How do you improve quality consistency on the production floor?

Quality consistency on the production floor improves when everyone works to the same proven process rather than their own level of skill. Standardized, interactive instructions available right at the workstation keep the same standard on every shift, whatever the worker's experience or language. In practice it comes down to three things: record how your best operator works, make that available at the place and moment the task is performed, and then measure the spread between shifts, not just the average.

Sounds obvious. The problem is that most plants have a standard that is described, not available. Those are two different things, and you can see the difference in the numbers at the end of the line.

Why does quality "float" between shifts?

Antonio Stradivari built around 1,100 instruments. He had a workshop, sons, apprentices and half a century of practice. He left not a single instruction behind. The knowledge of how he selected wood, graduated the plates and varnished the finished body lived in his hands and his eyes. When he died in 1737, the workshop kept going for a while, then the level dropped. For three centuries laboratories have been scanning, X-raying and analysing the Cremona varnish, because the only copy of the process was never written down.

On the shopfloor the same story plays out on the scale of a week, not a century.

First shift: an operator with fifteen years' experience assembles a component at a pace nobody else can match. He doesn't count seconds or check the drawing; he can hear when the clip has seated fully. Third shift: the same workstation, someone three weeks into the job, the same work card, twice the rework. Final inspection catches part of it, the customer catches the rest.

Nobody made a procedural error here. The work card was filled in. What was missing was something else: a reference. Because the real process didn't live in a document. It lived in the head of one person, who happened to be on holiday.

The three typical sources of spread always look similar:

• Skill instead of method. Every experienced operator has worked out their own, slightly different sequence of movements. All of them work. They don't give the same result.
• No single reference. Ask four people from the same workstation how step seven is done. You'll get four answers and four convictions that this is exactly how it's done.
• Knowledge in a few heads. The plant runs as long as three specific people are on the floor. That's not a process, that's a staffing dependency.

How do you standardize work without burying people in paper?

Taiichi Ohno, the father of the Toyota Production System, said there is no kaizen without a standard. It's worth finishing that thought: a standard that isn't at the workstation isn't a standard. It's a document.

A document loses to the workstation for very mundane reasons. The binder stands three metres away. The PDF instruction sits on a network drive in its 2019 version while a printout with a handwritten correction circulates on the floor. The operator has gloves on, both hands busy and a takt time that doesn't allow for reading fourteen pages. When it's time to decide whether the part has seated correctly, people don't reach for a description. They ask a colleague or do it the way they remember.

The result: the more documentation, the less standard. The paper grows, the knowledge stays in people's hands.

There is another way, and it has existed for a long time. The Ise shrine in Japan is dismantled and rebuilt every twenty years, and has been for over a thousand years. The ritual has religious meaning, but it also has a purely technical effect: each generation of carpenters learns the craft by practising it together with the previous generation. The knowledge isn't archived. It is shown in action, at real scale, on a real object.

That's the whole difference between a description and a demonstration. A description says what should happen. A demonstration shows what it looks like when it's done right.

How do AR instructions reduce errors and rework?

An augmented reality instruction does one thing no document can: it pins information to the physical workstation. The arrow points at this specific hole. The highlight marks this specific bolt and its torque. The next step appears only once the previous one is confirmed.

From the operator's point of view, several things change at once.

Hands stay free. No juggling a tablet, pulling off a glove, hunting for the right page. The instruction sits in the field of view, in the same place as the part.

The translation from flat drawing to physical space disappears. A drawing has to be rotated in your head and matched to the real position of the part. That effort is where most assembly mistakes are born. AR removes it, because it shows the step exactly where the step is performed.

Language stops being a barrier. The same sequence can be played back in Polish, Ukrainian, English or Romanian, with no translated binders and no supervisor acting as interpreter.

The reference becomes the expert, not the author of the procedure. The way your best operator works is recorded once and from that moment is available on every shift. Exactly what Stradivari never did.

Behind this sits a mechanism well known from learning research: learning by doing can be up to 24 percent more effective than learning by reading and listening. Someone who performs a step guided in the context of the workstation remembers it differently from someone who read about it in onboarding three weeks earlier.

The effect shows up in two places: less rework, because the error never happens, and a shorter ramp-up for new people, because they no longer wait for an experienced colleague's free moment.

How do you measure the effect?

If a rollout doesn't change the numbers, it wasn't a rollout. Four indicators are enough to settle it.

• Rework. The number of corrective operations per hundred units, counted at the same workstation before and after.
• Non-conformities. Defects and complaints in PPM, split into caught internally and caught by the customer.
• Spread between shifts. The same indicator calculated separately for each shift, then the gap between the best and the weakest.
• Time to autonomy. How many days pass from a new person's first day to the moment they work unassisted and within the quality norm.

The third point matters most and is forgotten most often. An average can improve simply because the best shift went even higher. Consistency means something else: the result stops depending on who came to work today. Until the spread between shifts narrows, the standard works on one shift, not in the process.

Collect the baseline before the pilot, over two to four weeks. Without it, any conversation about the effect is a conversation about impressions.

Where do you start?

Not with a plant-wide transformation. With one workstation.

Pick the one that generates the most rework or the most questions to the supervisor. Measure its results for two to four weeks. Record how the best operator works and turn it into an interactive instruction available at the workstation. Run it on all shifts, including the night shift, because that's where a standard is tested most honestly. After a month, compare the four indicators with the baseline.

One workstation, one month, your own numbers. That's enough to decide about scale without leaning on someone else's case studies.

Aikando does one thing in this process: it turns the way your best people work into a standard available to the whole team, at the workstation and in the worker's language. Knowledge stops being a trait of a few people and becomes a trait of the process.

Stradivari left no instructions because in the 18th century there was no way to record them. Today there is. The only question is whether the knowledge of your best people stays in the company or walks out the gate with them.

Book a demo: info@captix.eu