Turning Process Data Into Intelligence

I have always been a big fan of MS Dhoni. A few days back I was wondering how would it feel to bowl to such an explosive player. Co-incidentally I came across an article showcasing his performance from a match on various type of deliveries. It is beautifully depicted in the image above.

The use of data science in cricket has increased significantly in past decade. Teams decide their strategy purely on data now. Decisions right from electing to bat first , to bowl particular type of deliveries to a batsman etc are being taken based on such ‘performance analytics’

Based on the data shown in the image above, I would refrain from bowling a short length to MSD as he has a strike rate of 100 for those kinds of deliveries. I wouldn’t want to ruin my bowling career would I. My aim would be to bowl dot balls if not to get him out. So, I aim in the yellow zone- Yorkers and full length to get the desired output- a wicket! More performance data I have, better are my chances of finding a batsman’s weak spot.

Imagine the amount of value you get if you apply such analytics to your manufacturing processes. When you have time series process data in hand-coming from DCS, PLCs, historians etc you just need the right tools to process the same and analyse to get the desired output -better yield, better capacity utilization , optimized energy consumption etc.

One of my customer’s has been able to improve the yield significantly within 3-4 months of deployment of an IoT platform. They refined the warning and the control limits after creating a baseline to get real time alerts. Using machine learning algorithms, the platform turned the process data into intelligence and started predicting the batch yield. Yes,the user’s domain knowledge was also crucial in this case. We are now moving towards ‘prescriptive analytics’ - the IoT platform has started prescribing changes in the process parameters to enable better yield. This is like- say suggesting to change the fielder position after an unwanted delivery has been bowled before it even reaches the batsman.

One important thing here is, an IoT platform should have all the necessary tools to churn value from your time series process data. Having to plug the data to some other app / system for further analysis defeats the purpose of centralized intelligence (although inter-operability should be seamlessly possible) I would love to discuss further about possible use cases for various industries. Feel free to reach out to me on this :)