Process Improvement: Methods, Pitfalls & Real Wins

What is process improvement?

Process improvement is the systematic discipline of making an operational process faster, cheaper, safer or higher in quality through structured observation, analysis and change. In manufacturing it covers everything from a fifteen-minute Kaizen event on a single station to a two-year Six Sigma programme across multiple plants. It is also one of the most oversold topics in the market. Every consultancy owns a methodology, every methodology claims to be the definitive answer, and every plant I have walked into over 25+ years has already tried at least two of them.

As a Six Sigma Black Belt who spent three years on the DMAIC side of the Johnson Controls headliner lines, and then a decade running global improvement programmes across four continents, the pattern I see is brutally consistent. The methodology matters far less than people think. The baseline measurement matters far more. Plants that spend energy arguing about Lean versus Six Sigma are usually plants that have neither a clean baseline nor a closed feedback loop, and without those, no methodology produces sustained results.

The methodology landscape, honestly

Five methodologies show up in almost every improvement conversation. The textbook framing presents them as competing choices. The practitioner reality is that they are complementary tools, and the right question is never "which one do we use" but "which one fits this specific problem".

In a typical automotive tier-one plant that I know well, all five run in parallel. Kaizen drives daily-shift improvements at station level. Six Sigma takes on the dimensional stability projects where statistical analysis is required. Lean shapes the material flow. TPM governs maintenance cycles. TQM is the governance layer above all of it. Choosing between them is not the job. Knowing which to reach for is.

The common engine: PDCA and DMAIC

Strip the branding and every improvement methodology runs the same loop. Deming's PDCA (Plan, Do, Check, Act) and Six Sigma's DMAIC (Define, Measure, Analyze, Improve, Control) are the two most common formulations. In practice they describe the same five activities in different language.

Observe the process as it actually runs · Measure its current performance · Analyze what causes the gap between current and target · Change a small number of variables · Verify that the change held and standardise it.

Everything else in an improvement methodology is scaffolding around this loop. Lean adds a vocabulary for waste types. Six Sigma adds statistical rigour to the measurement and analysis steps. Kaizen reduces the cycle time and pushes it down to the operator level. TPM applies the loop specifically to equipment failure modes. TQM scales the loop organisationally.

The practical consequence: if your improvement programme is not yet running this loop cleanly on a single real problem, arguing about which brand of methodology to adopt is premature.

Why most process improvement initiatives stall

Having audited improvement programmes on four continents, and having watched my own programmes succeed and fail, the failure modes are depressingly repeatable. Five patterns cover the majority of stalled initiatives.

1. The baseline is wrong. The measurement of "before" is based on operator tallies, end-of-shift estimates or ERP confirmations. After the improvement, the "after" measurement is taken differently, usually more rigorously because management is watching. The 12 % improvement on paper is partly method change, partly real change, and nobody can tell them apart. This is the core thesis of my book: a low but honest baseline is worth more than a perfect one that lies.

2. The project picks the wrong problem. Classic symptom: the improvement team attacks the visible queue or the loudest complaint instead of the constraint. Six months of effort, no throughput change, because the real bottleneck was somewhere else. This is also the case for improvements on a non-bottleneck station, which by definition cannot improve system throughput.

3. The Control phase is skipped. DMAIC has five letters for a reason. The Control phase, which standardises the gain and builds the feedback mechanism, is the least glamorous and the most frequently abandoned. Six months later the process has drifted back, nobody noticed, and the lesson learned is "improvement does not stick in our plant". It stuck. Nobody watched it.

4. The cycle time is too long. A classical Six Sigma project with a four-month timeline is a luxury. On a shopfloor where conditions change every week, the answer comes too late to act on. Kaizen-style daily loops on top of a measured baseline usually outperform one big DMAIC project, because they compound.

5. Data lives in different systems. Quality defect data sits in one database, availability data in another, order status in ERP, operator feedback on paper. The analysis that would reveal the root cause requires joining them, and nobody has the time. The improvement team works with what is reachable, not with what is relevant.

Practical rule: the first measurable sign that an improvement programme is working is not an OEE increase. It is a drop. When automatic measurement replaces operator estimates, the honest number is almost always 5 to 20 percentage points below the previously reported number. That drop is the baseline from which real improvement can start. Plants that fire the MES vendor at this point are plants that will never improve anything.

What actually works: focused improvement on a measured baseline

The improvement programmes I have seen deliver sustained, verifiable results share four characteristics. None of them are methodology-specific.

Real-time baseline. OEE, availability, scrap rate, cycle time, captured automatically at machine level, continuously. Not shift-end, not weekly. The baseline is updated every cycle.

Problem selection based on Pareto from data. The top three causes of loss, in quantified terms, over a rolling window. The improvement team works on #1, not on the loudest complaint. When #1 is solved, #2 becomes #1 and the loop continues.

Small hypothesis, fast test, honest verification. One variable changed. Two weeks of new data. Compared cleanly against the prior two weeks under the same conditions. If it worked, standardise and move on. If it did not, say so, keep the data, move on. No stigma, no dashboard massaging.

