Throughput

Throughput is the rate at which good parts leave the process, and it is the metric most small shops misread the worst. The temptation is to count everything the machines made and call it a productive day. The honest count is the one that subtracts rework, scrap, and parts still sitting in the rework cage. A shop with high machine utilization and low throughput is a shop in trouble, but the dashboards will not tell you that until the customer does.

"Busy is not the same as productive. Throughput is the only number that knows the difference."

How throughput works

Throughput is calculated by dividing good output by elapsed time. The two parts of that calculation are where the disagreements live. Good output means units that passed final quality, not units the machines produced. Elapsed time means the wall-clock window you are measuring, not the time the machines were running. The first distinction strips out the hidden factory of rework. The second strips out the trick of timing only the productive hours and ignoring the changeovers, breakdowns, and waiting.

Anatomy of a throughput number

A real throughput number has three components worth pulling apart:

• The count. Good units, full stop. Anything that needed rework before passing final does not count until it passes.
• The window. A consistent interval. An hour, a shift, a day, a week. Mixing windows produces nonsense.
• The mix. Throughput on easy parts is different from throughput on hard parts. Note the mix or the average lies to you.

Throughput is also the output of the slowest step in the chain, not the average of every step. If a shop has five operations and the fourth runs at 60 units an hour while every other operation runs at 100, the shop's throughput is 60. The four faster operations are producing inventory that will sit. This is why lean shops fix the constraint first and ignore the rest.

Where throughput fits on the shop floor

Imagine a 25-person CNC job shop running parts for two HVAC contract customers. The owner has been tracking machine hours and reporting 85 percent utilization to the customers, who are happy. Then a shipment misses its date by ten days and the relationship cools. The owner walks the floor and finds the answer: 85 percent utilization, yes, but throughput on the shipping dock has been declining for six weeks. The rework cage has grown. A worn fixture on the second op produces parts that pass at the machine and fail at final inspection. The machines are busy. The good output is shrinking.

The fix is not better utilization tracking. The fix is putting a throughput board at the shipping dock with a daily target and a daily actual, updated by hand. The fixture gets noticed within two days because every operator can see the gap on the board, and the conversation moves from "we ran hard today" to "we shipped what we promised." Throughput is the metric that surfaces problems machine-hours-tracked will hide.

Common mistakes with throughput

• Counting gross output instead of good output. Parts the customer rejects do not count. Parts the customer accepts after the third reorder do not count three times.
• Picking the wrong measurement point. Throughput measured at the first operation tells you what the first operation made, not what the shop shipped. Always measure at the last step that actually matters to the customer.
• Treating throughput as a target. Pay-per-unit schemes inflate the count and ruin quality. Throughput is a result of good flow, not a goal you push toward.
• Mixing windows or mixes without noting it. A weekly throughput number that averages across different product mixes is a number that lies. Either pin the mix or report by mix.
• Reporting throughput from a dashboard nobody on the floor sees. Throughput lives on the shop floor. If only the office sees it, it is a vanity metric.

Throughput and related Lean tools

Throughput is one of the canonical lean KPIs and the rate-counterpart to throughput time, the duration a single unit takes to travel the value stream. It is closely tied to overall yield and first-pass yield, since only good output counts. And it is the result side of capacity utilization: chasing utilization usually drops throughput, while improving flow lifts both at once.