Lesson 2: Tell me how you measure me

Ruth Adler, the TOC coach Daniel has invited, walks the mill with him. Above the hot-rolling mill hangs a big electronic board: 512 tonnes/hour.

Ruth: “Who looks at that number?”

Daniel: “Everyone. Viktor, who runs the rolling mill, gets his bonus on it. If he meets his tons per hour, nobody asks him anything else. If he misses it, nothing else he did matters.”

Ruth: “Then let’s see what that number does to your whiteboard.”

The mode of managing is expressed in its measurements

How an operation is managed shows up in what it measures, and the measurement controls behaviour. In almost every operation there is a local efficiency measurement. In steel it is tons per hour; elsewhere it is units per shift, utilisation, earned hours or cost per piece. For decades tons per hour was the prime measurement in steel: meet it and you can get away with anything; miss it and nothing else counts.

“Tell me how you measure me, and I will tell you how I will behave.” — Eliyahu M. Goldratt

So let’s follow tons per hour, step by step, the way Goldratt did.

Step by step through the tree

The start. Tons per hour is the prime measurement, and most people behave in line with the way they are measured. So every department tries to maximise its tons per hour.

Branch 1: fast items first. In most departments some items take less time per tonne than others. A thick plate rolls more tonnes per hour than a thin one.

Viktor has orders for 2-inch plate and 1-inch plate, and a month-end tons-per-hour target. What will he roll first?

Whatever is due first
The thick plate: more tonnes per hour
Whatever the planner says, in due-date order

To look good in the measurement period, departments roll the fast items at the expense of the slow ones. The slow items, whatever their due dates, wait. Result: poor due-date performance and too many customer complaints.

Branch 2: produce for stock. Not producing at all scores zero tons per hour. So when there are no orders, departments produce anyway, for stock, even with no market request in the short or medium term. Result: inventories too high and, with money tied up in steel nobody ordered, payback periods too long.

Branch 3: bigger batches. Every set-up, changing the alignment of the rolls, means hours without producing, which lowers tons per hour. So departments pull orders forward from future weeks to make each batch bigger. Result: work that isn’t needed yet is done now, and work that is needed waits behind it: long lead times, high inventory, poor due-date performance.

Branch 4: stealing. Steel is a V-plant: from one slab come hundreds of end items, differing in width, thickness, surface and grade. Every step is a divergence point where material can go to more than one order. To feed a big efficient batch, a department takes material intended for one order and uses it for another, which then has to be made again from scratch. Goldratt called it stealing. It feeds every UDE on the board, including the last one: when one department keeps taking another’s material, relationships between departments go bad.

Go over the tree again, arrow by arrow, and check one thing: as long as local efficiency is the prime measurement, the undesirable effects are bound to exist. No planning system, reorganisation, merger or psychologist removes them, because none of them touches the cause.

Exercise: which behaviour driven by local efficiency is this?

The cutting line uses plate reserved for an urgent export order, because it completes a full batch of another size.
The furnace keeps running at the weekend to make slabs “we’ll surely need”.
The mill rolls next month’s orders of the same grade now, to avoid a roll change.
The thin-gauge order slips again; the thick plates are always rolled first.

Four types of plant

Goldratt classified operations by the shape of their flow, because each shape produces its own typical UDEs:

V-plants: few raw materials, many end products, many divergence points (steel, textiles, paper, chemicals). Typical damage: stealing.
A-plants: many parts converging into few products (machinery, aircraft). Typical damage: parts waiting at assembly for the one that’s missing.
T-plants: a limited set of parts combined at the end into many variants (appliances, electronics). Typical damage: stealing at final assembly.
I-plants: a straight line, one route.

The details differ, but the root is the same in all four: efficiency used as the prime measurement.

But there must be a good reason

If one measurement does so much damage, why is it the prime measurement almost everywhere? Goldratt insisted that defining the core problem needs one more step: understanding why the system goes along with the problematic measurement. There must be a good reason. That is the next lesson.

Check your understanding

Answer all five. Get four right to complete the lesson.

1. Why does tons per hour lead to poor due-date performance?

Workers are lazy
The machines are old
Fast items are rolled at the expense of slow ones, whatever their due dates
Customers order late and change their orders after they are confirmed

The measurement rewards what is fast, not what is due.

2. Producing for stock with no orders happens because…

Not producing scores zero on the efficiency measurement
The forecast says the products will be needed soon anyway
Stock is cheap
Customers prefer it

Idle time looks like failure under a local efficiency measurement.

3. “Stealing” is typical of which plant type?

I-plant
A-plant
None
V-plant, with its many divergence points

And of final assembly in T-plants.

4. The tree shows that, as long as local efficiency is the prime measurement…

The UDEs can be fixed one by one
The UDEs are bound to exist
Inventory will fall
Only V-plants suffer

The measurement is the common cause.

5. Before declaring the core problem, TOC also asks…

Who introduced the measurement, and when it was last reviewed
How much it costs
Why the system continues to comply with it
What competitors measure

There must be a good reason, and it will turn out to be a conflict.