Zen And The Art of Building an IoT Enterprise
I don’t have decades of experience building an IoT Enterprise. However, my PhD research involved developing (sensor network and robotic) systems and algorithms, deploying them in real world and then making sense of the collected data. I did the same during my academic stint at IIIT Delhi and have been following upon it here at Zenatix. So that does provide me with more than a decade of exposure to building and deploying systems from the perspective of exploiting the collected data for real and meaningful applications.
Photo Credit : ShutterStock,
IoT of today was sensor networks of yesterday. The applications might be strikingly different but there are huge similarities in terms of constraints of the systems, applications motivated by the desire to collect data that is never seen before and eventually automate the whole process. Sensor networks were largely motivated by defense applications and infrastructure monitoring (e.g. bridges), eventually also leading up to environmental sensing applications (this is the area I worked in during my days at UCLA). Largely, it catered to environments that were not easily accessible. IoT on the other hand, extends the environments to pretty much everything that surrounds us - manufacturing, transportation, retail, healthcare, buildings, utilities and cities, among others.
In the sensor network days, pretty much every early research paper would start with the motivation that this environment that is inaccessible has to be monitored with sensors that will be air dropped, they make their own network and then start sending data that is never seen before. As far as I know, this image was never realised (or it may have been for some defense applications). IoT in a similar manner was much hyped and over the past couple of years, some realism has stuck. Some recent projections are still quite humongous - 24 billion IoT devices by 2020 (growing annually at 41% CAGR), 6 trillion of investments in IoT solutions over the next 5 years which will generate 13 trillion of ROI by 2025 with governments and business being the primary driving force (consumers being the third in all respects - devices installed, money spent and ROI achieved). These statistics have been borrowed from here. At this point, I must also say that IoT should not be confused with the smartphones. Yes, smartphones are connecting us and collecting information about our each and every action, but there are fundamental differences (one of them being smartphone as a generic device while IoT devices/applications being vertically focused in one of the many environments that I mentioned earlier) and therefore the two should not be put together.
While I mentioned about IoT hype being similar to that of sensor networks, my personal belief is that with the pace at which technology advancements are happening and with focus on environments around us (which are much more accessible and therefore can provide rapid improvements in systems and algorithms), IoT proliferation will undoubtedly happen - the only question is “When” and “How Quickly”. The second belief I have is that the driving force for this eventual pervasive connectivity will not be the desire to just connect things, but to use the “dark” (never seen before) data to remove inherent inefficiencies and to integrate a suite of services around a given vertical. A follow up of that belief is that for any IoT application to succeed, one should be clear about the value it delivers and whether it make sense to spend money connecting the necessary things (because this connection will add the cost overhead and this cost overhead should make sense) involved in delivering that value. Looking from that lens, a web connected microwave or a smart home which allows me to switch off my lights sitting in my bed seem far behind when compared with applications in healthcare, energy analytics, mobility and industry.
Further, IoT of today should be looked upon more as a means to connect the physical and the digital world and not an end in itself. One may think of IoT devices as “hardware APIs” connecting the physical and the the digital world. However, one should not underestimate the effort involved in building and scaling up these hardware APIs due to the aspect of physical installation required for these IoT devices. Some argue that eventually things will come with standard APIs to connect to internet. However, I am one amongst the skeptics who are not sure if all connected devices from different manufacturers will come with standard and open APIs for any third party to connect to them. After all, who would not like to exploit the data coming out of these devices. Even if one can not build capacity to internally analyse the data, one can at least monetise that data through someone who may want to build enhanced applications and in process build a significant dependency on their systems for the third party to sell their enhanced application.
I presume that as and when the devices get connected (whether its because of someone like us putting in our own gateways or the ecosystem has been developed with devices coming with in-built open and standardised APIs for connectivity), the platforms that integrate these connected devices to provide advanced level of analytics eventually driving every ounce of efficiency from the system and/or integrate a suite of services to improve the overall benefit to the end consumer will be the winners. The key here is “once the devices are connected”. And with lots of upfront investment required to replace the existing devices with their smart and connected counterparts, this is unlikely to happen in the next 5 years.
