Digital Transformation, the learning tool

by Murat Saglam

If you think animals instinctively know everything to survive from birth, you are wrong. Many essential tools in the survival kit are acquired by adaptation, more precisely put, by learning, and learning is what we call structured, planned adaptation.

In order to survive and thrive in today’s business setting, you need not only good business instincts but also an adaptation via digital transformation. Many get it wrong, digital transformation is not the goal to succeed rather it is the ultimate survival tool offering a learning infrastructure.

I am sure you would agree that successful organizations are the ones that value adaptability. Based on a McKinsey study, 75% of companies are transforming their organizations [1]. These companies know they need to change, they seek new ways of doing business. However, based on the same study, companies has a 10% chance to succeed if they don’t have a ‘holistic’ transformation plan.

What makes a holistic plan better than ‘a plan’ ? Is it more intricate? A more detailed one? A meticulously crafted version? Yes to all, but not sufficient. The key differentiator of a better plan is the strength of the intrinsic learning mechanism.

OK, if you are in production business, how do you learn better now? How can digital transformation help you?

Let’s remember basics. Learning is noticing that you are not doing something good enough, changing the way you do it and getting better at it.

Noticing is the key element here and where the 3D concept comes into play [2]. The first D of the 3D is Data. In order to notice issues, bottlenecks, you need to have data representing your operation. With the digital transformation, the doors are wide open now. You can collect a lot of operational data from your crew, planners, target setters and more data directly from machines using wide range of IoT sensors. Having all these data consolidated in one place is priceless. If you are at this maturity level of digital transformation, congratulations! But, I have to say that there is another stop until the Noticing Station, that is the second D, Dashboard. With the second D, the data becomes information. You can visualize your present data next to past performances, or you can set KPIs and observe how your actions reflect into KPIs. That awareness alone could tell that you need to change. This is great, but what is beyond noticing? Is this enough to learn now? Theoretically yes (I reserve here for another blog to talk about Reinforcement Learning [3]), but in practice usually no! In order to learn effectively, we need something more directional.

Let me visualize my point with an example. I am a planner in a production business and my task is to create a monthly production schedule. Only at the end of the month –if I have any data at all– I can be aware of the fact that the schedule adherence is 70% and I have 20 backorders, 30 work orders with passed due dates. The fact that I can quantify these is good to some extent. Now my digital transformation maturity score is 2 out of 3. I need this 3rd D to tell me something more directional. No, 3rd D is not direction. What should I do to increase adherence, decrease backorders, minimize passed due dates? The answer is a forward model. Ok, yet another new term but bear with me. A forward model is a concept that predicts the consequences of your actions before they actually happen. In our example, a forward model is a digital tool that predicts the adherence, number of backorders and passed due dates based on a plan that I intend to deploy to the shopfloor. Learning with past data, the forward model also estimates my present production speeds, so when I schedule a work order with a certain size, I can better estimate the duration I need to allocate. That means I don't need to solely rely on theoretical speeds. Thus, my scheduled work orders are not a group toppled domino stones anymore and they are less sensitive to prolonged production runs, or to theoretically ‘unexpected’ but statistically ‘expected’ failures. Having these predictions even before actual production happens is great. Now as a planner, I can exploit this cycle: plan, simulate, observe, re-plan … until I am happy with the expected outcome. This way, I can refine my Decisions, oh which is by the way the final D of 3D.

Apparently, effective learning relies on an accurate forward model, which has its own –machine– learning journey. To train the forward model, machine learning demands analytical tools for sure but more importantly requires data with large Volume, high Variety, high Velocity, high Veracity, see 4Vs of Big data [4].

I know I know it is 3Ds, 4Vs but you got my point, we need data that can only be fueled with Digital Transformation; and be it either human-, machine-, organization-, or whatever-learning, what we are sure about is the fact that Digital Transformation is right at the heart of it.

[1]: https://www.slideshare.net/aipmm/70-26633757

[2]: http://scm3d.com/

[3]: https://en.wikipedia.org/wiki/Reinforcement_learning

[4]: http://www.ibmbigdatahub.com/infographic/four-vs-big-data

Go to top of page