Traceable identifiers
Map product, work order, process and lot. Keep split and merge histories instead of assuming one-to-one relationships.
MANUFACTURING / CAPABILITY
Trace the conditions
around every defect.
A data foundation for investigating quality variation: align quantities, process conditions and inspection results before designing the dashboard.
Explore the delivery scope ↓OPERATING MODELILLUSTRATIVE
- 01Production & inspection
- 02Align lots and timestamps
- 03Compare process conditions
- 04Investigate on the floor
Conceptual data-to-operation flow, adapted to the agreed project scope.
What makes
this workflow difficult.
Daily reports, lot inspections and second-level sensor logs have different grains. Joining them directly can multiply rows and distort defect rates or downtime. Define the mapping from work order to finished lot first.
THE SHORT ANSWER
A data foundation for investigating quality variation: align quantities, process conditions and inspection results before designing the dashboard.
DESIGN & DELIVERY
Design in the order the work happens.
Explicit reporting cutoffs
Separate preliminary and finalized totals, and define the treatment of rework, retests and scrap. Preserve explanations for differences from daily reports.
Investigation-oriented views
Navigate from defect category to lot, conditions and original records. Treat correlation as a lead for investigation, not proof of causation.
DATA READINESS
Required data and the conditions to check.
A consistent format does not ensure operational usefulness. Confirm meaning, refresh and exceptions with owners, and keep missing information explicit.
| Data | Required fields | Checks |
|---|---|---|
| Production | Order, product, quantity and process time | Rework duplicates and lot splits |
| Inspection | Result, defect type and inspected quantity | Different denominators for sampling and full inspection |
| Equipment | Asset ID, state and timestamp | Clock drift, missing values and units |
EVALUATION
Measure value in the workflow.
These are candidate evaluation measures. Establish a baseline and agree on data, period, targets and exclusions before work starts. They are not claimed performance results.
| Measure | Evaluation method |
|---|---|
| Traceability | Traceable lots divided by in-scope lots |
| Reporting effort | Active time from source collection to reconciled report |
| Defect rate | Agreed defect count divided by inspected quantity, with a fixed retest rule |
Replacing MES or control equipment is not a prerequisite. Start with available exports and expand integrations after confirming access conditions.
Define deliverables
and acceptance.
Review data mappings, workflow, prototype and evaluation evidence before choosing production scope. Estimates distinguish assumptions because permissions and interface terms affect implementation.
Select one line or asset group and one goal: quality investigation or anomaly notification.
QUESTIONS BEFORE YOU START
Questions before starting
Can we start without sensor data?
Yes. Start with production and inspection reconciliation. Predictive maintenance needs a separate assessment of sampling and event history.
Do you guarantee a defect reduction?
No fixed reduction is promised. First validate traceability and investigation effort, then evaluate quality interventions through plant trials.
RELATED FIELD GUIDE
MANUFACTURING / GUIDE
Evaluating manufacturing AI: lots, time and false alarms before accuracy
Before commissioning factory AI, define how evaluation data is separated and how operators will review results. This guide explains common pitfalls and a practical evaluation brief.
Read the guide ↗YOUR OPERATION / OUR STARTING POINT
Start with the work you do today.
Tell us about the workflow, systems and reporting or documents that need attention. We will define a suitable phase, deliverables and assumptions.
Manufacturing — discuss your project How scope and estimates work ↗