Yield, scrap, rework and OEE are defined locally, which makes network comparison a reconciliation exercise.
Operational and manufacturing analytics for medical device production, built around the constraint that separates this industry from every other manufacturing environment.
In most factories, analytics identifies an optimal parameter setting and somebody changes it that week. In a validated medical device process, the same finding starts a conversation about change control, requalification, documentation and the risk assessment that justifies touching a process somebody already proved. The analysis is identical and the economics are not, which means the insights worth pursuing are the ones whose value exceeds the cost of acting on them.
In general manufacturing, a process drifting outside its usual range is an efficiency signal somebody investigates informally. In a validated medical device process it is a departure from an approved state, which means a nonconformance, a documented investigation, a disposition decision and potentially a CAPA. That changes the economics of detection in both directions: finding drift early is more valuable because it avoids the record, and surfacing marginal signals is more costly because each one carries process overhead.
The second problem is that the data needed to understand a deviation lives across MES, ERP, QMS, equipment historians and paper, and reconciling it is usually manual.
Process change carries a qualification cost
Part number, work order, serial, asset tag and complaint reference all describe the same unit and none of them join.
A deviation generates a record, not just a conversation
PLM, MES, ERP, QMS, CRM and service platforms each do their job well and were never intended to interoperate.
Equipment data and quality data live apart
Which means they are asked rarely, answered partially and never repeated the same way twice.
Metrics mean different things per site
Production metrics hide quality losses
Output measured before inspection and rework lets a line appear productive while consuming capacity on units that will be reworked or scrapped.
The build record is the start of every field investigation
Years later somebody needs what the line recorded, in a form nobody designed for that purpose.
A substantial share of manufacturing improvement in a validated environment requires no process change at all: scheduling, material flow, changeover, maintenance timing, operator support and detecting drift before it becomes a deviation. Those deliver value without triggering change control, and they are the right place to start while the harder parameter work is scoped properly.
Step 1
Separate improvements requiring a validated process change from those that do not, since the second group is where early value sits.
Â
Step 2
Define KPI rules for production completion, unit acceptance, first pass yield, rework, OEE downtime, ideal cycle time, scrap valuation, product mix and historical definitions.
Step 3
Join equipment data to quality outcomes, which is the correlation most device manufacturers cannot currently make.
Â
Step 4
Establish lot and serial traceability through the process, covered on our data platform page and required by every investigation.
Â
Step 5
Detect drift before it becomes a deviation, since avoiding the record is worth more than explaining it.
Â
Step 6
Model the cost of acting, so prioritization accounts for qualification rather than only for benefit.
Â
Step 7
Keep the device history record intact, because the analytical copy is not the controlled record.
Â
Step 8
Design for the investigation, so manufacturing context is retrievable years later without a reconstruction.
Â
Step 9
Respect the boundary with process control, since analytics reads from the line and does not act on it.
Â
Step 10
Validate the analytics where output supports a quality decision, to the position your quality function defines.
These capabilities apply standard manufacturing analytics within a validated environment, where the constraint on acting shapes what is worth building.
Measure
Metric Definition and Harmonization
Yield, scrap, rework, OEE, first pass yield and throughput defined once so network comparison is possible rather than contested.
Production Performance Analytics
Output, cycle time, changeover, downtime and utilization by line, shift, product and period, with the definitions stated.
Quality Cost Analytics
Scrap, rework, investigation effort and disposition time, which is the cost of quality most organizations under-measure.
Equipment Performance Analytics
Availability, performance and quality contribution per asset, joined to maintenance history and process parameters.
Correlate and Warn
Process to Outcome Correlation
Linking equipment parameters, environmental conditions and material lots to inspection results and field performance.
Drift and Early Warning Detection
Identifying parameters trending toward limits before a specification is exceeded, which avoids the record rather than explaining it.
Predictive Maintenance Foundations
Where equipment data supports it, with honest assessment of whether the failure history is sufficient to predict anything.
Root Cause Support
Assembling the manufacturing context behind a nonconformance so investigation starts with evidence rather than with data gathering.
Trace and Govern
Lot and Serial Traceability
Component lot through build to finished unit, which every investigation needs and which frequently stops at the work order.
Device History Support
Manufacturing context retrievable for a specific unit years later, without reconstructing it from several systems.
Site and Network Comparison
Performance across plants on harmonized definitions, which is the comparison most multi-site manufacturers cannot currently make.
Analytics Governance
Validation where output supports a quality decision, with definitions versioned since a change restates history.
What CaliberFocus does, and does not do?
Manufacturing analytics identifies meaningful losses, helps teams investigate causes and verifies whether improvements worked. We prioritize actionable improvements that avoid validated process changes while larger initiatives are scoped. The platform reads manufacturing data without affecting line control. Process changes, disposition, validation strategy and deviation criteria remain with your manufacturing and quality teams.
The third column is the honest constraint on acting, and it should drive sequencing. The analyses at the top deliver value in weeks. The ones at the bottom are worth doing and need a change control conversation before anybody starts.
| Analysis | What It Shows | What Acting on It Requires |
|---|---|---|
| Downtime and Changeover | Where line time is lost. | Scheduling and practice changes without process requalification. |
| Scrap and Rework Cost | Quality failure costs by cause. | Improved visibility; most causes can be addressed through existing mechanisms. |
| Maintenance Timing | Whether maintenance is early, late or reactive. | Maintenance schedule changes, generally outside process validation. |
| Material and Supplier Correlation | Whether a lot or supplier correlates with outcomes. | Supplier action, covered on our Document AI and Quality pages. |
| Drift and Early Warning | Parameters trending toward limits. | Intervention within the validated range, offering high value with a low barrier. |
| Equipment Parameter Optimization | Whether alternative settings could improve performance. | Change control and requalification, with real value and cost. |
| Process Design Change | Potential improvements to the process itself. | Full change assessment, validation and documentation. |
Final yield reported as first pass yield is the most common way rework disappears from a manufacturing report. The unit of measurement has to be explicit too, since unit-level first pass yield and process-step first pass yield answer different questions and are routinely quoted interchangeably.
