Clustering and Supply Chain Analysis

Throughout my life (more so during my summer internship), whenever I have been confronted with large amount of data for analysis, I have turned to Excel. And more often than not, it has served my purpose. But only two days into my Business Analytics class and I have realised the power of SPSS. I am still quite an Excel loyalist, but ‘excelling’ at SPSS does seem to be quite beneficial.

The basic objective of SPSS is that it gives us access to a plethora of decision making tools. Collecting data is one thing, but interpreting data and then developing a strategy based on those interpretations is a completely different challenge.

In the first two classes we have dealt in depth with Cluster analysis. Cluster analysis, as the name suggests involves division of the sample into clusters based on certain common attributes. There are two major types of clustering:

i)Hierarchical Clustering: Building of hierarchies from individual elements by progressively merging clusters. Done when number of items to be clustered is less than 50

ii) K-Means clustering: Assigning each point to a cluster whose centre is the nearest. The centre represents the arithmetic mean of each dimension separately over all the points of the cluster. Done when number of items to be clustered is more than 50.

OK. Enough of the theory. Let me try to apply this cluster analysis to some other concept. As an operations student with a profound interest in supply chain, I will try to apply it to the function of distribution.

Distribution of SKU’s from your finished goods warehouse to your distributors and from distributors to retailers seems to follow some kind of a divisive clustering model. You keep on dividing your clustered finished goods into smaller and smaller clusters as you go down the supply chain.

Divisive Clustering

In a similar way, reverse logistics follows an agglomerative clustering model. Imagine unsold newspapers going back from vendors back to the newspapers manufacturers for recycling. They will be clustered according to their location as they move up the supply chain.

Agglomerative clustering

Again, cluster analysis could be applied to milk run concept. One could create different clusters of distributors based on their location. A single vehicle could go from the warehouse to the distributors clustered locally. This type of clustering could help bring down the transportation cost and hence the supply chain cost. This could be done by k-means or hierarchical methods depending on the number of distributors to be covered.

As I said, SPSS is a decision making tool and hence it could be used to take decisions related to your distributors as well. Just create a distributors profile (by collecting information from retailers) and analyse how your distributors are performing with respect to variables like delivery time to retailers, responsiveness, service levels, facilities available at distributors’ warehouse and so on.

So that’s how I could interpret the applications of clustering and cluster analysis. Hope it helps!!

Posted by:

Anish Diwadkar

Operations 2