Can the Fourth Industrial Revolution succeed without 100% reliable technical data?

Recently I wrote about the question of the human role in the Fourth Industrial Revolution. In this article, I want to address another common question: How important is data accuracy in a highly digitalized manufacturing enterprise? Can the Fourth Industrial Revolution work if data is not 100% accurate?

Obviously, low data quality is a serious problem in a highly automated environment. But that does not mean that data must be 100% accurate for the system to function. After all, today’s enterprises have nowhere near complete data accuracy, but they still manage to plan, execute, and produce products that generate a profit. Further, we need to be clear that 100% accuracy is unrealistic no matter how precise or complete our data collection may be. Perfection is a nice goal, but it is never truly achievable on a mass scale.

So the real question is, how accurate does data need to be in order to make the Fourth Industrial Revolution effective?

Planning vs actual data

The big difference in the Fourth Industrial Revolution, when compared with revolutions in the past such as those brought about by CIM (Computer Integrated Manufacturing) in the 80s and ERP in the 90s, is the emphasis on distributed intelligence among a network of participants, instead of on a centralized plan.

This points to two different levels of data. Thus, when it comes to data quality, I believe we need to take different approaches toward planning and actual data.

Planning data will include all the master data in the planning and modelling systems which may or may not be fully coherent or correct. This data is used by ERP and supply chain planning tools to help forecast future operations or by simulation tools to model logistics and manufacturing dynamics. Clearly, planning and modelling will never be completely accurate, and in fact, plans need to have a degree of flexibility built into them so that enterprises can react when the unexpected occurs—as it always does! As the saying goes, “The best laid plans of mice and men…”

Actual data, on the other hand, is collected during execution from as-built material BOM, as-built operating sequences, actual lead time, actual cost, intelligent sensors data, as well as unstructured data from social media. If the systems are designed correctly, the quality and accuracy of the actual data should be very high, at least on the local level. This means that the real-time by hour and by minute interactions of operations—such as between suppliers and factories, or between warehouse and production—should become more driven by actual data than planning data, and this is the critical factor in the Fourth Industrial Revolution. We see this happening now in some industries; for example, in the automotive industry, “Pull” manufacturing models are widely used, with production data being fed to the warehouse or suppliers to activate “just-in-time” movement of supplies.

However, in order to make sense of the ocean of actual data for real-time decision-making, one must compare them to an existing plan or model. So what does this all mean?

Continuous data improvement

I believe the conclusion is straightforward. Namely, that there is no need to aim for 100% reliable data from the onset but rather to take a pragmatic approach that is robust enough to handle data quality variation. It will be important to have an enterprise architectural strategy to minimize problems in data quality and integrity that come from too many disparate systems. But with that in place, enterprises should be able to take a gradual approach with clear business objectives that reflect the constraints on data quality, while using actual data collected from Fourth Industrial Revolution devices to improve planning data in the long term.

This feedback loop of continuous improvement on data quality sounds simple, but is rarely implemented by enterprises today. For example, standard cost is typically revised only once a year during the budget cycle. Standard lead times or safety stock levels are rarely revised, perhaps once every year. Although new technologies on Big Data analytics and optimization are available these days to quickly extract insights from actual data in a global level, rarely are such insights applied frequently to update planning data. With the Fourth Industrial Revolution, it will be become more imminent to automate this data quality feedback loop.

The right enterprise data strategy to ensure data integrity combined with a pragmatic approach in implementing continuous improvement should be to enable business to benefit from the Fourth Industrial Revolution without 100% reliable data. After all, the success of the Fourth Industrial Revolution does not depend on robots and automation, but rather on closing the gap between the cyber and physical world.


If you liked this article, here are other materials you might also find interesting: