Yamazumi Software: Turning Measured Work Into a Balanced Line

What a Yamazumi chart shows

Each vertical bar is one station or operator. Each segment inside the bar is a single work element with its measured duration. A horizontal line marks takt time — the pace customer demand requires. Read together, the chart answers three questions at a glance: which station exceeds takt and therefore caps output, which stations carry slack, and how much of each bar is work the customer would actually pay for.

The power is in the segments, not the totals. A bar that runs 12 seconds over takt tells you there is a problem. A bar whose segments show 9 seconds of walking and 6 seconds of searching tells you what to do about it.

What Yamazumi software does that a chart tool does not

Drawing stacked bars is the easy part — any spreadsheet does it. Dedicated software earns its place through everything around the chart:

• Keeps each segment linked to the observation it came from, so a number can be traced back to the video frame that produced it.
• Recalculates station totals, takt compliance and balance efficiency the instant a work element is reassigned.
• Stores the before state, so an improvement can be proven rather than asserted.
• Carries the same work elements into standard work documents, so the chart and the operator instruction cannot drift apart.
• Produces the report without a second round of copying into slides.

In other words, the chart is the output. The value is the measured, maintained work-element library underneath it.

The data a Yamazumi chart is built from

A chart is only as sound as the observations behind it. Four inputs are non-negotiable:

1. Work elements — the process split into small, repeatable, individually timed activities, not one lump per station.
2. Measured times — several cycles per element, averaged, taken under normal running conditions.
3. Work classification — each element tagged value-added, incidental (necessary but not value-adding) or waste.
4. Takt time — net available production time divided by customer demand for the same period.

Estimates corrupt all four. A chart built from remembered times looks identical to one built from measured times, which is exactly what makes it dangerous.

Why the classification matters more than the total

Two stations can both sit at 62 seconds against a 60-second takt and require completely different responses. If the first is 58 seconds of value-added assembly plus 4 seconds of incidental inspection, it is genuinely full and work must move elsewhere. If the second is 38 seconds of assembly plus 24 seconds of walking, searching and double handling, nothing needs to move at all — the waste needs to go.

This is why classification is a required step rather than a refinement. Without it, rebalancing degenerates into shuffling waste from one operator to another, which changes the shape of the chart without changing the output of the line.

From video to work elements

Stopwatch studies remain valid, but they have structural limits: the observer's reaction time enters every reading, the observation cannot be reviewed, and the operator's awareness of being timed changes what is measured.

Video-based measurement removes those limits. The same cycle can be reviewed as often as needed, boundaries between elements can be set precisely rather than reflexively, a second engineer can check the split, and the recording remains as evidence when someone questions the number six months later. It also captures what a stopwatch never records — the reach, the repositioning, the second trip to the rack — which is usually where the recoverable time sits.

Takt time as the reference, not a target for every station

Takt is calculated from demand, not from what the line currently achieves: net available time divided by units required. If a shift runs 450 productive minutes and the customer needs 500 units, takt is 54 seconds.

A balanced line is not one where every station reads exactly 54 seconds. Stations that perform critical quality functions, or absorb variation from multiple product variants, are often deliberately left with margin. The engineering objective is continuous flow under takt with minimal waiting — not identical bar heights. Chasing numerical equality is one of the most common ways a technically correct Yamazumi study produces a worse line.

Testing a rebalance before you move anything

The costly part of line balancing is not the analysis — it is discovering after the workstations have been rearranged that the constraint simply moved downstream.

Working through the reassignment on the chart first answers the questions that matter while they are still cheap: what happens to station 3 if this element moves to station 4; whether one operator can absorb two elements without breaking takt; how many operators the line needs if demand rises 20 percent and takt tightens accordingly; whether overtime disappears through redistribution alone. Each of these is a few minutes of scenario work against measured data, and each one avoided a physical change that would have had to be undone.

Where the chart connects to the rest of the improvement work

A Yamazumi chart is rarely the only artefact a project needs. The same measured work elements normally feed several other documents:

When those documents are produced from one work-element library, they stay consistent. When each is built separately in its own spreadsheet, they diverge within a quarter — and the operator ends up following whichever one is printed at the station.

Reporting that survives contact with management

Engineering analysis that cannot be explained in one page does not get funded. A Yamazumi report that leads to action generally contains: the process studied and why; the current state with takt, station cycle times and the balance efficiency; the constraint and what specifically causes it; the proposed change with the expected chart after it; and how the result will be verified.

The before-and-after chart pair does most of the persuasive work, because it converts an engineering argument into a picture a plant manager can read in seconds.

Offline operation and manufacturing data

Production video is among the most sensitive material a plant holds. It shows layout, tooling, methods, cycle times and — unavoidably — identifiable people. In automotive, aerospace, medical device and defence supply chains, uploading that material to a third-party cloud service is frequently prohibited outright by the customer contract, not merely discouraged.

This is a genuine architectural fork rather than a feature checkbox. Software that processes video locally keeps the recording, the analysis and the exports inside the plant. Yamazo Studio runs fully offline for exactly this reason: no production video leaves the machine it was analysed on.

When a spreadsheet is still the right answer

Dedicated software is not the correct choice for every plant, and pretending otherwise wastes everyone's time.

The honest threshold is maintenance effort. When engineers spend more hours keeping workbooks accurate than analysing production, the tool has become the constraint.

Mistakes that make a correct chart useless

• Building on estimated times. The chart looks the same and the conclusions are wrong.
• Measuring one cycle. Normal variation gets promoted to a finding.
• Skipping classification. Waste gets rebalanced instead of removed.
• Ignoring takt. Even bars that fail to meet demand are not a balanced line.
• Looking only at labour. The constraint may be a machine, a material feed or an inspection queue.
• Treating it as a one-off. Product mix, demand and staffing all move; a chart that is never rebuilt describes a line that no longer exists.
• Measuring during abnormal conditions — training, trials, breakdowns — and calling it the current state.