When a manager asks why a number changed, a diagram of connected systems is rarely enough. The team needs to identify the source records, the processing version and the filters used for that specific result. Commission traceability around those investigations.
Related service: CRM, ERP and web application development- 01Capture the report context.
- 02Identify the dataset and refresh run.
- 03Find the transformation definition.
- 04Locate contributing source records.
- 05Explain the result or open a correction.
SECTION 01
Keep enough context to reproduce the question
Record which report version, filters, timezone and freshness state produced the number. Then identify the prepared dataset and transformation responsible. A generic link to the source application's home page leaves the investigator to reconstruct everything manually. A focused source reference should carry the relevant context without exposing credentials or private information in the URL.
W3C PROV describes provenance concepts such as entities, activities and responsible agents. OpenLineage describes metadata associated with datasets and processing runs. These references help structure evidence, but neither makes an incorrect source record true. Keep origin, transformation and quality checks as related but separate questions.
SECTION 02
Illustrative example: a dispatch count changes
Imagine a dashboard's dispatch total rises after a refresh. The drill-down shows the contributing shipment identifiers and the report cutoff. An investigation reveals that a source correction moved one shipment to the selected day. The operator can identify the correction and refresh run instead of guessing whether the chart's formula changed.
A different reader has permission to see only their region. Their investigation route must not reveal other regions' raw shipment records merely because the shared dashboard uses a common data pipeline. This hypothetical example combines useful traceability with the access boundary that a real implementation must preserve.
SECTION 03
Start with the measures people need to explain
Full record-level history for every transformation can be expensive to retain and operate. Prioritize the important outputs and the questions staff repeatedly investigate. Agree which evidence is stored, how long it remains useful and what happens when the source system later removes a record. Avoid implying that an old result can always be reconstructed if its required evidence is not retained.
Acceptance should take a sample number through the complete investigation route and reproduce its agreed calculation. Test a failed refresh, a changed transformation version and a restricted user. Handoff needs ownership of the definitions and investigation steps. A smaller, dependable trail for important decisions is more useful than a decorative architecture map no one can use to answer a concrete discrepancy.
SECTION 04
Frequently asked questions
Does lineage prove a number is accurate?
No. It explains origin and processing. Source quality and calculation correctness still need their own checks.
Must every reader access raw records?
No. Provide only the investigation detail that the reader is authorized to see, with an escalation route where necessary.
What is a useful acceptance exercise?
Choose a sample result, preserve its filters and cutoff, trace its source records and reproduce the calculation through the documented path.
PRIMARY REFERENCES
Sources and further reading
These references cover the standards, platforms, or published prices discussed in the guide. Worked examples and checklists are our editorial guidance.
- Provenance overviewW3C
- Dataset and run facetsOpenLineage
EDITORIAL METHOD
About this guide
We use AI to assist with drafting and editing. Catapult AI Work is responsible for the published content. Examples illustrate possible approaches; they are not client case studies unless identified as such.
Budget examples are not Catapult package prices. Check linked provider pages for current fees and plan limits before making a purchase.
Read the editorial policy