Data is the new oil of the digital economy!
Have you seen that statement before? Probably yes!
We are constantly seeing stories of companies that managed to traverse the boundaries of “traditional” business models into one where value is created through data. And usually either creating new industries or, as it’s usually put, “disrupt” existing ones.
So, do I think that is incorrect? No! I’m fully behind the notion that the exponential growth of data together with new and stronger analytical possibilities will provide us an extremely fascinating future.
But just like a bow tie, it’s not easy to succeed and it’s not for every occasion. Almost all IoT pitches that I see somehow contains a part where data is intended to be made into a new revenue stream. And often when I dig into it, there is little practical detail on how to do it. This is even more true when looking at “traditional” businesses that want to make sure they don’t fall behind by making their existing products data generating tech wonders. Some do succeed, but taking survivors bias into account, the overwhelming number doesn’t.
And the truth is that a large majority of successful IoT products are not successful due to their analytic capabilities, but due to being connected. It might add to some extent value to the service, but it doesn’t define them. Some connected products are just that, products that are connected, and they do not need to revolutionize how you earn your money.
But of course, while pursuing data driven benefits as the end goal is a risk and can lead down the wrong path, ignoring it is even more of a risk. There is more or less always some value to bring from collected data, and the challenge is coming to an understanding of what that benefit is.
But don’t expect the opportunity to present itself. It requires a lot of hard work and expertise to identify the possibilities. Which is of course a good thing, as otherwise someone else would already have done it.
Moving on to a few practical tips on how to approach data analytics within IoT.
1 - Understand the data that can be or is being collected. What does the gathered data actually measure and what is its granularity and precision?
2 - Map any other data that’s relevant. This can be both internal (e.g. historic consumer data) and external data (e.g. weather data).
3 - Identify opportunities based on the identified available data. This is the hard part and requires an innovative approach or a good understanding of other case studies. Here is the crucial time to remember that not all opportunities will be revolutionary. It’s usually easier starting with something simple that’s easily understood.
4 - Understand the properties of the opportunities. Do they rely on raw or aggregated data? Are they based on real time analytics or do they rely on historical data? It also requires a choice of analytical method (e.g. Machine Learning might be an option?). These choices significantly alters the architecture and infrastructure needed.
5 - Decide which data to gather and in what form to store it. Brian Krzanich (Intel CEO) predicts that autonomous vehicles will consume 4 Terabytes of data every day. Deciding which data is relevant to store is important to not get overwhelmed by it. Should some data only be analyzed in real-time and then discarded? Is some data only valuable when aggregated? For how long is the data useful?
While it’s definitely a challenge, it’s important to have the expertise to be able to understand the opportunities arising with exponentially more data being collected. But at the same time the value of the collected data shouldn’t be inflated. A practical approach with a focus on real opportunities rather than a promise of an elusive golden data driven future is important.
Partner and Senior Advisor Mikael Rönde, firstname.lastname@example.org, +46 (0)70 88 66 794