The direct answer
Mode labels tell the reader whether a number came from production, paper simulation, replay, or demo seed data. Without them, dashboards become marketing instead of operations.
Common failure modes
- A "win rate" that blends paper fills with live trades
- Staging data left connected to a production dashboard
- Demo environments shown to prospects without synthetic data disclaimers
- KPIs refreshed on different schedules but presented as synchronized
A minimal metric contract
For each metric you publish internally or externally, document name, mode, source system, as-of timestamp, and owner. Surface mode and as-of in the UI — not only in internal docs.
UI patterns that work
- Badge next to the metric (e.g., PAPER, LIVE)
- Tooltip with definition and lineage
- Separate tabs or filters per mode — never silent blending
- Demo pages labeled Demonstration — synthetic data
Why this matters for studio clients
When we build data intelligence layers, metric honesty is a deliverable — not a polish task. The same discipline applies to our own products like Parlay Gorilla, where research outputs must stay clearly separated from outcome guarantees.