The bottleneck is almost never where you think

Ask a shop owner where their bottleneck is, and you'll get a confident answer. Measure it for a week, and that answer is wrong about half the time. The loudest machine isn't always the constraint — sometimes it's the inspection bench, the deburr station, or a single experienced operator everyone routes work through.

CNC bottleneck analysis is just the discipline of letting the data tell you, instead of the gut.

Two signals that prove a constraint

A real bottleneck shows two fingerprints at the same time:

Queue grows in front of it. WIP piles up while downstream stations run hungry.
It runs near 100% scheduled hours. Everything else has slack; this one doesn't.

If only one signal is true, you have a scheduling problem, not a capacity problem.

The data you need (and don't need)

You don't need MES, ERP integration, or a $40k OEE platform to start. You need three things per job, per station:

Time the job arrived in queue
Time the job started running
Time it finished

That's it. With those three timestamps you can compute queue time, run time, and utilization for every work center in the shop. Our CNC machine tracking captures this with one tap and feeds it into work center scheduling automatically.

The four moves once you've found it

Once you've named the constraint, the playbook is well established (Goldratt's Theory of Constraints if you want the reading list):

Exploit it — never let it starve. Material, tooling, and the program live ahead of the constraint, not behind it. A digital expeditor is built for exactly this.
Subordinate everything else — upstream stations only release work the constraint can absorb. Otherwise you're just building WIP.
Elevate it — second shift, lights-out, or finally that new machine. But not before steps 1 and 2 are real.
Repeat — the constraint will move. Plan on it.

Common false constraints

The "slow" machine that's actually starved. It's not slow — it's waiting on material 40% of the time. Fix the downtime before you blame the iron.
Inspection. Often a scheduling and batching problem, not a capacity problem.
The Swiss/5-axis. Sometimes real, sometimes just unbalanced loading. The numbers will tell you.

Where to go next

If you're starting from zero, run downtime tracking for two weeks first — you can't analyze constraints without clean data. From there, manufacturing scheduling software and a real shop floor control loop turn the analysis into actual throughput.

The bottleneck is paying for itself every hour it sits unmeasured. Start the clock.