Sets allow developers to divide a dimension, on the basis of its own content or that of another column, into subsets and create new perspectives as a result. This allows isolation of data segments and consequent generation of new insights.

There are three different types of sets:

A data record can be found inside or outside the defined condition or allocation, i.e. a binary view is involved here.

The binary view and different types of sets can be illustrated by means of examples:

• Threshold - customer file at a store

A customer is to be considered regular as soon as they start buying there more than 5 times a month (threshold). A customer can therefore be either inside the segment of regular customers (> 5 purchases per month) or outside (< 5 purchases per month).

• Top-ranking of sales outlets

A company’s headquarters wants to ascertain the top 3 outlets based on sales. This classification is performed without any fixed turnover threshold.

• Manual selection - favourite customers

A craftsman decides to send a discount to his favourite customers. The group of favourite customers can only be created through manual assignment by a data-savvy developer.

Categorizations which cannot be represented by binary logic are not possible with sets.

Data preparation is now concluded and we know that sets can be used to segment data on the basis of dimensions. But how to create a set, and what is the difference between fixed and dynamic sets?

Fixed sets contain several dimensions and are simply selected in a diagram. They are therefore particularly suitable for ad-hoc analyses. Once selected, the elements of the set are fixed and cannot be changed. However, the resulting field can be reused, e.g. for filters or calculations.

Dynamic sets are created by right-clicking the dimension to be segmented. An editing window opens with three options:

• General - a defined set can also be created here; selected entries can be included or excluded. Any new data records are always added to the "miscellaneous group".
• Condition - setting of a threshold for another dimension.
• Topmost - determination of the top or bottom N entries after evaluation of a further dimension.

The condition and topmost functions both take new data records into account and categorize these according to the conditions specified. In addition, parameters can make thresholds and selection of N entries more interactive.