What is the usecase behind clustering?

[Bernd Welter]

Hello together,

this posting is just to give you a quick intro into the usecases of xCluster:
Clustering is a strategic usecase that has impact on long term decisions. Imagine a set of several hundred customers located in your area of responsibility.
Now quite often you assign these customers to specific sets whether it is a postcode region (quite simple because the customers address contains the cluster ID) or a weekday... Check the attached image:

sample with 3 clusters

it demonstrates the assignment of a given number of locations to the clusters 1,2,3.
Besides the ID each location has two important properties (in the context of clustering):

• the coordinates
• the quantity (e.g. the revenue, the service period or some other measurement...)

An optimal assignment takes care of the following target properties:

• a cluster is compact: cluster areas shouldn't overlap from a geographical point of view.
• the clusters are balanced: the sum of the quantities of a clusters elements is almost equal in all the clusters, e.g. because you want to have equal total revenue in all areas

This is the easiest approach to clustering. The other ones are more complicated and the visualization isn't that simple. I'll try to describe it anyway - imagine the following conditions:

• You have a list of visits to be scheduled over a given period (e.g. 24 weeks) aka horizon
• For each visit you have some additional constrains such as
  • on which weekday is it possible to visit this customer"?
  • is there a specific visiting pattern (periodical visit each 4 weeks...)

The extended usecase will then treat each working day within the given horizon as a cluster and take care that the sum of quantity within the working days is balanced considering the given constraints.

So much for now,
I hope this short intro opens your mind to understand the benefits provided by xCluster,
Regards Bernd