Portfolio metrics
One table, every company, the same metrics, normalised across every board pack format. This is the view the whole extraction pipeline exists to fill.
These controls are switched off The quarter, the view mode and the flagged-only filter all need figures that do not exist yet. They are shown rather than removed because the design expects them, and inert rather than working because switching them on would compare one placeholder against another. Each one says what it is waiting for.
Reporting periods are parsed from filenames and are provisional; some packs have none at all. Phase 2 extracts the real period from the pack contents.
Budget figures do not exist anywhere in the system, so there is nothing to switch to and nothing to compare against. Actual and budget arrive together with the fact table in phase 5.
Flagging a company as off plan needs a budget to be off against. Nothing computes one, so no company can be flagged and the filter would hide every row. Phase 5.
An export would be a spreadsheet of placeholders, which is worse than no export: it leaves the building without the marks that say the figures are not real. Phase 4.
Cross-portfolio comparison
The metrics the workshop judged reliably extractable or derivable.
No companies ingested.
Every figure in these tables is a placeholder, and the API says so itself: no fact has been approved yet, so there is nothing measured to show.
No variance arrows, deliberately. An arrow is a claim about direction, and there is no prior period to compare against.
No currency symbols, deliberately. Figures are stored in each company's own currency, nothing converts between them, and these companies do not share one.
There is no company page for this company. Company Detail is drawn entirely from the invented example set, so it has a page for each of those and none for a company whose board packs are actually here.
Harder-to-extract metrics
Reported less consistently across companies. Treat these with extra care once they are real.
No companies ingested.
Review queue
Extractions below 50% confidence, awaiting human approval.
Nothing is queued, and nothing can be. The queue ranks extracted figures by confidence; no figure has been approved and the confidence column is null on every row. It fills in phase 4.
Counting approved facts