When the pallets don’t all fit in the truck or the truck leaves the yard only half full, the corresponding master data is often not properly maintained. Error-free records are indispensable for smooth processes, and that goes for logistics as well.
But how should companies deal with data quality problems for which no precise rules can be defined? An innovative approach to this well-known problem involves the use of artificial intelligence, or machine learning.
Machine learning: generating knowledge from data
Machine learning is a subdomain of the popular research field of artificial intelligence. Machine learning algorithms are capable of discovering patterns and regularities based on existing data. These insights can be used to generate solutions – or new knowledge, so to speak – from previously unknown data.
An algorithm is initially configured by being ‘trained’ with sample data. The data should be free of errors and available in sufficient quantities. Once the training is complete, the algorithm can search for the identified pattern in new data and propose a corresponding solution.
Initial Data Analysis: How Good Is Your Data Really?
When machine learning is applied to the problem of potentially erroneous logistical data, it is possible to conduct what is known as a data plausibility check, which does not rely on set rules. For example, an algorithm is trained to check whether the combination of length, width, and height is plausible, in much the same way as a human would check the dimensions.
From Wrong to Right: A Future Without Faulty Data Entry
Before an algorithm is trained, statistical procedures are used to identify faulty products and correct them manually. This prevents the algorithm from learning incorrect data.
The corrected record provides a basis for the algorithm, which independently finds rules and dependencies, for example, the relation between length, width, and height and volume or weight.
All new product entries are then checked using this algorithm. If all the existing rules are fulfilled, there are no concerns about the new product. If a rule is not fulfilled, the user is warned to recheck the values he or she entered.
Improving Quality in the Long Term
You, too, can benefit from using innovative machine learning technology in your SAP system. We would be happy to support you in the process, from initial testing of data quality to preparation by training the machine learning model and integration in your SAP MDG system. The quality of the data will improve in the long term as a result, because incorrect entries will be prevented. To this end, we offer regular updates of the algorithm, so that new product categories can be analyzed as well.