# Example scenarios

## Fixed time scenario

### It is December 2019 and you want to identify people for a promotion to run in January 2020

To support a campaign that is promoting United States holidays, it would be useful to create a model predicting who will travel to the United States in the next month (Fixed Date of January 1<sup>st</sup> for everyone).

- The training date could be set, for example, to Jan 1<sup>st</sup> 2017, to *learn* from people who went to the United States in the following month and, hopefully, identify characteristics in their preceding behaviour that they had in common.
- The scoring date could then be set, for example to Jan 1<sup>st</sup> 2020, to *predict* which people to target, by scoring them based on how closely their recent behaviour matches the characteristics (behavioural features) of the people in the training data.
- Define one or more Evaluation Dates in between, to *assess* the success of the model.

## Event driven scenario

### Who is likely to buy a specific type of insurance cover within 3 months of making a holiday booking?

For this campaign we need to study peoples behaviours around their booking date (Event) as opposed to on one date that is the same for everyone.

- The training date could be set, for example, to Jan 1<sup>st</sup> 2020, to *Identify* people who made a booking on any date in the previous 6 months and then *learn* from their behaviours in the period up to three months after their booking date.
- The scoring date could then be set, for example, to Jan 1<sup>st</sup> 2021, to *predict* which people to target, by scoring them based on how closely their recent behaviour matches the characteristics (behavioural features) of the people in the training data.
- Define one or more Evaluation Dates in between, to *assess* the success of the model.