IoT generates a large amount of data - sensor data, behaviour data, location data, etc. While this data is usually stored for the long run, it is in its proper use that its value can be realized. And its uses can be various depending on your IoT project and the key issues you need to solve. One such issue could be to address overall operational planning (e.g. do predictive maintenance). In this case more traditional analytical models could be applied. The issue at hand could be however that you need to make almost real time decisions based on your sensor information (e.g. automated vehicles). In this case the value of your information has an extremely short lifespan and you need to employ proper ML algorithms to make that information used in that very short time. Of course training of the models should be based on a very large dataset, but your final model needs to be able to do split second decisions and monitor their implementation on the fly. Whatever your IoT project is, if you have large volumes of data generated, analytical tools can be extremely helpful in turning that data into actions.