decisions/2026-06-28-decision-first-dashboard-v1.md
ADR: Decision-First Dashboard v1
Date: 2026-06-28 Status: Accepted
Context
The first CI dashboard artifact was operationally useful but product-weak. It showed source health, reports, actions, and signals, but Arijit correctly challenged that it felt like a signal dump rather than a command center. The market-truth research showed that paid CI tools already cover monitoring, alerts, dashboards, battlecards, search, Q&A, and enablement workflows. Chowmes CI therefore cannot win by showing more monitored data.
The accepted wedge is an Algolia-specific decision layer: turn competitive movement into evidence-backed, owner-specific actions with quality gates.
Decision
The Algolia CI dashboard v1 will be designed as a decision surface first and an evidence browser second.
The first screen must answer:
- What changed?
- Why does it matter to Algolia?
- Who should act?
- Can Arijit trust the evidence?
Raw signals remain available, but they are the audit trail behind recommendations, not the primary homepage.
Rationale
- The feature matrix showed that generic monitoring and dashboards are table stakes, not differentiation.
- The Algolia relevance map showed that stakeholders care about role-specific implications, not raw alerts.
- The differentiation thesis says the system must compete on decision quality, action routing, and claim-quality governance.
- Baseline-only signals must not generate battlecard recommendations until true deltas are confirmed.
UI Direction
Dashboard v1 should include:
- Decision brief.
- Recommended actions with owner, rationale, controls, and evidence drawer.
- What changed this week.
- Implications for PMM, Sales Enablement, Product, Partnerships, Exec, and Arijit.
- Trust state and quality gates.
- Source health and collector strategy as an operational sidebar.
- Signal ledger as supporting audit trail.
- Report archive as supporting navigation.
Alternatives Considered
| Option | Why rejected |
|---|---|
| Keep signal-table homepage | It repeats the market's generic monitoring pattern and does not answer what action to take. |
| Build a full CI OS interface immediately | Premature. The accepted strategy is a narrow Algolia-specific decision layer first. |
| Make source health the primary view | Source health matters for trust, but it is infrastructure, not the product value. |
Consequences
- The dashboard data model needs first-class decision items, material changes, evidence references, and quality gate state.
- Static v1 can show the intended UX, but the next implementation slice must persist action decisions and compute true deltas from the ledger.
- Vercel should eventually serve a real app, but the current static dashboard seed is enough to validate the information architecture.