I’m David Hartwell, an IoT specialist at Tech Data. I hear a lot of references in the industry to “Edge to Enterprise” as a short hand for “All encompassing”. Sadly, I have seen this to really mean “The hard work’s done, sensors, communications and cloud are working well and here’s your data”; then the hard-pressed analysts in the organisation are tasked to do their clever stuff to make some sense of it. IoT has the potential to offer genuine insight, but rarely by simply presenting more sensor data. This can cause more harm than good, as it is very hard to know whether a particular sensor is showing an abnormal condition or not: you have to know context. So, a particular pump could show high vibration readings regularly, but after a quick chat with the operations director, he might say: “Yes, that pump often runs dry and cavitates, the design is not ideal, so I’m not surprised that one vibrates from time to time. It’s OK, we know about it.” This is context. Exact same model of one pump will vibrate, another one won’t, neither causing the operators any concerns.
Because of such complexities, the tendency is to present data to operations and maintenance staff without too much interpretation and let them apply their deep understanding of the plant and equipment along with their expertise. Unfortunately, this takes time and makes it difficult for senior management to act now. They have to use interpretation of historical data to inform them of system behaviour and use that to inform them of the right course of action. A typical scenario might be: “It’s OK, don’t send anyone out to that pump, it’s most likely running dry”. Operations folks make decisions like this all the time, that’s often the job – to make decisions based on a few basic pieces of data. Not surprising that it can go wrong, maybe that pump was not running dry, it was really pumping liquid at full power, but the bearings failed and now it’s broken down causing major process problems. Such is the life of operations and maintenance managers. IoT can easily be relegated to simple process control and provision of historical data. It’s true that such an IoT implementation might inform future behaviour, but often it is only really used to show failure or abnormal conditions that need remedial action now, in other words, there’s no distillation of the data to help with predictive analysis.
What if we could sense vibration, power, flow, temperature and anything else that might be relevant, wouldn’t that insight help us prevent another breakdown? “Well, yes” they’d say, “but, we’d be overloaded with data, we’d never see the wood for the trees”. It doesn’t have to be this way: I hope I’ve made the case for analytics and machine learning in an IoT implementation but even that is not enough in my opinion. Many IoT implementations start with putting sensors on equipment we want to know more about, and that sounds like a reasonable strategy, but I contend it really should be the other way. We should start at the screen: what the customer needs to know to act now, and then finish up at the sensor. So, rather than “Edge to Enterprise” it should be “Screen to Sensor”.
This is not a semantic distinction – it’s putting the user at the heart of the process. Like any new technology we all tend to get caught up in making it work, rather than making it work for us. That’s why when we start an IoT implementation – we start with understanding the people and what they want on their screens.