Manufacturing leaders often know where losses occur but lack the visibility to intervene before scrap, downtime, or performance losses accumulate. A National Institute of Standards and Technology (NIST) paper identifies delays between machining and quality assessment as a barrier to feeding results back into production.
A 2020 NIST report estimated $119.1 billion in annual U.S. discrete-manufacturing losses from preventable maintenance issues, using 2016 values. Downtime, defects, and lost sales make equipment reliability an output and revenue concern.
NIST’s statistical handbook distinguishes numerical measurements from pass/fail classifications. Measurements let teams track deterioration before more parts fail.
Emerj’s Marilie Fouché hosted Sebastian Dykas, Director of Manufacturing Engineering and Maintenance at Smith+Nephew, on the AI in Business Podcast to discuss improving manufacturing performance with real-time operational insight.
This article examines three approaches for manufacturing leaders seeking earlier intervention in production losses:
- Measurement feedback to prevent scrap earlier: Use a three-step framework to capture dimensions and adjustments, identify wear patterns, and trigger corrections before more material becomes scrap.
- Connected production and maintenance records for faster diagnosis: Compare disconnected and connected workflows to identify where equipment information can accelerate investigation of output shortfalls.
- Process controls or redesign to reduce persistent scrap: Determine whether tighter controls can stabilize the process or whether redesigning the operation can eliminate a recurring source of variability.
Listen to the full episode below:
Episode: Turning Manual Manufacturing Variability into Data-Driven Control – with Sebastian Dykas of Smith+Nephew
Guest: Sebastian Dykas, Director of Manufacturing Engineering and Maintenance at Smith+Nephew
Expertise: Manufacturing Engineering, Operations Leadership, Lean Manufacturing, Equipment and Process Validation
Brief Recognition: Sebastian Dykas serves as Director of Manufacturing Engineering and Maintenance at Smith+Nephew. He previously managed maintenance and freeze-dry engineering teams at Pfizer and held engineering, project management, and production leadership roles at Stryker. His experience also includes automated manufacturing processes for orthopedic implants and robotic welding research at Chrysler. He holds an M.S. and a B.S. in Mechanical Engineering from Oakland University.
Measurement Feedback to Prevent Scrap Earlier
Dykas describes manufacturers checking whether a part passes inspection without retaining its actual dimensions. A go/no-go gauge provides an acceptance result, but does not show how a feature moves toward a specification limit across successive parts. That missing history limits the opportunity to correct drift before more material becomes scrap.
His example is a drilled diameter that changes as the tool wears. An operator may adjust the machine to compensate, but engineers cannot examine the wear pattern if neither the dimensions nor those adjustments are retained.
In Dykas’s illustrative example, software identifies a tool-change interval of 80 pieces, and a control prevents further use until the drill is replaced to reduce errors.
His example provides a three-step framework for processes in which measurements can inform machine adjustments or tool replacement:
- Capture dimensions and interventions: Retain the measured feature and the machine adjustments made during the run, associating them with the product and machine so engineers can examine the same operation over time.
- Identify the pattern of deterioration: Compare successive measurements with the adjustments already made to determine when compensation or a tool change is needed.
- Return the decision to production: Feed measurements to the machine for adjustment or introduce a tool-replacement control before further parts fall outside specification.
The financial purpose is to reduce material lost to defective parts by moving corrective action earlier in the production run. The 80-piece interval illustrates that mechanism; each process needs its own measurement history to establish an appropriate intervention.
Dykas also describes machinists and inspectors measuring parts with several instruments, recording readings on paper, and eventually scanning the records. Scanning preserves the paperwork, but the measurements are not necessarily available for trend analysis or machine feedback.
In his proposed workflow, an operator measures a part, transmits the reading, and the machine adjusts while producing the next part. He says wireless measuring devices, software for automatic dimensional compensation, and integrated inspection equipment already support versions of this workflow.
For leaders assessing readiness, the question is whether the next measured part can influence the next production decision. Dykas’s checks are specific: does the machine know which product it is making, when the next part begins, and whether inspection readings are captured? Connecting those records supplies the history needed to identify drift and act before it creates additional scrap.
Connected Production and Maintenance Records for Faster Diagnosis
The challenge of delayed quality reporting becomes clear in Dykas’s example of a scrap event that develops over several days. Data reaches an analytical platform after final inspection, a new scrap category appears, and engineers work backward to identify the cause. He advocates monitoring individual processes, machines, and parts so teams can intervene closer to the point at which the problem develops.
In a second example, Dykas describes how an algorithm or AI could flag that an operator’s output is behind schedule during a shift. Someone investigates, finds faulty equipment, and enters a maintenance work order in a separate system that does not communicate with production.
