A system built only of cost figures saves its way into late deliveries. One built only of delivery figures buys dates with inventory and overtime. Four views keep the trade-offs visible, which is the entire point of balancing a scorecard.
What each view answers
| View | Question it answers |
|---|---|
| Finance | What does it cost, and what does it return? |
| Customer | What does the customer actually get? |
| Process | How well does the plant run today? |
| People and learning | Can we still do this in two years? |
The example plant ended with 24 figures, six per view. Finance carries conversion cost per unit, scrap cost, overtime rate, inventory turns, maintenance cost ratio and energy cost per unit. Customer carries on-time delivery, on time in full, customer PPM, complaint closure time, order fill rate and a satisfaction score. Process carries the OEE of the bottleneck line, first pass yield at final test, lead time, schedule attainment, unplanned downtime and rework rate. People and learning carries skills coverage, ideas per employee, absenteeism, training hours, near-miss reports and injury frequency.
Three rules keep the views useful. No view rests on fewer than three figures. Every view has an owner and a target. And people and learning, the view that gets dropped first, is the one that predicts next year.
The chain is what makes it a system
Four lists become a system only when somebody writes down how the views are supposed to connect. That written expectation is testable, which is what separates a scorecard from a sorted dashboard.
Write a chain backward, starting at the result you are asked for.
- Take a finance or customer figure with a target and a date.
- Ask which process figure has to move for it, and by how much. Write the number, not the direction.
- Ask which people and learning figure has to move for the process figure. This is the link that is usually missing.
- Write each link as one sentence with two numbers on each side.
- Give every link an owner, not only every figure.
The chain of the example plant
Training hours rose from 11 to 24 per employee and skills coverage from 68 to 84 percent, so a setup or a short stop no longer waited for one specialist. Unplanned downtime fell from 9.4 to 5.6 percent and the OEE of machining line 3 rose from 62.4 to 70.1 percent. The line made 40,500 units per month at the old level and 45,500 at the new one, and plant output rose from 43,000 to 48,000 units.
With less work in process in front of final test, manufacturing lead time fell from 11.5 to 7.8 days, so orders left the plant in the week they had been planned for. On-time delivery rose from 87.5 to 95.4 percent and on time in full from 82.0 to 93.1 percent. Fewer express freights and less overtime on almost the same cost base brought conversion cost per unit from EUR 48.00 to EUR 44.00: EUR 2,064,000 for 43,000 units, EUR 2,112,000 for 48,000.
The link that failed
One link did not hold. Energy cost per unit was expected to reach EUR 1.95 and reached EUR 2.05, because the extra output came from a line with a higher connected load. The chain was wrong.
That finding was worth more than the five links that held. The next version of the chain splits energy cost by line, and the plant metered its two largest consumers.
A chain is not a proof. Two figures that move together may both follow a third. Treat it as the best current explanation and let the next twelve months correct it. A plant that tests five links a year learns faster than one that reports forty figures and explains none.
The next part takes these figures down to the areas and the lines without copying the plant report onto every board.