Decision Tree: How do I set the Marketing Costs?

The Marketing Costs are set using the control shown below.

Example

To explain how the costs are applied, consider the following example of using a Decision Tree to plan a marketing campaign:

  • A Decision Tree has been built, based on an analysis selection consisting of all the people who responded to a previous campaign.

  • A selection of the best nodes in the Decision Tree is made.  These nodes have been chosen to give a high Analysis % (people in the analysis selection, i.e. responders).  The rules behind this selection will be used to choose the people who will be targeted.

    For example, the selected top nodes may have a base count of 100,000 and an analysis count of 20,000, giving an Analysis % of 20%.

  • In practice, these selection rules would be applied to another database, the prospect database. Figures based on this prospect database (e.g. Analysis %) would be slightly different to those seen during the modelling process, since new data is being used. However, if the model is good, it should transfer to different data.

Since only one set of data is available (that used to build the model), figures from the actual build process will be used as an approximation for what would happen when using a prospect database. The example figures above are used below in the profitability calculations.

  • The number of people who would be targeted is represented by the base count (100,000).

    At a Cost per record selected of £1, this would cost £100,000).

    Assuming a Fixed Cost for the campaign of £1000...the total cost of the campaign = £101,000 (100,000 x 1 + £1000).

  • The proportion of people who would respond is represented by the Analysis % (20%).

    This makes the big assumption that the people in this campaign behave exactly the same as those people who were used to build the Decision Tree.

    This would give us 20,000 responders (20% x 100,000).

  • The proportion of responders who actually bring in revenue is set by the Conversion Rate.

    A conversion rate of 1.5% would give us 300 new customers (1.5% x 20,000)

  • The amount of revenue we received depends on the Revenue per response [customer].

    A Revenue per response of £500 would give us a total revenue of £150,000 (300 x £500).

Definition of Inputs

The Marketing Cost inputs are as follows:

  • Fixed cost

    This is the cost forecast to occur in initiating the marketing activity independently of the volume of data mailed. For example, this cost might include the costs of copy writing and leaflet design.

  • Cost per record selected

    This is the cost forecast to occur for each record despatched. This cost would typically include the data licensing cost, the letter print costs, the mailing cost, etc.

  • Revenue per Response [Customer]

    This indicates the forecast revenue that can be expected for each new customer (successfully converted respondent).

  • Conversion  Rate

    This is the number of people who actually become customers as a proportion of those who are forecast to respond to the campaign.

Definition of Calculations

The Calculated measures are as follows:

  • Cost = Base Count x Cost per record + Fixed Cost

  • Revenue = Analysis Count x Conversion Rate x Revenue per response

  • Profit = Revenue – Cost

  • Revenue ROI = Revenue / Cost

  • Profit ROI = Profit / Cost

The Conversion Rate relates to how many of the people in the analysis count will become revenue generating responders/customers.  In deciding what rate to use, consider the following:

  • In the above example, the analysis selection is a group of responders to a previous campaign.  In this case it is reasonable to assume that a subsequent campaign will have a response rate similar to the Analysis % seen in the Decision Tree.  So the Conversion Rate, just relates to how many of the responders will go on to become customers.

  • In other cases, the analysis selection may not be based on a previous campaign.  For example, it may be people who went on holiday to Sweden and so are thought likely to be good candidates for a campaign promoting Scandinavia.  In this case the response rate is likely to be considerably less than the Analysis %.  The Conversion Rate needs to factor in two things now: how likely the Swedish holiday makers are to respond to the campaign and furthermore how likely those that do respond are to become customers.