Good or bad? - Assessing data quality correctly

Many companies have neglected their data quality in the past, for various reasons. The fact that their ERP systems performed only suboptimally as a result was something they were unaware of or accepted. With the digitization of processes and the rapid rise of artificial intelligence (AI), this is becoming an even greater risk. Because there is ample scientific evidence: Poor data quality affects efficiency and results, even with AI systems such as self-learning machines. Anyone who has not yet taken action should therefore do something.

Seven tips for better data quality

ERP manufacturer proALPHA advises subjecting data to a seven-stage performance test - and that includes the company's own inventories as well as information from other sources, such as credit reporting agencies and other third-party providers. The decisive performance parameters are:

1. CompletelyMissing information is more than just annoying. The more digitalized processes become, the more important it is that all the necessary data is available. If, for example, information on the components for a product is incomplete, the production process may come to a standstill or the end product may not meet the specifications. To ensure the performance of business processes, it is advisable to define mandatory fields and perform automated checks at selected process points. Nevertheless, companies should not fall into a "data collection frenzy". Since the General Data Protection Regulation (GDPR) came into force, the principle of economy has applied to personal data: only as much information may be stored as is actually necessary. Data that is no longer required must be permanently deleted. (Incidentally, this also applies to Swiss companies that have data from the EU area, editor's note).
2. CurrentNon-synchronized address and contact data is a prime example of outdated information. Precisely because sales employees only sporadically visit headquarters, the customer database was never up to date in the past. Mobile CRM solutions provide a remedy here. They also keep production-critical information such as warehouse data up to date. Nevertheless, not every status needs to be available in real time. Instead, companies should check their processes to see where faster data provision can increase productivity, and start there.
3. Consistent: Data records in different systems must not contradict each other. In practice, double data storage in several data silos and manual transfer "only" lead to additional work for data entry in the best case. In the worst case, errors occur, for example when transferring order data to quality assurance software. Inconsistencies resulting from this are relatively easy to get to grips with today, thanks to modern integration techniques.
4. Conform: Data must meet the requirements of the systems and the processes, for example, it must be in the appropriate, preferably standardized format. Date and currency formats are classics here. In the case of time stamps, it is also important to ensure that the respective time zone is recorded in addition to the hours and minutes. After all, there is a twelve-hour time difference between 8 a.m. in Shanghai and 8 a.m. in São Paulo.
5. Exactly: Data must be accurate. More precisely, they must be sufficiently accurate. After all, not every business process requires high-precision data down to the xth decimal place. Here, too, companies should first ask themselves: How precise do measured values and other data need to be? The required accuracy should then also be monitored on the system side by means of appropriate rules and data checks.
6. UniqueDuplicates not only unnecessarily inflate the database. They also lead to unnecessary queries. If they remain undetected, misinterpretations quickly occur. For example, if a supplier has several supplier numbers and thus key figures such as the contract volume for discount negotiations are not aggregated. Filtering out redundant data from an inventory is already possible with the on-board tools of a spreadsheet program. However, duplicates keep creeping in and the work starts all over again. An automated data quality manager offers a more sustainable way to clean data.
7. CorrectThis criterion refers not only to topicality but also to another essential aspect, namely the accuracy and reliability of the data. The current discussion about fake news underscores this once again: "cutting-edge" information about a supplier's economic difficulties does not necessarily have to be correct. The sources from which companies obtain business-relevant information must therefore be traceable and credible.

Sustainable quality management required

The relationship of companies to their data is often ambivalent: In some places, for example in sales or finance, its quality is sometimes openly mistrusted, while in other places, for example in production, the quality is often overestimated. Sustainable quality management for data helps in three ways: It prevents costly errors, increases confidence in the company's own data, and enables better decisions. Above all, however, properly managed data helps to digitize processes. To achieve this plus in data-driven efficiency, companies can learn a thing or two from top teams in sports: It takes profound and honest analysis, a clear strategy, and individual commitment to continuous improvement, at all levels of the team.

Source: proALPHA