Criteria View

Back to Introduction to Campaign Optimisation

Optimisation Criteria

Optimisation Criteria button

The Optimisation Criteria sub-view is where you specify how the “best” campaign should be chosen when an individual record is eligible for multiple campaigns.

Campaign optimisation interrogates the initial audience of all the campaigns to be optimised and, depending on the priority or the model rules defined, allocates individuals who could be in multiple campaigns to just a single campaign.

There are two types to select from:

Campaign Level Priority (standard)

The Campaign Level Priority (standard) allows you to rank the campaigns, with 1 being the most important. You would typically use this approach when there is a business logic to the importance of each campaign, as illustrated below:

To apply the Campaign Level Priority score:

  • Use the Optimise column to select the campaigns you want to optimise
  • Use the Priority column to set the priority order

Individual Level Score (advanced)

The Individual Level Score (advanced) is used in cases where the importance of a campaign depends on who the recipient is. In this case a variable is used to score the suitability of each campaign for each individual. For example, the relevance of the campaigns below depends on an individual’s preferences. PWE* models could be made to score how likely each person is to go to each event, and these are used to specify the criteria for the optimisation. Note that the audience selections for each campaign would probably express non-event specific criteria. The optimisation task then decides which campaign individuals receive.

Constraint Options

Start typing in the Variable cell to perform a search or use the drop-down button

To apply an Individual Level Score:

  • Use the Optimise column to select the campaigns you want to optimise
  • Use the Variable column drop-down menus to select the model variable

Score Values

Select a Score Values option to determine if a high or low score is best

In this example, the values are centiles and 01 would be the best - (Lowest is best). Alternatively, if the scores were some sort of predicted response revenue, Highest is best would be the appropriate choice.

Score Variables

  • Select a Score Variables option

This reflects whether the raw values held within each of the variables can be directly compared when evaluating if “lowest/highest is best”. Centile scales are comparable, since each variable is on a 00 to 99 scale. However, if some of the variables were from a Decision Tree*, they are likely to be on a different scale (for example 0 to 40 if there are 40 nodes). In this case, the “scales are different” option would standardise all the variables based on their mean and standard deviation values.

The initial audience can contain multiple selections comprised of includes and excludes but must flow into a single Message step.

You can use any non-text variable as a score. For selector variables, the code is used and converted to a number. Where codes can’t be converted, they are set to be the worst value (high/low depending on whether lowest/highest is best).

FastStats Selections

The optimisation process creates FastStats selections which contain the reserved records for each optimised campaign. You can use these to evaluate the effectiveness of a run.

The selections are automatically saved into the Files explorer, located in Public:/PeopleStage/_Campaign_Optimisation. By default, selections are available in your FastStats system for 7 days before being archived.

This folder is automatically hidden in PeopleStage and cannot be accessed through the PeopleStage Library.


Concept Link IconSee Also

Added in Q2 2019