IDs, abbreviations and the combinations of both often featuring additional characters such as the underscore are almost exclusively understandable for data-oriented users. However, the bulk of users have a very different focus than the data itself. Therefore, it is important to translate the technical terms into terms that are familiar to the user.

Few self-service users know or want to know the relationships between database/data lake artefacts. Therefore, a semantic layer must map these and provide them in such a way that no matter what the user does, the correct relationships are always used.

The report creator should also not make his or her own calculations since the process itself can become complex on the one hand and the semantic layer should act as a “single point of truth” on the other hand. There are development procedures that temporarily allow such self-calculated numbers. In the medium term, however, these must then be incorporated into the model in a manner that ensures quality.

Just as important as basic and advanced metrics is the so-called time intelligence. An absolute number alone does not say much at first. But if you put it into context, e.g. “same period last year”, you can do more with it. Other forms of time intelligence are year to date, quarter to date or month to date values. Delta values between current and historical values are also often reported as performance indicators.

Ever tried to make a report and kept waiting for the results to show? The frustration rises very quickly. Therefore, it is important that the response times are in the lowest second range (<3 sec.). With all the possibilities that modern reporting offers, the reporter also wants to try things out to get the best possible result.