Raw data vs views: keep the evidence, shape the answer

Learn how source records, imported tables and task-specific views differ, with a worked example based on a GitHub issue review.

Three layers with different jobs

A source record is what the original system knows about an issue, order or document. Keeping a record close to its original form gives you something to revisit when a definition changes or a summary looks wrong. Check the record’s coverage and freshness before drawing conclusions.

An imported table makes choices about what one row represents, which fields to keep and how to convert their types. A view applies a second set of choices for a specific task. A database view typically keeps the query definition. A materialised view stores the query results and needs refreshing.

Work backwards from a planning question

A team wants to know which open GitHub issues to discuss at weekly planning. Choose one row per issue and include the issue number, title, source link and last update. Set a review cutoff, such as fourteen days without an update.

The result is a queue for human review. An old issue might be waiting intentionally or have active discussion elsewhere. Record when the source was queried, which repository was covered and whether every relevant page was retrieved.

Write down the definition and test it

Record the purpose, row grain, filters and time basis. Distinguish when the source changed from when you retrieved it. Decide how to handle missing timestamps, duplicates and issues whose status changed since the previous review.

Test the rule with cases that should qualify, cases that should be excluded and cases with incomplete data. In an OrgSDK project you can turn a verified transformation into a reusable action and keep the evidence behind each run.

Keep coverage and access visible

A list of GitHub issues has limited coverage. State which records you used and keep links so the reader can verify conclusions. Choose access permissions independently of the columns displayed in a view.

Explore three GitHub repository review questions ↗