# Open, refresh and verify

## Current status

The native report definitions pass Microsoft's PBIR validator. The prepared data and relationships pass independent checks. Desktop refresh, DAX runtime results and native screenshots remain unverified because Power BI Desktop is not installed here.

## Open locally

Use a current Power BI Desktop version with PBIP and enhanced PBIR support. If necessary, enable the relevant preview options under File → Options and settings → Options → Preview features, then restart Desktop. Open `powerbi/RetailPerformance.pbip`.

The `DataFolder` parameter initially points to this workspace's prepared data. After extracting elsewhere, close Desktop and run the adjacent `Set-DataFolder.ps1`, or update that parameter in Power Query. It must end with a directory separator and point to the included gzip CSV files. Do not rename the PBIP file to PBIX.

Select Refresh. Review any data-source privacy prompt for the included local files. No account, credential or cloud publication is required to review a local report. Do not publish to a Power BI workspace until its audience is chosen.

## Verify before sharing

- Compare cards and totals with `outputs/metrics.json` and the case study. Source-level precision is preserved; display rounding is expected.
- Change the Month and second slicer, clear selections, and confirm filters compose as intended. Retail reference slices are in `outputs/filter_reference.csv`; demand monthly references are in `outputs/results.csv`.
- Confirm titles, labels, card values, table columns and annotations fit at the normal Fit-to-page view. Use keyboard navigation as well as the mouse.
- Confirm that selecting a chart does not make a percentage denominator misleading. Clear visual selections before comparing full-period totals.
- Save as PBIX from Desktop only after successful refresh and review, if a compiled single-file deliverable is needed.

## Refresh scope

This is a fixed historical snapshot, not a live connection. A Python rerun regenerates the prepared data from the audited historical sources; Desktop Refresh imports that snapshot. To add later data, validate the source contract, extend dimensions, rerun the notebook, and reassess model evaluation. Do not treat another Desktop Refresh of the same snapshot as new evidence.

## Report maintenance

Native source files are created by `powerbi/author.py`. Edit the generator for durable layout changes; rerunning it overwrites generated PBIR and model definitions. Data preparation lives in `powerbi/prepare.py`. Current measure expressions are documented in MEASURES.md.