Bosch x Almetra

“We could see how much they produced, but not how they were producing it - and that was missing.”

Zsofia Toth Horanszky
DATA ANALYST, BOSCH

A LINE THAT LOOKED EFFICIENT - ON PAPER

In Bosch’s Miskolc plant in Hungary, four identical manual assembly lines produce e-bike batteries, each designed to deliver 1,500 units per shift. Performance reports were consistently positive: targets met, minimal interruptions, stable output. Yet on the shop floor, the story felt different.

Operators and supervisors noticed irregular rhythms in the process. Some stations seemed perpetually busy, while others showed long idle times. Despite the green indicators, the line did not feel balanced. The Continuous Improvement (CI) team suspected untapped potential.

THE CHALLENGE

The Miskolc team wanted to answer two critical questions: Could the same output be achieved with fewer operators? Or could higher throughput be reached with the same resources?

Traditional tools offered no reliable way to answer this.

MES in place

Tracked what the line produced, but not how work actually flowed. Cycle time variation, walking and reaching motions between stations, and micro-delays were invisible.

Manual time studies

Delivered only isolated snapshots. Engineers walked the line with stopwatches and tablets, often capturing “best behaviour” rather than normal day-to-day performance.

Green reports

Since targets were technically being met, there was no automatic trigger to challenge the status quo. The CI team suspected that hidden waste was being masked by overstaffing, uneven workloads, or inefficient operator motion.

Without continuous, unbiased observation, true efficiency potential remained hidden.

ALMETRA’S APPROACH

The suspicion was there - but proof was not. To move forward, the Miskolc team needed something they had never had before: a way to see the real process continuously, not just in occasional checks or time-study snapshots.

What began as a simple camera connection quickly became a new source of truth for the entire line. Using bird’s-eye and station cameras, Almetra transformed ordinary footage into objective, continuous visibility of how work actually flowed, station by station.

The system automatically analysed the video streams, detecting operator presence, motion patterns and activity levels, and correlating these behaviours directly with station-level output. Suddenly, everything that had been hidden - micro-delays, motion between stations, real cycle-time variation - came into view.

• Data was captured automatically and continuously, eliminating observation bias and any need for manual logging.

• The insights reflected true everyday behaviour, not the “ideal” performance often observed during inspections.

• Inefficiencies surfaced immediately, backed by video-based evidence rather than assumptions or gut feeling.

What the data revealed

1. A clear link between staffing levels and output

For the first time, the team could see how production changed depending on how many operators were active on the line. The line consistently met - and often exceeded - output targets with four operators instead of five. In several cases, operators stepped away from the line and production still stayed on target, revealing significant unused capacity.

2. Uneven workloads between stations

The activity-per-station view showed that some operators were working close to full capacity while others had long idle periods. Almetra also revealed why: operators frequently left their stations to fetch materials or support neighbouring stations. This unnecessary motion amplified the imbalance and helped the team pinpoint exactly where layout adjustments and clearer task boundaries were needed.

3. Operators not always following standard work

Cycle-time variance charts highlighted abnormal rhythms - short cycles mixed with long ones. When the team reviewed the automatically linked video clips, they saw that some operators weren’t following the defined SOP. This objective visibility made retraining straightforward and brought the line back to standard.

4. Excess buffer building between stations

Almetra’s cycle-time view showed frequent short cycles followed by longer pauses, a clear sign that operators were building buffers ahead of slower stations. Video clips confirmed the pattern and helped the team visualise how work was being unevenly distributed across the line.

THE RESULTS

After reviewing the data and retraining operators, the line became noticeably more stable and efficient. Cycle times moved closer to the defined standard, and buffering between stations dropped sharply. Almetra helped the team prove that the line could reach higher output with less resources.

+19% shift output

Shift output target rose from 1,500 to 1,780 units.

-20% labour need

Operators required went from 5 to 4 (validated potential).

+48% units per operator

From 300 to 445 units per operator per shift.

-70% buffer building

Buffer building dropped from ~50% to ~15% per shift.

-75% standard work deviation

Deviation from standard work fell from ~20% to ~5% per shift.

BROADER IMPACT

The project changed how the Miskolc plant approached continuous improvement. Instead of relying on spot checks, assumptions, or operator memory, the team gained continuous evidence of how the line actually operated. The change was simple but fundamental: the plant now had real data - showing what happened between cycles, where time was lost, and how staffing changes affected output.

“Before, we couldn’t even analyse data - because we didn’t have any data to analyse.”

Zsofia Toth Horanszky
DATA ANALYST, BOSCH

Almetra did not replace existing systems; it filled a gap they could not cover. The MES showed what was produced, while Almetra showed how it was produced - and where inefficiencies were creeping in.

“For the first time, we could understand not just the results, but the reasons behind them.”

Szabolcs Laszlo
INDUSTRIALIZATION ENGINEER, BOSCH

With that visibility, the team uncovered 19 percent hidden capacity, stabilised cycle times, reduced buffers, and validated that the line could run with fewer operators. None of this came from guesses or stopwatch snapshots; it came from continuously observing the real process and acting on what the data revealed.

For plants with manual or semi-manual work, the lesson is straightforward: you can’t improve what you can’t see.

Ready to see what AI could reveal on your line?