Value Finder requirements for data sets
If your organization is currently still on a Pega Customer Decision Hub version earlier than 8.6 that does not include Value Finder, you can import a data set with strategy results and customer attributes to an 8.6 environment and analyze the data there by running a Value Finder simulation test.
Ensure that your data set meets the import requirements so that Value Finder can analyze the data. For example, the data set must contain all strategy results and appropriate customer attributes that Value Finder can use to describe under-served customers. The data set must also contain the data for at least the eligibility and arbitration stages, as well as the propensities from adaptive models.
Since the structure of a strategy framework can vary and propensity values are not always available at the eligibility stage, you might have to create a new strategy to obtain the necessary eligibility and arbitration stage results from your specific strategy framework, and to attach the necessary propensities.
When creating a data set for Value Finder analysis, follow these guidelines:
- Run a simulation and save the strategy results for all customers. Then, join these strategy results with customer data. Use a data flow to merge these sources and to ensure that there is at least one entry for every customer in the original audience.
- Do not mix inbound and outbound results as propensity ranges can vary.
- If applicable, include multiple records for the same customer ID. There may be
multiple records for the same customer ID because:
- The Next-Best-Action Designer strategy consists of different stages: eligibility, applicability, suitability, and arbitration. The data set must include a pyStage column that contains at least the Eligibility and Arbitration values. The other stages are optional. If a stage is empty, this indicates that all actions were filtered out because there are no strategy results for that customer after eligibility.
- A customer can have one or more strategy results (actions or treatments) at each of these stages.
- Include a pyPropensity column that indicates the propensity for each of the strategy results obtained from the adaptive model.
Include the following fields in the data set:
- pyStage – In this column, include at least two values: Eligibility and Arbitration. You can also leave it empty when there are no actions for a customer after eligibility.
- pyPropensity – In this column, include the propensities as obtained from the adaptive model.
- Customer attributes:
- Include any fields that are useful in describing groups of under-served customers, for example, age, gender, or current product holdings.
- You can exclude customer attributes such as names, addresses, or other uniquely identifiable data. A Value Finder analysis does not require this data.
If you use the following fields in the prioritization, you can include them in the data set. Value Finder does not use these fields for analysis, but they can be useful for further insights:
Tabular data example
The following table shows a sample data set which Value Finder can analyze to identify gaps in customer engagement. A data set must include all customers, so that Value Finder knows how large the original population is and can identify how many customers are left without actions after eligibility.
If the decision strategy returns a result for a customer at every stage of the decision strategy funnel, include all the results in the table. If a customer does not pass the eligibility stage, you can leave the result for the stage empty, so that Value Finder knows how many customers are left without actions after eligibility.
There are multiple records for some customer IDs. For example, there are three records for customer 1, which represent three strategy results. This customer is eligible for two actions (OfferX and OfferY), one of which was selected for the customer at the arbitration stage (OfferX).