How it works
From request to published report
TrueFigure works the way a good reporting specialist works. It asks the questions it needs answered, writes them down, builds the report, and shows you the figures to check.
1. Your data, refreshed nightly
TrueFigure reads the Power BI dataflows you already refresh, straight from your Azure Data Lake, so there is no second pull from Planon or SAP and no extra source cost. Every entity in every dataflow is loaded into your own Postgres database each night, exactly as the source names it.
Calculated columns such as SLA status and PPM status are computed at refresh, as Power BI does. Measures are evaluated when a visual asks for them and cached until the next refresh, so slicers and cross filtering stay fast.
Daily snapshots start from the first sync, so "how many jobs were open at month end" and "is the backlog growing" can be answered later even though the source only shows today.
2. A semantic layer the agent reads
Every table and column has a plain English meaning, every metric a plain English definition and an approved SQL expression, every business term a glossary entry. The agent reads this before it builds anything, and it can only use approved metrics in a published report.
Your specialist approves each entry once. From then on, every report uses the same definition of "open job", "reactive" or "SLA met", and a change to a definition flags every report that depends on it.
3. The agent asks, learns and builds
It asks
"Which date should the month come from?" "Does a cancelled order count towards WIP?" If something is undefined, the agent asks your specialist instead of guessing.
It records the answer
Each answer is saved as a pending definition. Your specialist reviews it on the Knowledge page and approves, edits or rejects it. Approved knowledge is used in every future request.
It builds a spec, not code
The output is a structured report specification: pages, sections, slicers, visuals, drill through, access. Your front end renders it in your branding. The agent never writes DAX and cannot set a colour or font.
4. Check before anyone sees it
- Every draft renders live on real data, side by side with the current published version and a plain English summary of what changed.
- The agent lists its assumptions and the figures to verify. It never claims a number is validated.
- Known good figures from your existing reports are stored as checks and run automatically.
- Specialists can edit the specification by hand, in a visual builder, or by carrying on the conversation. Every change is a revision you can undo.
5. Publish, version, govern
- Publishing creates an immutable version with notes, who published and when. Viewers only ever see the latest.
- Report level access by group, row level access by contract, region or client, enforced in the database.
- Requests from the business flow through a request queue with states from submitted to published.
- Usage, cost and an audit log of every agent call, change, publish and export.
The result
A branded, published report with slicers, cross filtering and drill through, on a phone or a boardroom screen, with "Data as of" in the header so nobody asks how fresh it is.
Attend SLA, last 6 months
By job type
| Type | Jobs | SLA |
|---|---|---|
| Reactive | 241 | 96.5% |
| PPM | 48 | 98.4% |
| PPM remedials | 9 | 88.9% |
| Projects | 4 | n/a |
Watch a report being built
Bring a report request from your backlog. We will build it live on sample data and show you every step above.