![](/images/ORS/Volunteering/ProBono/Pro-Bono-case-studies/Bloodwise%20image.jpg)
The Approach
Since the transactional data contained a mix of categorical and continuous variables a two stage approach was used to identify different segments of supporters.
Initially, by means of a clustering algorithm, the volunteer team of analysts grouped all supporters characteristics into a smaller set of components. This has the benefit of reducing the dimensionality in the data.
Finally, once the data had been grouped into a smaller set of variables we applied a hierarchical clustering algorithm. The outputs of this algorithm revealed eight distinct segments.
Once eight segments were identified, the next step was to gain an insight about what makes the segment different from each other. A decision tree provided a set of rules to classify the segments. These rules were coded into an Excel macro so Bloodwise can classify their new supporters into segments.
"We identified quick wins which will allow us to optimise our promotional activities around sporting events" – Insight & Analysis Manager