Well-designed procedures provide the foundation. Manufacturing excellence depends on how consistently people apply them on the shop floor.
1. Why do gaps emerge between approved procedures and actual shop-floor practices, even at sites with established quality systems? What do these gaps reveal about how work is managed?
A procedure is written at one point in time, but operations keep moving. Equipment is upgraded, batch sizes change, products move between blocks and new shifts begin. When documents do not keep pace, people working under production pressure may start adapting how a task is done.
The reasons are often practical. Procedures grow longer with each CAPA. The written sequence may not match the physical layout. Planned cycle times may not leave enough room for in-process checks and documentation. Informal methods also pass from experienced operators to new ones.
An adaptation can feel practical to the person making it. But an unauthorised departure from an approved procedure may introduce risks that have not been assessed. The difference must be promptly reported through the deviation process, and QA must assess its potential impact on the product. The team then decides whether practice must return to the approved procedure or the procedure needs to be revised. A revised method can be used only after formal risk assessment and approval through change control. Until then, operators follow the approved procedure, with any interim controls agreed by QA.
A gap can also tell us whether procedures are being kept current and whether people feel comfortable saying, “This step cannot be done as written.”
The most useful tool I know is the procedure-versus-practice walk. The area manager and a QA colleague take a frequently used SOP into the area and watch two operators from different shifts perform the task, document in hand. They note the differences and ask what makes each step difficult. The purpose is to understand what is happening on the floor.
When procedures are revised, involve the operators who perform the task. A clear sequence, one action per step and controlled visual aids at the point of use can make the approved method easier to follow across shifts.
2. When the same deviations recur, what can the pattern reveal about operator behaviour, supervisory intervention and the way teams respond to production pressure?
When the same event happens with different people on different shifts, I would look at the task, its controls and the conditions around it before focusing on any individual.
Most sites classify deviations by type. That helps with reporting, but it may not reveal the underlying pattern. For learning, I suggest putting two years of deviations in one sheet and sorting them by shift and hour, week of the month, proximity to changeovers and restarts, experience of the people involved, and the exact process step.
The patterns can be revealing. Documentation errors late in the night shift may point to fatigue or an end-of-shift recording load. Events after changeovers may indicate setup or handover problems. Peaks in the last week of the month may reflect workload.
Consider a packing hall where most line clearance deviations occur during evening changeovers, when one supervisor covers two lines. Before attributing these events to individual carelessness, the investigation should examine supervisory coverage. If overlapping changeovers mean the supervisor cannot be present at both lines, staggering them addresses the operating condition that contributed to the problem.
I would identify critical steps in each area, such as granulation end point, line clearance or filter integrity testing, and plan supervisor presence at those steps across shifts.
A monthly review of recurring deviations by the production and QA heads, with the area owner presenting, helps keep attention on the problem areas. Before closing a repeat investigation, someone should observe the step on two different shifts. “Human error” should not be where an investigation ends; it should prompt the question of what made the error possible. CAPA effectiveness depends on whether the event stops recurring, not simply on the closure date.
3. How can manufacturing leaders create an environment where operators report near misses, process abnormalities and procedural difficulties before they become quality events?
Operators often notice drift first: a compression machine that sounds slightly different, a label roll that feeds unevenly or a step that consistently takes longer than planned. If that observation reaches the right person early, a potential deviation can be addressed sooner.
A simple reporting channel helps. It could be a near-miss card or register at the line, in a language people are comfortable using, or a verbal report recorded by the supervisor. This does not replace the deviation system; it brings issues into it earlier. QA reviews the reports, and anything that may affect the product is handled under site procedures.
What happens after someone raises a concern is just as important. The first few responses from a supervisor can set the tone for the team. Saying, “Thank you, show me,” and going to look at the issue together makes it clear that raising a concern is part of the job.
Handovers are another point where early signals can be caught or lost. Outgoing and incoming supervisors should discuss what happened, what nearly happened and what remains open. A short start-of-shift huddle can bring the same questions to the wider team.
There also needs to be visible follow-through. Acknowledge reports within an agreed time and show what has changed and what is still in progress. If a fix needs approval or budget, explain the timeline and interim control. People need to know their reports have been heard.
Honest mistakes and concerns raised in good faith should lead to learning. Deliberate disregard of procedures must be handled separately.
4. Beyond audit findings and deviation metrics, what observable behaviours indicate that quality ownership has become part of everyday manufacturing rather than a responsibility delegated to the QA function?
Metrics tell us what has already happened. Everyday behaviour can tell us something about how work is being done.
