Marketing Segments and Segment Trees > Creating and Editing Segment Trees >

Splitting Segments or Branches


Use the Split/Sample option to split any segment or tree branch into smaller groups. As you split or sample, the resulting branch is always the same size or smaller than the parent branch. Table 17 shows the available sampling methods.

Table 17. Sampling Methods for Splitting Segments
Sampling Method
Description

Random selection

The Random method selects a purely random subset of members based on the percentage or count you add for this method. For random selection, you are not required to select a field.

Nth selection

The Nth method splits the branch into N number of equally sized groups. The selection is based on the sequence with which the records are returned from the database. For example, if you select N = 4, the branch would be split into 4 equally sized groups representing 25% of the original group. For this method, you are not required to select a field.

Ranked

The Ranked method sorts the members of the branch by a field you select and lets you choose the top or bottom portion of the ranked list. For example, if you Rank by Total Customer Revenue, you can select the top 10% of the customers with the highest revenue.

N-tiles

The N-tile method sorts the members by a field you select and then splits the list into equal sized groups. For example, you may create quartets or deciles based on Total Customer Revenue.

Values

Similar to instructions for creating filters in segment criteria blocks. For more information about adding a filter using an aggregate field, see Adding Criteria Blocks to Segments.

Mixing Random Splits with Other Splits

You can create segments that mix random splits with other splits. However, this affects how targets are assigned to each split. Random splits exclude targets only when those targets are in immediately adjacent random spits. This point is best illustrated by some examples.

Example A: Three Random Splits

1 - Random Split 10%
2 - Random Split 20% (This split excludes targets that are in split 1.)
3 - Random Split 30% (This split excludes targets that are in splits 1 and 2.)
Effect: All three splits contain different targets, no target exists in more than one split.

Example B: Mixed Random and Nth Splits with Possible Recurrence of Targets

1 - Random Split 10%
2 - Nth Split [N=10]
3 - Random Split 20% (This split does not exclude any targets.)
4 - Random Split 30% (This split excludes targets that are in split 3, but does not exclude targets that are in other splits.)
Effect: Because Split 4, a random split, follows another random split, it does not contain any targets that are in split 3. Because there is a nonrandom split after split 1, targets from split 1 could appear in splits 3 and 4.

Example C: Mixed Random and Nth Splits with No Recurrence of Targets in Random Split

1- Nth Split [N=10]
2- Random Split 10%
3- Random Split 20% (This split excludes targets that are in split 2.)
4- Random Split 30% (This split excludes targets that are in splits 2 and 3.)
Effect: Because the random splits are adjacent to each other, splits 3 and 4 exclude any targets that are in previously adjacent random splits. However, targets in split 1, an Nth split, may recur in any of the random splits.

When random split cells are separated, they are evaluated independently and do not stay mutually exclusive.

To select a field for splitting or sampling criteria

  1. Click the action button for a tree branch to split or sample.
  2. In the menu, choose Split/Sample.
  3. In the Apply Split/Sample dialog box, select a sampling method to use.

    NOTE:  The contents of the dialog box changes when you select different sampling methods.

    The available subject areas and fields appear in the selection pane. The list of subject areas is based on the target level for the segment tree.

  4. In the revised Apply Split/Sample dialog box, complete the fields.
  5. Click the subject area that contains the desired field.
  6. Click the folders to expand the subject area.
  7. Click the field name.

To create random samples

  1. Enter the number of groups to create (1 to N).
  2. Select whether to define the group sizes by Percentage or Count.
  3. Enter a percentage or count for each group, and click OK.

    The sum of the percentages cannot exceed 100%.

To create Nth samples

  1. Enter the number of groups to create (N).
  2. Click OK.

To create ranked samples

  1. Select the field to rank by from a subject area in the selection pane.
  2. Enter the number of groups to create (1 to N).
  3. Select whether to define the group sizes by Percentage or Count.
  4. Enter a percentage or count for each group, and click OK.

    The sum of the percentages cannot exceed 100%.

To create N-tile samples

  1. Select the field to use for the N-tile from a subject area in the selection pane.
  2. Enter the number of groups to create (1 to N).
  3. Select whether the top group has the lowest or highest values for the selected field, and click OK.

Using the Values Sampling Method

Instructions for creating Values samples uses instructions that are similar to creating filters in segment criteria block. For more information about adding a filter using an aggregate field, see Adding Criteria Blocks to Segments.

 

Contents

Copyright © 2013, Oracle. All rights reserved.