In a previous article, “Productivity or Kosu: Faster or in Less Time?”, I introduced kosu as a more honest way to talk about labour productivity. The core idea was simple: instead of counting pieces per person per hour, kosu measures how much direct labour time you actually consume per good unit. That earlier piece showed how kosu removes a lot of confusion around PPPH, efficiency percentages, and the difference between “running faster” and “doing the work in less time”. Kosu works beautifully when you focus on one product, or at least a stable product mix. Operators understand the measure, engineers can improve it, and managers can see whether labour content genuinely comes down over time. The follow-up question many readers asked is this:
“How do we measure productivity with kosu when we produce dozens or hundreds of products with different work content?”
This post answers that question. It explains why average kosu breaks down in mixed-model production, and it shows a practical way to roll up kosu across mixed-model production without falling back into standard hours, earned hours, and negotiated efficiencies.
Why Product Mix Breaks Average Kosu
Imagine an assembly area that builds two products. Product A requires 10 minutes of direct labour per unit. Product B requires 100 minutes. Yesterday, the area produced mostly A. Today, the same area produced mostly B. If you calculate a simple average kosu, today’s value will be much higher than yesterday’s.
Did productivity deteriorate? Not necessarily. The product mix changed.
This illustrates the core measurement issue. Kosu is intentionally blind to product mix. At product level, that is a strength, because it tells the truth for each unit: how much labour did this product actually consume?
“Kosu is intentionally blind to product mix.”
At plant level, however, leadership wants a different answer:
- Are we actually reducing labour content across the value stream?
- Are our improvement activities visible in one consolidated indicator?
- Can we compare productivity trends across periods with very different mixes?
If you only look at average kosu, mix movements dominate the signal. You end up debating whether “we built a heavy mix” rather than whether processes improved. This is exactly the point where many organisations quietly abandon kosu and reintroduce standard hours, earned hours, and labour efficiency percentages.
Why Standard Hours Solve The Math But Hurt The Behaviour
Standard hours, earned hours, and labour efficiency percentages have one big advantage: mathematically, they normalise out product mix. Each product gets a standard labour content. Actual labour is compared against that standard. Efficiency is reported as a percentage.
From a calculation point of view, this is neat. In practice, however, several recurring issues appear.
Standards become political. Routing times are negotiated rather than analysed. Engineers hesitate to update standards because cost prices and budgets depend on them. Every time a standard is revised, visible improvement disappears from the KPI. Operators stop discussing how to reduce labour content and instead argue over whether the standard is fair.
None of this means standard hours are inherently wrong. They simply answer a different question. Standard-hour efficiency asks, “How did we perform against today’s standard?” Kosu asks, “How much labour did we actually consume?” If your strategic intent is to remove real labour content from your processes over time, you want to keep that second question at the centre of your conversations.
Measurement Goal: One Coherent System From Genba To Hoshin
So the challenge is not to invent yet another efficiency formula. The challenge is to preserve the strengths of kosu and make them usable at three levels:
- At the shop floor, operators still discuss actual labour content per product and process, to guide their kaizen efforts.
- At plant level, management needs a single, consolidated indicator that is independent of product mix but still rooted in real labour minutes.
- At executive level, Hoshin Kanri requires a long-term productivity trend that is logically connected to what happens at the genba, not an accounting artefact.
Ideally, all three levels are supported by one coherent measurement system: the same underlying logic from operator to CEO. The approach in this article does that by combining product-level kosu with a fixed reference per product and a simple roll-up KPI: the Kosu Index.
Introducing A Fixed Reference Kosu
The first step is to establish a fixed reference kosu for every product at the start of a strategy cycle.
This reference kosu is a snapshot of actual labour content at the beginning of the improvement journey. It is not a costing standard, not a target, and not a budget basis. You measure how much labour the process actually consumes for each product and then freeze those values for the duration of the cycle.
Suppose that at the start of the cycle, Product A has an actual kosu of 120 minutes per unit. Over time, you run several improvement activities: better layout, reduced walking, clearer standard work, perhaps minor automation. When you re-measure, actual kosu has fallen to 98.2 minutes per unit. The improvement per unit is 21.8 minutes.
If you produce 11 units of Product A during a shift, the total labour removed for that product is 21.8 times 11, so 240 labour minutes. You can repeat this calculation for every product produced during the shift: reference kosu minus actual kosu, multiplied by quantity. The sum across all products gives you the total amount of direct labour removed compared with the beginning of the Hoshin period.
The important point is that this still starts from actual labour. You are not comparing people to a theoretical “should”. You are comparing today’s reality to the historical reality of where you started, expressed in the most tangible unit on the shop floor: labour minutes.
