Building Groups Based on Relations

Grouping is not necessarily a very unusual task - customer groups always come to my mind spontaneously. Common characteristics are often used, such as sales, and often also socio-demographic characteristics.

But relationships can also be the starting point for grouping, for example, to identify influences that may not be directly measurable.?Relationships can also provide interesting insights for analyzing cooperation in teams (watching a soccer match you can sometimes immediately notice that collaboration doesn't really work, in companies it's often a bit more difficult to spot something like that).

The basic principle of this kind of grouping is very simple: if there is a relationship between two elements and one of these elements also has a relationship with another, then they belong to a group.?In the following example you can see very well that in many cases there is only one connection (e.g. between "A" and "B" or "D" and "H"), but that all orange elements are connected at least by "detours". However, there is no connection to "Blue" and "Green" in the example.

With a limited number of elements, it is no problem to recognize something like this, but with an increasing number of elements, it's getting more difficult.

However, if you want to investigate relationships in organizations or even in social networks, for example, it might be too complicated to do it by using this "visual" method.

At this point Alteryx is helping us again.?I have represented the "relationships" from the image in a small table:

??The "starting point" is not always uniform - "A" is, for example, once the sender (call, e-mail), in other cases the recipient. In practice, such a list will probably contain many more entries.

To derive groups from this list, we can use the MAKE GROUP tool. The configuration is very simple - we just specify the relevant fields:

The tool then returns the corresponding classification:

As manually assigned before, the elements A, B, C, D and H end up in a group named according to the first element found (here "A"), F, G and I form the second group, while E is not assigned to any group due to missing relations.

In the example shown here, this is probably not very impressive - it could have been done manually. With a few thousand entries, however, it certainly looks different!

The MAKE GROUP tool will probably not be used in many workflows, but it can save a lot of work for specific applications. And as you know, that is always very important to me ...