At the line, records should be completed as each step is performed, so the batch record is current when you walk in. Ask an operator why hardness is checked at a particular stage, and the answer should relate to the product, not simply to QA. People should pause and escalate when something looks different. Colleagues reminding each other about gowning or line clearance should be part of normal work. The area should run consistently on a Sunday night and on an inspection day.
Supervisors need to be present at critical steps, and handovers should cover quality alongside output. The production head should present the area’s deviations and CAPAs and own the actions, while QA provides independent oversight and retains authority over disposition decisions. Production should call QA early when in doubt, and QA should spend time on the floor helping resolve problems.
Near-miss reporting and the share of issues first raised by production rather than found during QA review are useful indicators. Neither proves maturity on its own: an increase in reports could reflect greater trust or more problems. They need to be read alongside severity, recurrence and the effectiveness of the response. More reports with falling severity and fewer repeat events is an encouraging combination.
The FDA’s Quality Management Maturity programme has brought this subject onto leadership agendas. I would combine an anonymous survey with structured floor observations and small focus groups from different areas. Each method can help identify issues the others may miss.
5. In sterile and biologics manufacturing, how can human behaviour and adherence to aseptic practices influence contamination control and process reliability? What practical approaches help quality leaders address these risks?
In a cleanroom, people are one of the most significant sources of contamination. The revised EU GMP Annex 1, in force since August 2023, treats personnel as part of the site’s contamination control strategy, alongside facilities, equipment and processes. Even a well-designed line depends on how people move, intervene and respond when something goes wrong.
Intervention discipline is essential. Interventions should be defined, risk-assessed, described in procedures, represented in aseptic process simulation and recorded when they occur. An unplanned intervention must be escalated, documented and assessed under the site’s quality procedures.
Coaching helps build the right practices. Operators need to learn deliberate movement, how to keep their hands out of the path of first air over exposed product and closures, and how to avoid contact with the face or non-sterile surfaces. Pairing new aseptic operators with experienced colleagues before they work independently in Grade A can help. Reviewing recorded interventions together can also be useful where site policy and employee consent allow.
Gowning should be observed during routine production, not only during qualification. Time spent in full sterile gowning, fatigue, planned breaks and task rotation also need attention.
Environmental and personnel monitoring results become more useful when trended by shift, activity and intervention as well as by location. Isolators, RABS and closed systems reduce exposure, but they still require discipline around glove integrity, transfers and aseptic connections. The same attention is needed for open manipulations in biologics.
Operators must feel supported to stop the line, reject a component or report immediately when sterility may have been compromised.
6. As pharmaceutical manufacturing becomes increasingly digital, how can quality leaders ensure that technology strengthens deviation detection, data integrity and timely decision-making while preserving human accountability?
Electronic batch records, MES, LIMS and electronic QMS have reduced manual calculations, transcription and late entries. The principle I apply is simple: the system detects and flags, while a named, trained person reviews, decides and signs. Batch disposition, deviation classification and investigation conclusions remain human decisions supported by good data.
The way systems are configured matters. When eBMR and MES workflows are designed with operators and supervisors, the screen sequence can match the physical flow of work, helping people enter data at the time of the activity.
Alarms need careful management. Too many prompts risk becoming background noise. Every critical alarm should have an owner and a defined response, and acknowledgement patterns should be reviewed regularly. Frequent alarms are a reason to investigate possible process drift, equipment condition or setup. They are not, on their own, a reason to change a limit.
Review by exception works when QA defines the exceptions, the logic is validated and performance is checked periodically. Audit trail review should be risk-based and conducted by trained reviewers. Unique user IDs, no shared accounts, periodic access reviews and segregation of duties remain fundamental.
Trending process parameters, monitoring results and near misses in near real time can help teams detect drift before a limit is breached. Analytical and AI tools can help identify where to look, but people still need to assess what the findings mean. Operators should understand that the electronic record is the record of their work, and that ALCOA+ principles apply just as they do to paper records.
So following are three actions for next week
1. Start an observation: ask the area head and a QA colleague to walk one frequently used SOP against practice on two shifts, and raise every difference through the quality system.
2. Start a review: ask the production and QA heads to review the top five recurring deviations together and observe those steps across shifts. Resulting changes go through approval; those needing people, equipment or budget get an owner and a date.
3. Strengthen a routine: add “What nearly went wrong, and what is still open?” to shift handover and the daily meeting, and make sure every item gets a visible response.
Teams don’t judge leaders by the questions they ask. They judge them by what happens after someone speaks up.