Low cost sensing together with low cost and pervasive connectivity is ready to be exploited today. This implies that someone who wants to leverage the platform play for tomorrow has to start with devices and gateways to retrofit and connect the existing devices and be the first one to access the “dark” data, build the use cases and the models around these uses cases and eventually be ready with the IP once the connected devices become the reality. Data will be the eventual differentiator and early access to that data starting today will separate out winners from losers.
Our journey at Zenatix has been somewhat similar. We wanted to be an energy analytics company but very soon realised that pervasive connected devices (to drive operational energy efficiency that we wanted to target) are absent in today’s world. We had two options in front of us - build up hypothetical use cases from the small amount of data that existed (and simulate the rest of the data) or to push ourselves to be an “IoT driven” energy analytics company (building the hardware that retrofits the existing systems and hardware with connectivity and provide ourselves with interesting datasets whereby even some basic analytics throw up interesting business opportunities to exploit both in immediate and near future). We chose the later option and while every new customer we deploy our systems with (we now have more than 400 systems deployed in the field) has resulted in interesting new use cases (further strengthening our hypothesis), the journey has been challenging to say the least. In my earlier three articles, I have discussed in detail about these challenges and three key differentiators that we have built which provide us with significant competitive edge. At a high level, these three technology differentiators are:
Robust and extensible hardware - something that on one hand is hard to achieve with prototype systems (such as Arduinos) and on the other hand is challenging to do with your in-house developed systems, especially in the Indian context, due to the lack of supporting ecosystem. The only eventual solution is to slowly and steadily build this capability in-house and be ready to take off once the robustness at a reasonable scale (a few hundred systems running properly in the field for an extended period of time) has been achieved
Software and hardware developed in an integrated manner - Avoiding the hardware to only offer stack will limit the offering and scope significantly. Joint development will give better control over the whole system - Apple is an example.
Technology backend to support the logistics around installation and maintenance of the hardware in the field - this has a lot of similarities with the technology behind the logistics requirement in e-commerce for which enough has been written and talked about.
Realising the right product-market fit in an IoT domain (one that delivers real value to the customer) could be a challenge. Narrowing down the specific aspect of the data inference and the corresponding customer segment to focus on will likely be an iterative process. Starting from a wide array of potential customers, narrowing down segments where value is significant in a quick manner will be useful. Assuming that the technology (hardware, software and logistics backend) is built to support scale, eventually the value delivered will increase with deeper analysis of the collected data and/or with incremental (and easy to do) additions to the existing hardware and/or integrating a suite of services around the primary use case. The key therefore is that once a product-market fit is identified, get to a reasonable scale as fast as possible. This will allow you to do things that will increasingly make it difficult for the newcomers to catch up with you, such as:
Your hardware is already deployed - A switch for hardware is not as easy as a software switch and can not be done overnight. Your competitor (even if they come with big pockets and offer the solution for free) still have to physically visit each site, remove your system and install theirs. Even if they are confident of doing it, the customer may be reluctant to spend their energy again if the value delivered is not significantly larger that what is being already delivered (and it will not be a simple price game). The onus therefore is on you to ensure that you keep increasing the value being delivered which takes me to my next differentiator.
Advanced analytics with deeper domain expertise - If you are amongst the first to get access to the “dark” data and also the first one to understand the deeper nuances of different verticals within the identified customer segment, then you can build significant IP as well as provide increased value proposition which will be hard to compete with for any new incumbent.
Leveraging the data for other applications - Once your systems start generating data and you have models to ingest the incoming data, this can be extended to use cases beyond the customer segment you are currently focused on. For example, as we optimise the air conditioning for our customers, we collect data that provides insights about performance of AC in different environments (from hot and dry in Delhi to humid in Mumbai). This data can be very useful for the AC OEM players to advance their systems further.
Disclaimer: The views expressed in the article above are those of the authors' and do not necessarily represent or reflect the views of this publishing house
Around The World