Daily Decisions Catalog: What data each decision needs

This concise reference maps common shopfloor decisions to the data, role, cadence, and safe automation level required. Use it when you design dashboards, alerts, or AI assistance so the tool supports the actual human decision.

Decision pattern: Stop or continue a running machine

• Data needed: cycle time, part count, vibration/temp (if applicable), quality inspection flags.
• Who decides: operator or automatic stop with supervisor escalation.
• Cadence: immediate (seconds to minutes).
• Automation guardrail: auto-stop allowed for safety-critical thresholds; otherwise alert + confirm by operator.
• Measure success: % of correct stops, false stop rate, time-to-restart.

Decision pattern: Trigger maintenance triage

• Data needed: anomaly score or threshold crossing, recent failure history, spare parts availability.
• Who decides: maintenance supervisor with input from operator; automated ticket creation is acceptable with human triage.
• Cadence: minutes to hours.
• Automation guardrail: create ticket automatically but require human priority assignment; avoid auto-scheduling without capacity check.
• Measure success: reduction in repeat failures, MTTR, percent of tickets with sufficient diagnostics.

Decision pattern: Contain a quality escape

• Data needed: inline quality sensor flags, SPC chart violations, recent change history (tooling, material, operator).
• Who decides: quality engineer + line supervisor; operators perform containment actions per standard work.
• Cadence: immediate to shift-wide (minutes to hours).
• Automation guardrail: alerts for likely defects; do not auto-quarantine without human confirmation unless proven safe.
• Measure success: time to containment, number of escapes prevented, false positive rate.

Decision pattern: Short-term scheduling or sequence adjustment

• Data needed: current WIP, machine availability, due dates, changeover times, constraint machine status.
• Who decides: production planner or supervisor; optimization suggestions can be presented to the human scheduler.
• Cadence: minutes to hours (daily adjustments).
• Automation guardrail: present ranked options with consequence estimates; require human approval for final sequence changes.
• Measure success: on-time delivery, schedule adherence, changeover reduction.

Scaling examples

For a small job shop: focus first on operator-visible KPIs and a single automated alert → manual triage flow. For a midsize plant: map 5–8 repeat daily decisions, instrument the constraint and quality-critical points, and roll out operator dashboards plus escalation rules.

Use this catalog to avoid the common mistake of automating without defining the decision, the data required, who retains authority, and how success is measured.