What to do with the results
Having identified the most likely people to go to Sweden, we can use this information either to:
- Predict new prospects for use in a Marketing Campaign.
- Describe existing customers to gain a better understanding of the customer base.
Marketing costs¶
If you are using the Decision Tree to help plan for a Marketing Campaign, it may be useful to enter some Marketing Costs. These can be entered to help plan how many customers to target in a campaign and therefore which nodes to base a campaign selection on. See How do I set the Marketing Costs?
| Marketing costs settings | Calculation |
|---|---|
.png) |
The settings shown mean that a campaign will cost: - £1 for every person targeted. - An additional fixed £1000. - Of the responders, 1% will become customers. - Each of these will bring £750 revenue. |
These figures are applied to the analysis and base counts and the profitability figures can be seen in the Tree Grid. Right Click on the column headers and use the Column Chooser to make the columns below visible.
.png)
The calculations are explained more fully in How do I set the Marketing Costs?
Note, that the Decision Tree at present does not use a holdout sample. As a result, the analysis count is only an indication of the likely response rate if the selection rules to be applied to a different sample.
Having looked at the above table, you might decide to target the top 6 nodes as highlighted. This would still be profitable (and also makes the rest of the example work!)
Creating a selection for a marketing campaign¶
Having selected a number of nodes, you can drag them off from any of the tree views to create a FastStats Selection. The selections below were created by selecting the top 6 nodes using the Gains Chart, and then dragging these off from the Box Tree.
| Gains chart | Box tree |
|---|---|
.png) |
.png) |
We can see that highlighting the 6th node on the gains chart has highlighted a number of nodes on the Box tree view. If we drag off from one of these nodes we get the selection below:
.png)
The highlighted nodes can all be traced back to node 1, hence we get a simplified query containing the rules for node 1.
Understanding current customers¶
The individual Node Rules, or the resulting Selection Query, provide insight into the different types of customer who go to Sweden.
The above tree would suggest that:
- The most likely people to go to Sweden have an income of over £20k - £80k.
-
If we select Node 19, the 'best' node, we can see that the most likely records to go to Sweden have an income between £30k and £50k and 6 distinct occupation types.
.png)
These characteristics could be used to identify people on your existing customer database who have the characteristics of those that have already been on holiday to Sweden, but have not yet been themselves. You could then run a cross-selling campaign.