Stage 3: IoT Connectivity
The third stage is where IoT solutions truly come into their own: interconnectivity. At this point, devices talk to each other, and we start to see a connected ecosystem take shape.
The technical challenges to build a connected product are even more difficult. Of course, we still need all the expertise from stages one and two, but now we need an even greater skill level to be successful.
We ask a lot of our connected devices, yet these embedded systems work on constrained hardware. Integrating various services, especially when their points of origin are so different, is a major hurdle. Security becomes even more difficult, and we really need to think about building in security from the start; for instance, we’ll want to embed a hardware security module (HSM) chip into our circuit board.
One of the hardest parts of IoT development is that we need to make every little bit count. While a more powerful computer can afford to dedicate a little disk space or processing power to applications that are only ‘nice to have’ or even downright unnecessary, IoT devices lack this luxury.
That’s why tooling like Nerves is so useful: it lets us build a custom Linux system that only has what we need and nothing more. However, actually knowing what to include and what to drop requires a lot of technical knowledge.
Stage 4: Predictive Analytics
This is the stage where we really start putting our data to work. Predictive analytics for IoT looks at trends like sensor data, user engagement, and other metrics that we get from our devices. We can then use that big data for tasks like predictive maintenance for industrial IoT.
Stage four is where data scientists become more critical. These professionals use tools like Python, PyTorch, and AWS SageMaker to build, train, and deploy machine learning models, but that’s only a small part of the job. Foundational to any successful data science project is an analytical framework, a way of thinking critically about data and business problems. Sometimes, the hardest part is just finding the right questions to ask.
We can’t, however, just throw a bunch of numbers at a data scientist and expect a fully-fledged predictive analytics model in return. We need a cross-disciplinary approach where our data scientists work closely with our engineering teams to develop a data pipeline. After all, if our hardware engineers don’t know what data our analysts want to use, how will they know which sensors to choose? Likewise, our software developers need to understand the data scientists’ priorities so that they can figure out if they need to derive any variables, whether to aggregate data or simply push it to the cloud, and even which data points need to go to which databases.
Stage 5: Prescriptive Analytics
Taking our data-driven approach one step further, this stage is defined by prescriptive analytics, which builds off the predictive power of stage four analytics by then recommending future courses of action. IoT companies can use prescriptive analytics to offer long-term value to users because they have the potential to make our lives easier, more convenient, and more enjoyable.
On the tech side of the equation, stage five includes many of the same elements of stage four, yet they’re all required to function at a much higher level. Just when it comes to data science, for instance, we drastically expand our scope; we’re no longer just using a single model, such as anomaly detection for preventative maintenance, but rather we’re now using a quilt-work of interwoven ML models to pull off some truly spectacular feats. These may include things like Natural Language Processing (NLP) for speech recognition/voice commands, algorithms that optimize according to the OCEAN personality model, and much more.
The result is something that begins to truly resemble Artificial Intelligence (AI), and so it’s not hard to see how these challenges span further than just data science. Our hardware team, for example, will need to find creative ways to embed even more processing power into the most compact spaces, such as with GPUs for edge computing. Moreover, a stage five product is never truly complete. Agile practices such as continuous integration/continuous deployment (CI/CD) are crucial if we want to continue to provide a world-class IoT experience.
Stage 6: Ubiquitous Computing
The final stage of the IoT maturity model is ubiquitous computing, an endgame where virtually every aspect of daily life includes some interaction with the digital world. Currently, this stage only exists in science fiction, but we might be closer than you think.
The tech that it would take to get here is immense, and all we can really do is speculate at this point. We do know, however, that it will take a collective masterwork in engineering, software development, data science, user experience design, and more. Building a collection of talent in these domains is the biggest obstacle preventing us from entering the world of ubiquitous computing.
We have a long way to go. Let’s start building.
It should now be clear how much more difficult each progressive step is than the last. Even the transition from a stage two device to a true stage three IoT product is a massive leap, as it requires expertise across many domains and forces us to master many different technologies.
Even though today’s most advanced tech companies boast stage five maturity, we still don’t have anything close to ubiquitous computing. Thankfully, many of the greatest minds across the globe are working to advance thousands of different technologies.
That doesn’t mean the current state-of-the-art isn’t already changing the world.
Ready to dive deeper? Download the full PDF of the IoT Maturity Model here.