4 Step Process:
- We understand who you have been mailing through retrospective analysis of your mailing campaigns
- We produce counts (illustrated as buckets) against the different combinations of variables. A customer may feature in one or more of these ‘buckets’
- We look at responders to see which ‘bucket’ has the highest concentration of these customers. We also do the same for data that did not include any responders.
- Summary: this gives us insight into which variables produce responders and which variables produce non-responders.
1. We look at historic mailings to see what variables are included within (there could be thousands) These are the ingredients that will modelled against the people within each bucket. These will give us insight into the names that will have the highest propensity to purchase
2. We then look at including these variables and filling up ‘buckets’ of responders who match those variable
3. By running the previous data against the customers held in each ‘bucket’ we are able to identify the best buckets – and therefore those responders who are most likely to purchase
4. Through the process of elimination we are able to identify not only the best buckets of data, but also how each ‘bucket’ responded to each variable. For example, in the diagram on the left, bucket including variables GD and M included the data with the highest responders, but how did other buckets respond to variable M?
These are representative of different buckets of data with different variables in to represent which we can check response against to see what the key markers are for the most responsive ingredients