Pega Platform: Data Foundations for Data Science
Pega Platform
The Pega Platform domain tests whether you can prepare and govern data for analytics inside Pega. Expect questions on case data modeling, data pages, data transforms, data sets, data flows, and how these artifacts support predictive modeling and reporting. The exam is not about Java or SQL tuning; it is about knowing where a feature comes from, when it refreshes, and whether it is available at scoring time.
Core rules
- A data page is a rule that loads a page or page list on demand. Scope (thread, requestor, node) and refresh strategy determine reuse and staleness.
- Data transforms map source values to target properties and run actions in order. They are not a general-purpose join engine.
- Data sets define sources and sinks for data flows; use them for batch extraction and model-data preparation.
- Savable data pages persist changes only when explicitly saved; read-only data pages do not.
- For modeling, validate label definition, feature availability, and timestamp consistency before training.
Worked example
A team wants to predict case escalation. They define a savable data page D_CaseFeatures with requestor scope, sourced from a report definition, and a data transform that copies priority, channel, and open age. Reasoning: requestor scope is wrong for batch scoring across many cases; the page would be reused incorrectly. They switch to a list data page keyed by case ID, refresh always for batch, and use a data flow to write a modeling extract. They also confirm open age is computed as of the snapshot timestamp, not the current date.
Traps
- Treating data transforms as joins; they cannot merge unrelated sources without a page or report.
- Forgetting that data page scope changes reuse and can leak data between cases.
- Using current-date features in training, which creates leakage and serving skew.
- Assuming a data set is a database table; it is an abstraction over a source.