Top 5 priorities to move from reactive to predictive quality management

Who has never experienced it? The “fire fighting” when it is already too late. It is also well known in the context of quality management and quality costs for, e.g., recalled products are only one of the possible effects. Another impact will most certainly be the “headache” of the quality manager.

Very often, activities and measures have been defined to move away from reactive quality management. But it has never quite worked out that way.

So is “artificial intelligence” now able to deliver on this promise? Our opinion is: “No”. No, because even artificial intelligence must first learn based on historical data. Hundreds, even more thousands, of data sets will be needed to teach the desired predictive insights to an algorithm. But once this is done, the algorithm can discover and predict results, and here in particular correlations (for example between measurement criteria and external influences), with a very high (and probably unprecedented) precision. With the help of this data, predictions can thus be made, which can then be backed up in actions and with measures before the error escalates.

But what do you need to look out for or what can you prepare today to prepare for the introduction of artificial intelligence with “Predictive Quality” capabilities?

  1. Collect data. As explained earlier, every algorithm needs historical data. The more, the better. This data can be, for example, historical measurement data, or machine data for maintenance, photos of products and their defects. How many data points you require will depend, among other things, on the complexity of the problem. But you will probably never hear the phrase “we have too much data” from a data engineer.
  2. Involve your IT department early on. Data becomes particularly valuable when it can be collected and analyzed beyond the boundaries of the individual department. Your IT will be absolutely necessary for this. In our experience, this and many other reasons speak in favor of involving internal IT as early and in good time as possible.
  3. Democratize quality management. Only if your organization accepts quality management standards, actively reports quality issues and tries to remedy them, will “predictive quality” fall on fertile ground. Therefore, try to convince as many stakeholders as possible of the relevance of quality management and take them along on the journey.
  4. Link insights to actions. Only linking the prediction to actions and measures (or at least to alerts in your system) will condition improvements and thus create real added value.
  5. Determine a method to calculate ROI. If measures in the future lead to, e.g., quality costs being reduced before they are incurred, it will become increasingly difficult to measure the success for the effort expended or to justify the costs. Create a framework and methods in advance of how you will demonstrate ROI today, but also in the future.

We look forward to discussing the topic of predictive quality with you and to show you how our software can support you in this process and what capabilities our CAQ system already offers “out of the box” in this area.

Further information: 
Predictive Quality

How Much Training Data is required for Machine Learning?

Predictive ROI