Why are they Different?
Big Data and IoT (the Internet of Things) may at first glance appear to be very similar. After all, both collect lots of data and analyze it to extract information. Therefore we might think that techniques and architectures developed for Big Data can be successfully applied to IoT. Unfortunately we would be wrong.
Why are they different? Think of Big Data and IoT as doctors. Both use information to diagnose and treat diseases. So far they sound the same. Let’s say they are treating the same disease even — the flu! Surely all doctors treating the flu approach it in the same way? Well .. no. Dr Big Data is the epidemiologist. It tries to extract patterns from data about millions of flu cases to find ways to increase the health of the entire population. That’s great. But Dr IoT is a general practitioner. Its job is to diagnose when their individual patients have contracted the flu, and determine the best course of treatment for that specific person, based on their unique medical history.
The Requirements Implications
Why does that mean that Dr IoT’s analysis must be completely different from Dr Big Data’s? Now that we have drawn the medical analogy, I think you know the answer yourself. If you visited Dr IoT concerned you might have the flu, and Dr IoT told you they were running an epidemiological study and asked you to come back in 6 weeks when she expected to have a response which would probably work for 33% of the population, you would not be impressed. Timing aside, you want a treatment you can expect to work for you, and you would expect Dr IoT to be able to explain why it will work for you.
So here is the difference: IoT’s results must be specific, and they must be explainable. They must tell us what is happening to a specific subject, and why we can trust this answer to be true. In other words, they must be deterministic. We are going to adjust a control input, schedule a maintenance visit or invest in a major refurbishment or upgrade. Making the wrong response to an issue wastes a lot of money, something businesses are loathe to do. Even worse, the wrong action could cause damage itself, even more damage than living with the original problem.
The Key Difference in IoT vs Big Data Analytics
What does this mean for IoT analytics? Determinism. Big Data can rely on statistical techniques to extract answers that will be correct some of the time. It can rely on machine learning which may be true much of the time, but cannot be explained or verified. In many applications, such as market analysis, “some” might translate into a very small proportion — increasing sales conversion by a few percentage points can justify a very large digital marketing campaign indeed — and being able to explain why a relationship exists may not be necessary to take action — again, the fact that 5% of customers prefer blue is valid in its own right. In contrast, in IoT, a human must be able to justify using information to make weighty decisions for a specific subject in situations with low acceptable risk. Therefore, IoT requires deterministic analytical models, which always get the answer right, in a way that can be validated through a logical process of deduction. Practical application examples include inventory management, environmental controls and equipment/machine maintenance.
Determinism is the difference!