Closed loop back to operations. The improvement is not documented in a PowerPoint and filed. It is written into the standard work, enforced in the MES, visible on the shopfloor screen, and audited automatically. The control phase is built in, not added.

Under these four conditions, a plant will improve regardless of whether it calls the programme Lean, Six Sigma, Kaizen or something of its own invention. Without them, no branded methodology will save it.

What changes with a modern MES

Traditional improvement is event-driven. A quarterly Kaizen workshop, a four-month DMAIC project, a yearly TPM campaign. Each event has a start, a middle and an end, and the space between events is where gains evaporate. Modern MES changes this in three concrete ways.

1. Continuous baseline. Every cycle of every machine is recorded. The baseline is not reconstructed once a year; it is visible in real time, with complete history.
2. Automatic Pareto analysis. The top causes of unplanned downtime, scrap and speed loss are ranked live, not compiled after the fact. The improvement team spends its meeting time on decisions, not on data assembly.
3. Closed-loop control. Standard-work instructions, setup parameters, inspection frequencies and reason-code logic live in the MES. A change made by the improvement team is enforced from the next cycle onwards, not hoped for.

The practical effect: the improvement cycle compresses from months to days. A plant running this loop generates more measurable improvement in twelve weeks than most traditional programmes deliver in twelve months, and the improvements stick because the control phase is automatic.

A real case: Neoperl

Neoperl is an international manufacturer of precision water-flow components, headquartered in Müllheim with additional European sites in Bulgaria, the UK and Italy. The starting point was fully automated assembly lines with PLC-level alarm data that nobody was using systematically. The improvement team ran quarterly Kaizen events; the lines kept stopping for reasons that were never fully categorised.

The SYMESTIC engagement began as a focused proof of concept on a single line: four weeks to validate that PLC alarms could be captured, categorised automatically and correlated with quality defects, all without operator intervention. The PoC succeeded, the first three lines went live, and the rollout has been expanding continuously since. Critically, Neoperl deliberately framed the deployment as a KVP tool (the German term for continuous improvement), not as a monitoring system. The goal from day one was structured organisational learning, not dashboards.

The measured results after the first year of operation:

• 10 % fewer stops, through automatic capture of short-stop causes that operators previously did not record
• 8 % higher availability, through structured root-cause analysis on the top alarm categories
• 15 % less scrap, through correlation of PLC alarms with downstream quality defects
• 15 % productivity gain, through targeted actions driven by weekly Pareto reviews

None of these came from new machines. All came from running a disciplined improvement loop on a measured baseline, with the control phase enforced by the MES itself. When I say that methodology matters less than baseline and loop, this is the evidence.

FAQ

Should we pick Lean or Six Sigma?
Neither as a standalone choice. Use Lean for flow and waste problems, Six Sigma for variation and capability problems, Kaizen for daily shopfloor problem-solving. Most mid-size plants benefit from a combined approach, often called Lean Six Sigma, but the label matters less than the baseline you are measuring against.

Where do we start if we have no improvement programme today?
With measurement, not methodology. Until you have an honest real-time baseline of OEE, availability and quality at machine level, any methodology you adopt will produce arguments rather than results. Measure first, choose method second.

Does process improvement need software?
Not formally. You can run PDCA on paper for a single line. At scale, continuous improvement without software quickly collapses under the weight of data collection and reconciliation. Plants with more than 10 to 15 connected machines benefit substantially from an MES that captures, aggregates and feeds back data automatically.

How do we know if our improvement programme is working?
Sustained, measurable change in the baseline metrics that matter, verified against an honest pre-state. Soft signs like "more engagement in shopfloor meetings" are welcome but not sufficient. If OEE, availability, scrap or lead time are not moving after six months, something in the loop is broken.

Why did our previous Lean/Six Sigma programme stall?
Most likely one of the five failure modes above: dishonest baseline, wrong problem, skipped Control phase, cycle time too long, data in silos. The first three together cover 70 percent of the stalled programmes I have audited.

Can AI replace traditional process improvement?
No, and the framing is wrong. AI makes the analysis step faster and catches patterns that human analysts miss, particularly in multivariate process data. It does not replace the Define, Improve and Control phases, which are organisational and operational. The strongest improvement stacks combine classical DMAIC discipline with AI-assisted analysis on a real-time baseline.

How does SYMESTIC support process improvement?
By providing the measured baseline, the Pareto analysis, the standard-work enforcement and the closed feedback loop that a serious improvement programme needs. Typical results in the first year: 5 to 15 percentage points OEE gain, depending on the starting point. See SYMESTIC Production Metrics.

Related: OEE · MES · Bottleneck · Production Quality · Production Control · Operating Time · SYMESTIC Production Metrics

About the author

Christian Fieg

Head of Sales at SYMESTIC. 25+ years in manufacturing including Johnson Controls, Visteon, iTAC and Dürr. Six Sigma Black Belt with three years on the DMAIC side of the Johnson Controls headliner lines. Led global MES and traceability rollouts across 900+ machines in China, Mexico, USA, France, Tunisia and Russia. Author of OEE: Eine Zahl, viele Lügen (2025). · LinkedIn