These two examples show how inspection information arrives after defects occur, while equipment information remains disconnected from an output shortfall. Dykas’s proposed integration gives leaders a way to examine those handoffs in their own plants:
- Record the loss or shortfall: Inspection identifies scrap, or production tracking shows output behind schedule.
- Investigate after the signal appears: Engineers trace the scrap backward, or someone visits the operation to discover faulty equipment.
- Keep the findings in separate systems: Maintenance records do not communicate with production, leaving teams to reconstruct the relationship between equipment condition and output.
In place of this reactive process, Dykas recommends a connected workflow that links production, quality, and maintenance information earlier in the investigation:
- Capture information during production: Monitor the relevant machine, process, part, and output while work is running.
- Bring production and maintenance information together: Connect equipment alerts and work orders with production records so teams can examine the same interruption.
- Trigger investigation sooner: Use an output or equipment alert to direct a person to the operation, then use the findings to establish preventive controls or investigate process changes.
This comparison helps leaders locate delays by checking when a problem is detected, where its equipment history is recorded, and whether those investigating the shortfall can access both sources of information. Earlier investigation can limit the amount of defective output produced before corrective action; connected maintenance information can also help teams understand equipment-related interruptions.
Beyond diagnosis, Dykas suggests using connected production, equipment, and maintenance records to adjust preventive maintenance schedules based on actual machine usage rather than fixed intervals.
For implementation, Dykas recommends appropriate software partners and internal capability. Engineers, quality teams, and operations leaders can learn to build applications for their own environments. He cautions that commissioning a universal digital architecture for the entire manufacturing process can be costly and time-consuming.
Dykas explains why collecting data in individual applications is not enough:
“They can become very quick at creating the tools that can be very specific for certain types of manufacturing environments that will give you the data you need. The difficulty of that is: how do you integrate all those things together to understand what’s actually going on in the whole plant? How do you integrate the CMMS? How do you integrate all this other software that you have so you can get the whole story?”
— Sebastian Dykas, Director of Manufacturing Engineering and Maintenance at Smith+Nephew
Process Controls or Redesign to Reduce Persistent Scrap
Manual finishing means an operator performs the final work on a part’s surface, such as using an abrasive belt after machining. Dykas describes results that vary by operator or day. When the operation itself depends on that craftsmanship, capturing more data may expose the variation without making the process repeatable.
Dykas recommends as a rule of thumb: improve measurement and process controls where variability can be managed, and investigate process redesign where variability remains inherent to the operation.
He illustrates the process-redesign path with five-axis milling equipment that may produce the part while eliminating subsequent manual finishing. The investment changes the production sequence, removing a source of variation rather than only monitoring it.
In his hypothetical $4 million equipment case, the proposed replacement eliminates a manual belting operation. Leaders can structure the comparison around the costs and workflow changes he identifies:
- Scrap: Identify the losses associated with the manual step and which scrap categories the replacement process could remove.
- Staffing dependence: Include attrition and reliance on operators with the required craftsmanship.
- Maintenance and consumables: Compare the belts and maintenance supplies used by the existing operation with the requirements of the alternative.
- Production steps: Identify which operations disappear; his example reduces a seven-step process to five.
These considerations give leaders a specific capital-investment question: could removing the operation reduce its associated waste and recurring costs enough to justify the replacement equipment? Dykas’s example illustrates the investment logic, rather than disclosing an actual purchase or achieved return.
He assigns different parts of that investigation to different teams. Continuous-improvement colleagues stabilize and improve current processes. Sustainment teams examine how to monitor stable machine-based operations. Advanced-manufacturing engineers work with operations and equipment builders to investigate alternative ways to make parts that generate persistent scrap.
Dykas recommends pursuing both known sources of production losses and opportunities where digitization is easier to implement. Digitization supplies information about current production; equipment investigation tests whether a different process can remove a known loss. This separates two investment decisions: improve the information and controls for an operation that can be stabilized, or investigate equipment that removes a persistently variable step.
Dykas says manufacturers often know their largest scrap problems. Investigation must establish whether the process can change, what replacement equipment would require, and whether the company is willing to invest.
He calls for leaders to protect that investigation time. Engineers occupied entirely with daily production problems cannot examine where control is missing or work with suppliers on alternatives. Giving them time and autonomy allows the team to develop a process plan while operators continue meeting production requirements.
Dykas recommends working with equipment builders and industry partners whose experience extends across medical devices, aerospace, and other manufacturing environments. Their role is to examine the existing operation and help determine whether digital controls or a different production process can address its losses.