Data quality: Protecting the Achilles heel of digitization

The Achilles' heel for efficient digitization is not technology. The decisive weak point is the data. If one too many zeros creep into the digitized process or important data is missing, common sense no longer has a chance to intervene. This weakness can only be eliminated with a stringent fitness program for data. Here are the ten most important questions about data quality and how to ensure it:

1) Can data quality be measured and evaluated reasonably at all?

Experts describe up to 15 theoretical dimensions of data quality. In practice, the situation is simpler: Automated process steps must first and foremost work with complete, up-to-date and, above all, unique data. This is because duplicates in the parts master or in the customer data weaken efficiency. If, for example, two data records exist for a customer, this can lead to the customer being treated as a C-customer, although it actually belongs to the B-customers.

Not all information is equally important. Different spellings of an address do not have the same effect on process fitness as a missing list of conditions for a vendor. Therefore, measuring data quality always includes an evaluation of the errors found. Particularly critical cases also require an escalation workflow to eliminate errors promptly.

2) Is it worth the effort? Isn't everything just as chaotic again after a short time as it was before a cleanup?

Anyone who has already tried to get a grip on data quality with several projects will feel like Sisyphus in the Greek myth. No sooner have you reached the top than the stone rolls back down the hill and the effort starts all over again. In fact, experience shows that the effect of a project-by-project cleanup fizzles out after a while. As with running training, you have to keep at it and implement a data quality program.

3) Where is the best place to start?

Ideally, a company starts where better data delivers the fastest added value. This can be in purchasing, because supplier addresses, conditions and replenishment times noticeably accelerate operational procurement. A start in production and logistics can also help to maintain parts masters cleanly. Parts are then fully assigned to their groups and all necessary weight data is available for shipping. Depending on the industry and company, sales and service can also benefit particularly strongly if address and contract data are up-to-date.

4) Can a data quality program be established without analysis gurus or Excel specialists?

Today, modern analysis programs do not require any programming at all. Defining the rules is not witchcraft for a user who is somewhat familiar with the system. Once the rules have been defined, the employees in the specialist department receive an indication of which data needs to be corrected. Ideally, they can click directly through to the data record concerned. This saves time. Short-term corrections also ensure a rapid learning effect. The frequency of errors is thus reduced quite naturally.

5) How quickly can such regulations be adapted to new requirements?

Today, changes can be made at very short notice. This is because modern analysis tools require neither programming by a software manufacturer nor an IT expert. However, companies must ensure that rule changes do not contradict each other or lead to problems elsewhere. Without data governance, this is not possible.

6) Do all data need to be in one system for ongoing data review and cleanup?

Anyone who postulates this is thinking outside the box. The vast majority of companies today work with more than one system. Today, inspection software, so-called data quality managers, effortlessly integrate data from several sources and inspect them together.

7) How do you get a handle on the issue internationally?

There is no way around master data management. Clear responsibilities are the be-all and end-all: who is responsible for which data, who may and who must change which data - and where, in which system. Master data management regulates which data is maintained centrally and which locally, and ensures the necessary synchronization.

8) How can progress be reliably measured and documented?

The reporting of regular analysis runs must not be limited to the identification of individual errors. It must also enable controlling in the sense of a "state of the data". This status report shows the specialist department as well as the management that data cleansing is worthwhile and that the efforts are bearing fruit - and sustainably so. Healthy competition can even develop among the departments.

9) How does a continuous data quality program work?

Data Quality Manager software checks the previously extracted parts masters, customer files or other data against a set of rules. For example, zip codes in Great Britain are alphanumeric, whereas in Germany or Austria they consist only of digits. Checks against external databases, such as those that check zip codes and streets for plausibility, are also possible. In addition to pure error detection, the software categorizes whether the errors are serious or have a minor impact. The errors found are then transferred to the target system along with an evaluation. In most cases, this is an ERP system. There, the employees can then directly clean up the data. If an exception is detected, it is noted in the rules and regulations. All this can be done today without an employee or a consultant having to program a line of code.

10) How often should data sets be audited?

There are no fixed guidelines for this. The frequency depends very much on the particular company, its processes and data sets. Like any fitness program, it is tailored to individual goals and performance parameters. The key is to continuously and regularly check and measure progress.

Most companies are now aware of their Achilles' heel and are prepared to actively do something about data quality. Those that have already started report a double training effect: On the one hand, data quality management ensures greater production and process reliability internally - and thus for well-founded decisions. In addition, reliable statements about delivery dates and availability increase customer and supplier satisfaction and accelerate collaboration.

Source: proALPHA