How to get a grip on data quality in your scheduling

The automation of scheduling promises a number of advantages. I already reported on these in detail in my last blog post. However, reaping all these benefits is not a matter of course. On the one hand, it must be ensured beforehand that a number of important prerequisites are met, and on the other hand, the introduction of automated scheduling – like every process improvement and every software introduction – requires a considered and structured approach. Otherwise, disappointed expectations are inevitable. The design of the introduction process will be the subject of one of my next articles.

In today’s article, however, I would like to focus on what I consider to be the most important prerequisite for automating scheduling and shed light on the question: “Do you have all the data necessary for automated scheduling under control?” If you have to answer this question in the negative, it will be difficult to impossible to actually reap the benefits of automation. An old IT adage says: “Garbage in, garbage out”. To ensure that you are well prepared for the introduction of automated route planning, I will make a plea for data quality today.

Does your data reflect operational reality?

Automation always means that a machine supports humans in processes and decisions. However, the machine can only use the digitally available data. If this data does not reflect the operational reality or is of poor quality, even an automated result will be unsatisfactory. Automation then brings no advantage. On the contrary: if automatically calculated results or plans are followed, even though they are based on an unclean database, this can lead to significantly worse results.

Why is data quality an issue at all; after all, the data is already needed within the manual processes? The answer is as simple as it is obvious: manual processes always make use of a great deal of implicit knowledge on the part of the actors. A procedure along the lines of “What do you mean, the truck is 20 cm too short? That’s nonsense, the trucks are all longer than the data say; it has always worked so far”, a digital process cannot reproduce this. Nor should it. The widespread combination of incorrect data and manual correction by “gut feeling” should generally be replaced by correct data. For this reason, it is very important to give high priority to the quality of one’s own data. Unfortunately, failure to do so often leads to frustration among users and a rejection of new software tools.

The focus is on the following data in particular:

Order and article data

• Do the specified time windows actually correspond to the customer’s preferences or is there implicit or empirical knowledge or are the real preferences possibly different or flexible? (Example: The appointment is in the afternoon, but the customer prefers to be delivered early in the morning).
• Is the article data (capacity requirements, special properties, etc.) correctly available?

Location data

• Are the opening hours known, up-to-date and correct?
• Do we have correct and accurate address or localisation data?

Vehicle data

• Are the vehicle master data and the capacities cleanly recorded?
• Are relevant vehicle characteristics and aids (e.g. transportable forklift) recorded centrally?

It helps to realise that this data is actually needed anyway. With manual decisions, only each decision-maker uses his or her own version of the data – a combination of explicit data and gut feeling. With automated decisions, however, this data must all be available digitally in ONE version. An additional advantage of clear, digitally available data, which should not be underestimated, is that it makes all decisions clearly traceable. This makes subsequent evaluation much clearer. The effort to bring the data to a good quality level is therefore worthwhile anyway.

Bild: iStock (subtik)