A model-data dashboard pulls information out of models and project systems, stores it in a simple table or database, and shows it as charts and figures that answer specific questions. Good dashboards answer questions people actually ask every week: are the models getting better, which issues are overdue, how many assets have complete data, how much concrete has been placed. Bad ones show 40 charts nobody reads.
At a glance
| Audience | Project directors, BIM managers, coordinators, clients, estates teams |
| Data sources | Model exports (IFC, schedules), CDE and issue systems (via APIs), clash results, programme, asset data |
| Pipeline | Extract → clean and shape (one row per thing, consistent names) → store (spreadsheet, database, data model) → visualise |
| Tools | Business-intelligence tools (Power BI, Looker Studio, Tableau and similar), spreadsheets, custom web pages |
| Refresh | Daily or weekly, automated where possible |
Start from the question
| Question | Measure | Source |
|---|---|---|
| Are models getting healthier? | Warnings, proxies, unnamed spaces per model per week | Model checks or IFC script |
| Are we closing issues? | Open issues by age and owner; opened vs closed per week | Issue log or BCF |
| Is the asset data ready for handover? | % of assets with required properties, by system | IFC or COBie check against the AIR |
| Is construction on track? | Planned vs installed elements by zone | 4D programme and progress records |
| Is the design within budget? | Key quantities by revision (façade area, concrete volume) | Quantity extracts per revision |
One question, one chart, one owner who acts on it.
Shape the data before charting it
- One row per thing (per element, per issue, per check), with columns for its attributes.
- Consistent names: "Level 03", "L03" and "3rd floor" become one value.
- Keep history: store each week's extract with a date, so trends are possible.
- Document definitions: what counts as "overdue", what "complete" means.
Worked example
On a large project, the BIM manager reports "model quality" monthly from screenshots. Instead, a script runs every Sunday on each discipline's latest IFC, counting proxies, elements with no classification, and spaces with no number, and appends the counts to a table with the date. A three-chart dashboard shows the trend per discipline. In the first month, the structural model's unclassified elements rise from 2% to 18%: a new modeller used the wrong template. It's fixed in a week, not discovered at the next data drop.
Common mistakes
- Building charts before agreeing the questions.
- Manual data refreshes that stop when the person who did them is busy.
- Comparing numbers whose definitions changed between weeks.
- Dashboards that nobody owns and nobody acts on.
Where to read more
Business-intelligence vendors publish tutorials on data modelling and visualisation. Vendor APIs (CDEs, issue trackers, APS) document how to extract project data. Guidance on information management performance measures appears in ISO 19650 national guidance.