Manufacturing data comes from systems built for execution, control and record keeping rather than for analysis. The third column is what each one demands before it is usable, and underestimating this is why manufacturing analytics programmes take longer than planned.
| Source | What It Holds | What It Demands |
|---|---|---|
| MES | Work orders, operations and build genealogy. | The strongest structured source and the anchor for other systems. |
| Equipment Historians | High-frequency process parameters. | Volume handling, time alignment and linking tags to work orders. |
| QMS | Nonconformances, deviations, CAPA and investigations. | Controlled access and links from records to the relevant build. |
| ERP | Materials, lots, receipts, orders and costs. | Lot genealogy traced to serial numbers, not just work orders. |
| Inspection Systems | Measurements, images and pass/fail results. | Association with individual units, covered on our Vision Inspection page. |
| Paper and Hybrid Records | Records that have not been digitized. | Extraction through our Document AI page and clear reporting of coverage gaps. |
Why Time Alignment Matters
Equipment historians, MES, inspection systems and quality records often use different timestamps. Without a common time basis, correlating process parameters with production events can produce misleading results. Reliable time alignment is essential for trustworthy manufacturing analytics.
A manufacturing data platform holds a copy of information that exists in controlled form elsewhere. That copy is useful for analysis and is not the record, and the distinction matters when somebody asks what was actually built. Keeping the governed original authoritative while the analytical layer stays fast and flexible is the architecture that works, and conflating them creates either an unusable platform or an unreliable record.
Record boundaries
- The controlled device history record remains authoritative, with the analytical layer explicitly a derived copy
- Analysis supporting a quality decision traceable back to the controlled source rather than resting on the copy
- Engineering controls capable of supporting the frameworks that apply, which may include FDA regulation, ISO 13485, GxP expectations and electronic record requirements
- Applicability, interpretation and validation strategy determined by your quality and regulatory functions rather than by us
Traceability
- Lot to serial genealogy maintained, since it determines investigation and field action scope
- Equipment, operator, shift and environmental context associated with the build where recorded
- Historical process state retrievable, because the investigation asks what the line was doing then
- Lineage from any reported figure back to the source records behind it
Governance
- Metric definitions owned and versioned, since a changed definition restates history with no production change
- Site-level variation in definitions surfaced rather than silently averaged in network reporting
- Change to analytical logic governed where output supports a quality decision
- A named owner for the metric set, distinct from whoever built the pipelines
Security
- Industrial network boundaries respected, with data egress rather than platform access into the control environment
- Process parameters and yield data treated as commercially sensitive intellectual property
- Access governed by purpose, since quality investigation and operational reporting are different grants
- Supplier and material cost information protected to the relevant agreements
Yield is calculated differently depending on whether rework counts as a pass, whether scrap at first operation is included, whether engineering units are excluded and where the measurement boundary sits. Every site answered those questions reasonably and differently, and network reporting then compares numbers that are not comparable. The reconciliation is usually more valuable than the dashboard.
Does rework count as a pass
The single largest source of yield definition variance between sites.
Where is the measurement boundary
First pass at an operation, at a stage, or at final, which produce very different figures.
How are engineering and validation units treated
Included, excluded or separately reported, and sites differ.
What counts as downtime
Planned, unplanned, changeover and minor stoppages are categorized differently everywhere.
How is a nonconformance distinguished from a scrap event
Since one generates a record and the other may not, and the boundary varies.
What period and what product mix
Because a favourable mix explains more apparent performance difference than process capability does.
Manufacturing analytics is usually reported as dashboards delivered and data sources connected. The outcomes that matter in a regulated manufacturer are whether deviations were avoided, whether the cost of quality is visible, and whether sites can finally be compared.
| Category | What We Measure | Why It Matters |
|---|---|---|
| Explain Where the Line Loses Capacity | For one production line, whether you can report scheduled and produced units, first pass acceptance, rework, scrap, lost production time, planned and unplanned downtime, ideal and actual cycle times, bottlenecks, recurring equipment interruptions, material lots linked to failures, capacity consumed by rework, and reconcile production counts with quality dispositions. | If this requires several departments and spreadsheet reconciliation, the opportunity lies in improving that reconciliation. |
| Deviations Avoided | Drift detected and corrected before a specification limit is exceeded. | Preventing a deviation is more valuable than explaining it afterward. |
| Cost of Quality Visibility | Investigation, disposition and CAPA effort quantified alongside scrap. | These costs often exceed material costs and are rarely measured. |
| Definition Convergence | Whether sites calculate core metrics consistently. | Consistent definitions are essential for meaningful network comparisons. |
| Investigation Start Time | Time from a nonconformance to assembling the relevant manufacturing context. | Shows where investigation effort is spent. |
| Actionability Ratio | Insights acted on compared with total insights delivered. | Unaddressed insights erode confidence in the programme. |
We will reconcile how your sites define the core metrics, assess whether equipment and quality data can be correlated reliably including time alignment, quantify the cost of quality beyond recorded scrap, and separate improvements you can act on now from those requiring a change control conversation.
One conversation reconciling how your sites actually measure, and a written definition, correlation and actionability assessment.
Security & Compliance
