sigfox probably doesn’t care. In the same way that providing a cheaper taxi service is just the route to a larger goal of being the global provider of Transport as a Service for Uber, so getting the number of connections that enables them to develop their IoT as a Service model is the goal for Sigfox. To achieve that they need to get to the hundreds of millions of connected devices as quickly as possible. If they can do that and become the IoT infrastructure provider of choice they will lock out competition. But some other very big players are lurking in the wings. And the mobile operators still believe that it is their rightful destiny.
Sigfox faces the quandary which the Uber model gives us. Thought leadership and commoditisation aren’t good bedfellows – you generally need high margin products if you really want to develop value. Right now, most of the IoT is comprised of companies sparring for thought leadership, but unable to deliver on that until commoditisation arrives. The arrival of multi-mode communications chips combining proprietary and new cellular protocols may be just the catalyst the industry needs to break that deadlock and move towards commoditisation. But it is a two-edged sword. Sigfox needs to act quickly to get subscriptions up. These new chips could accelerate that, but they could also have the side effect of moving power back into the hands of the mobile operators. The next year will be an interesting time.