To sample at the offer level
Use this method to sample if your offer distribution is skewed.
For more information about working with flowcharts and processes, see the IBM® Campaign User’s Guide.
1.
In your post-optimization flowchart, configure your Extract process.
a.
Retrieve optimized contacts according to package and offer code.
For example, package A contains offers 1, 2, 3 and package B contains offers 1 and 4. You need 5 Extract processes, one each for optimized transactions: A1, A2, A3, B1, and B4.
b.
(Optional) If you must segment by cell for statistically valid hold-out control groups, retrieve optimized contacts from the CellCode field.
2.
(Required only if you performed step 1b.) To resegment back into the original cells to create statistically relevant control groups, add a Segment process to the flowchart and segment by the CellCode field. This method is how you put customer IDs back into their initial groups, for example, High-value and Low-value.
3.
Randomly sample each cell into a target cell and a control cell using the Random Sample option in the Sample process.
4.
Connect all of the target and control cells belonging to a single package into a single contact process.
For more information about channeling multiple cells in to separate contact processes, see To include one Extract process per package.
Copyright IBM Corporation 2013. All Rights Reserved.