From Product-Level Improvement To Area-Level Numbers
To turn these product-level improvements into a KPI that managers can read quickly, you translate the data into two aggregate numbers for each area and period.
The first is Actual Labour: the total direct labour consumed to produce that period’s mix, calculated using current kosu values. The second is Reference Labour: the total labour that would have been required to produce the same mix if each product still had its frozen reference kosu from the start of the cycle.
Mathematically, you can think of it as summing actual kosu times quantity for Actual Labour, and reference kosu times quantity for Reference Labour. The difference between these two totals is the labour content removed. It answers a simple question that works well in tiered meetings: compared with our starting point, how many labour minutes did we take out of our processes in this area during this period?
Defining And Interpreting The Kosu Index
To make the Kosu-signal easy to compare over time and across areas, you introduce the Kosu Index:
- Kosu Index = Actual Labour ÷ Reference Labour × 100
If the index is 100, the area uses the same amount of labour as at the baseline. If it is 95, today’s mix required only 95% of the labour that the same mix needed at the beginning of the cycle. If it is 85, labour content has been reduced by 15%.
A short example makes this concrete. Imagine one shift that produces 11 units of Part A and 5 units of Part B. At baseline, A had a reference kosu of 120 minutes and B had 160 minutes. After improvement, you measure actual kosu of 98.2 minutes for A and 144 minutes for B. Actual labour for the shift is around 1,800 minutes, while reference labour for the same mix is 2,120 minutes. The difference, 320 minutes, is the labour content removed. The Kosu Index is 1,800 divided by 2,120, about 85.
In plain language: today’s production consumed 15% less direct labour than the same mix required at the beginning of our improvement journey. Most operators, engineers, and managers can immediately understand that statement.
Rolling The Kosu Index Through Tiered Management
One of the strengths of this approach is that consolidation stays mathematically clean. Each area reports Actual Labour, Reference Labour, Labour Difference, and its local Kosu Index. At the next tier, you do not average percentages. Instead, you add Actual Labour and Reference Labour across areas and then recompute the index.
Suppose Area A reports 920 hours of actual labour against 1,000 hours of reference labour, which gives a local Kosu Index of 92. Area B reports 490 hours actual against 500 hours reference, with an index of 98. At plant level, you simply combine the labour numbers: 1,410 hours actual against 1,500 hours reference. The plant Kosu Index is 1,410 divided by 1,500, approximately 94.
The logic is identical whether you consolidate two areas or twenty. The KPI can be rolled up from an assembly cell to a value stream, to a plant, and into Hoshin reviews without distortion, while always anchored in real labour minutes.
Why This Approach Fits Lean Thinking
On the surface, the Kosu Index resembles earned-hour efficiency methods. Both compare a current period with some reference. Underneath, however, the philosophy is different.
Standard hours compare performance against today’s standard, which is periodically revised and often used for costing, pricing, and budgeting. Each revision risks resetting the improvement story and entangling productivity discussions with financial constraints. The Kosu Index compares today’s actual labour against a fixed historical reference captured at the start of the strategy cycle. Nobody negotiates that reference once it is set. Nobody updates it after each improvement. Nobody uses it for costing. Its single purpose is to answer whether labour content is decreasing over time.
This keeps the measurement system close to Lean intent. My first kosu article framed productivity as the difference between running faster and doing the work in less time. The Kosu Index is a way to track “in less time” across a complex product mix, without importing the behavioural side-effects of standard-hour efficiency. It encourages conversations about labour content removed, not standards achieved.
“The Kosu Index is a way to track productivity improvement across a complex product mix.”
Practical Considerations For Implementation
A few practical conditions are important if you want this system to work in real plants. The reference kosu must remain fixed for the whole strategy cycle; if you change it halfway, you recreate the same visibility problem you have with updated standards. Standard work should continue to evolve with every improvement. The frozen reference is a measurement baseline, not a work instruction. Kosu itself stays the primary improvement metric at product and process level. The Kosu Index sits above that as a consolidation tool.
Implementation is easier if you start with one area where direct labour data is reasonably reliable and the product mix is representative. Pilot the approach there, use your existing kosu experience from the earlier article to teach the logic, and see what kinds of conversations it triggers compared with your current productivity metrics.
In that sense, this article is the natural “Part 2” to the introductory kosu post. The first piece made the case for kosu as a clearer productivity measure. This one shows how to keep that clarity in mixed-model environments and how to connect kosu into tiered management and Hoshin Kanri without falling back into the standard-hours trap.
“As long as productivity is defined as ‘meeting the standard’, improvement will always compete with gaming the numbers.”
The moment you define productivity as “reducing actual labour content”, kosu and the Kosu Index stop being theory and start becoming a leadership tool. The uncomfortable step is not the maths—it’s deciding what kind of productivity story you want your organisation to tell.
