Sampling audiences for simulations
The audience is a set of individuals, that you are targeting with your marketing message. Sampling is the process that allows you to focus on a representative subset of a large population, and therefore you do not have to analyze large groups of subjects. You can also use sampling to reduce the amount of time needed to run a simulation that delivers accurate projections when extrapolated across the entire population.
Pega Customer Decision Hub provides separate inbound and outbound samples of your production customer data that you can use as the audience input for a simulation. Sampled customer data is captured in your Production Environment and migrated to the Business Operations Environment (BOE) by a data migration pipeline. The inbound and outbound sample datasets in the BOE can be refreshed at scheduled intervals. For more information about scheduling data migration, see: Configuring the system for running simulations of decision strategies.
The Pega Customer Decision Hub inbound sampling algorithm randomly samples up to 20% of your inbound decision requests (default sample size is 5%). The final inbound sample includes only one random interaction per subjectID (for example, account, customer). If you use a multilevel customer entity, the sampling algorithm captures all the entity data required to decision the next best action for the sampled request.
For example, if a decision for the head of household requires customer data for all members of the household, then customer data for all members of the household and their devices will be sampled so the interaction can be replayed in a BOE simulation. In addition to customer data, the container payload, other inbound contextual data, and Interaction History for the customer are all captured and migrated to the BOE as a part of the inbound sample.
The Pega Customer Decision Hub outbound sampling algorithm randomly samples up to 20% of your customer data from production (default sample size is 10%). If you use a multilevel customer entity, the sampling algorithm samples at the top level (for example, account) and includes all underlying entity data (for example, customer, device) in the sample. In addition to customer data, the Customer Insights and Interaction History for the customer all captured and migrated to the BOE for the outbound sample. For more information, see Enabling the migration of historical data for Scenario Planner.