Do this on the floor
Find the weak link
One real job followed to cash, a notebook, and a week spent on the thing everything waits on.
Your business is a chain
Output is set by the weakest link. Strengthen any other link and the chain holds the same weight. Speed up the fast stations and you get a longer pile in front of the slow one.
In The Goal, a scout troop arrives when the last boy arrives. Speeding up the fast hikers only stretches the line. The slow hiker, Herbie, sets the pace. Two moves matter: put him at the front so the line stops spreading, and lighten his pack.
In a shop the stretched line is half-finished work. The pack is everything the constraint is asked to do that someone else could do, or that should not be done at all.
Walk one job
Follow one real order from the first enquiry to the payment. Sketch every handoff. Write the wait at each handoff in hours or days. The longest persistent wait is your first suspect.
If you cannot get a live order, walk the path of the last job that shipped and reconstruct it with the people who touched it. Ask the expediter where they always have to go.
you are improving a station with no pile in front of it, or you cannot point to one job you followed to cash.
Three kinds of bottleneck
Settle the goal first. For a commercial business that is more money now and in the future. Safety, quality, and keeping good people are conditions you must not break.
| Kind | What you see | What to do first |
|---|---|---|
| Physical | A machine, a person, a bay, a truck. A pile in front. That resource is busy while others wait. | Squeeze it. Buy capacity only if you still must. |
| Policy | A rule, batch size, approval, or measure that caps output long after its reason expired. | Change the rule. After a constraint breaks, old rules stay and become the new one. |
| Market | Spare capacity inside the walls and not enough good orders. | Stop hunting plant bottlenecks. Sell reliability you can actually hit. Do not fill the week with cheap work that wrecks the price of the good work. |
For the longest wait you found, write one sentence: physical, policy, or market. Then run the Bottleneck Finder. If the two answers disagree, believe the floor.
"We need another CNC" is a request. Ask what rule decides the work that machine is given.
Squeeze it before you spend
Do these in order:
- Find the constraint.
- Squeeze it. Keep it fed. Give it only good work. Stagger breaks so it runs while its operator eats. Move the quality check in front of the pile so it never polishes rubbish.
- Make everything else wait. Other stations should sometimes be idle. Running them flat out buries the constraint in half-finished work.
- Spend only after the free capacity is used.
- Start again when the constraint breaks. Throw out the old rules that assumed it was still there.
Do not start a job until the kit is complete: drawings, materials, fasteners, a free bay, the answer to the open question. A job started incomplete parks itself on the constraint and steals an hour of company output.
The parts are cheap. The process that starves the machine you already own is the expensive thing.
Pick one free change and run it for five working days. Write down what the constraint did each day: hours running, hours waiting, hours on rework.
Feel why a "balanced" line fails
Five stations. Each rolls a die each round. That is its capacity, because real work fluctuates. A station can only pass on work it actually has. Common sense says a line that averages 3.5 at every station should ship 3.5. Run it.
Bad luck accumulates. Good luck evaporates. The fix has three parts.
| Part | On the floor |
|---|---|
| Drum | The constraint sets the beat. Everyone else marches to it. |
| Buffer | A cushion of time in front of the drum, and only there. The drum must not starve. |
| Rope | New work is released at the drum's pace. Do not start jobs to look busy. |
If you run projects, the same physics applies. Safety padded into every task gets eaten. Put one buffer at the end. Do not start every project at once if they all sit on the same person.
Walk your own line and mark where the drum, the buffer, and the rope would sit. If you cannot point to a rope, you are releasing work because someone is free, not because the drum can use it.
Three numbers
Most accounts tell you where money went. These three tell you what to do next.
| Measure | Meaning | Ask |
|---|---|---|
| Throughput (T) | Revenue minus truly variable cost (materials, freight, sales commission). The rate the system makes new money. | Does this help us sell more, faster? |
| Investment (I) | Money tied up: stock, plant, unpaid invoices. | Does it lock cash or release it? |
| Operating expense (OE) | Money to keep the doors open: wages, rent, power. Direct labour sits here in most shops, because it does not rise with each extra unit. | Does it raise the monthly burn? |
Profit is T minus OE. Return is that profit divided by I. Cost-cutting stops at zero. Throughput can keep rising. Because most operating expense does not move when you sell one more unit, extra throughput hits profit almost intact.
Paste last month's lines into the Throughput Converter. Correct the buckets it guesses. Then take one real decision on your desk and write T, I, and OE next to it.
This week
Monday: name the constraint from the walk. Write one free change with a due day. Before you approve any spend, glance at T, I, and OE. Friday: read the constraint's hours and the pile in front of it.
If the pile shrank and output did not rise, you squeezed the wrong thing, or you ate the buffer, or the constraint is the market. Did T move, or did you only rearrange I?
Say it to one other person in the building, in their nouns:
- We are a chain. Output is the weak link.
- We squeeze that link before we spend.
- Everyone else marches to its pace, even if that means standing idle.
- When it breaks, we start again, and we throw out the old rules that assumed it was still there.
If you want someone on the floor with you, that is the Floor Walk.