How Mercidia builds recommendations
Evidence first, uncertainty visible, and a human decision recorded beside the next move.
Lead with an observation
A Mercidia pointer starts with a timestamped signal from connected records. It identifies what moved, where the evidence came from, and what remains unknown before presenting options for a human decision-maker.
- Source records, freshness, metric completeness, and known exclusions stay attached to the pointer.
- Co-occurring changes are described as correlation, never as proven cause.
- Possible explanations are labeled unconfirmed and paired with evidence that could confirm or reject them.
Separate measured impact from investigation
Only positive measured loss can appear as a numeric opportunity. Heuristic findings and unquantified signals remain useful investigation prompts but do not become measured savings by presentation alone.
- Expected impact is distinct from realized impact.
- Confidence is based on observable inputs such as sample volume and completeness.
- A recommendation carries assumptions, alternatives, tradeoffs, monitoring, and rollback criteria.
Preserve the decision
The operator can approve, modify, defer, dismiss, assign, or review a pointer. The decision, owner, review date, evidence snapshot, and eventual outcome become operating memory instead of disappearing into a meeting thread.
See the evidence loop
Start with the fictional sample workspace, then connect your own sources when you are ready. Mercidia keeps observed values, assumptions, uncertainty, and decisions visible together.