Your machines are running. Your operators are busy. And a job you quoted at three weeks is on its fourth week. The usual reaction is to look at the machines: are they slow, is the programming slow, do we need another shift? Often the answer is none of these, because the machine was never where the time went.
A job in a CNC shop spends most of its life waiting: for material, in front of a busy machine, in a queue for inspection, at a supplier for plating. The hours when it is actually being cut are a small part of the journey, and almost nobody measures the rest.
This article shows how to trace one job from order to dispatch, split its time between working and waiting, name what it waited for, and decide what to fix first. You need a pen and one job card. No software.
Two Clocks Running on Every Job
Every job has two clocks:
- The machine clock: hours of setup, machining and inspection. This is the time you quote, plan around and pay for.
- The job clock: the calendar time from the day the order is accepted to the day the parts leave. This is the time your customer feels.
The two are far apart. A job can need ten hours of machine time and still take three weeks. The gap between them is waiting, and it grows quietly because it doesn't show up on any machine report. A spindle that is idle for one job is usually busy with another, so machine utilization looks fine while the job sits.
That is why speeding up the machine rarely speeds up delivery. It shortens the small clock. The big one is somewhere else.
The Words, in One Place
These terms get mixed up. Here is how this article uses them:
| Term | Meaning |
|---|---|
| Lead time | Customer's order accepted to parts dispatched. What the customer feels. |
| Throughput time | Often counted from release to the floor to finished. Some shops use it for lead time, so check before comparing numbers. |
| Touch time | Hours someone is actually working on the job: setup, machining, deburring, inspection, packing. |
| Queue time | Time a job waits before the next step starts on it. |
| Waiting time | All the time that isn't touch time: queues, material, outside work, decisions. |
Two simple formulas, and that is all the arithmetic in this article: waiting time = lead time − touch time, and flow efficiency = touch time ÷ lead time × 100. Use the same unit for both (working days of eight hours is easiest).
How to Trace One Job (About an Hour)
Pick a job that has already shipped, or one that is running now. A late job is a good choice, because the delay is fresh in people's minds. Then:
1 List every step from order to dispatch
Include the ones nobody thinks of as steps: waiting for material, drawing check, release to the floor, deburring, inspection, outside processing, packing, dispatch. Use the job card, the planner's sheet and the supervisor's memory together.
2 Write a start and an end for each one
Date and time if you can get it, the day if you can't. The gap between one step ending and the next starting is the waiting time. Don't skip the gaps. They are the point.
3 Mark each gap: working or waiting
A step is working when someone is setting up, machining, inspecting or packing the job. Everything else is waiting, including a job sitting at a machine that is running something else.
4 Write what it waited for
One short reason per wait: material, machine busy, program not ready, inspection, outside work, owner's decision, drawing query. If nobody can say, write unknown. A lot of "unknown" is a finding by itself.
5 Add up working and waiting separately
Convert both to the same unit (working days of eight hours is easiest). Then divide hands-on time by total elapsed time. That percentage is the job's flow efficiency: how much of its journey it was actually being worked on. Books and trade articles also call this the value-added ratio or process efficiency, and call the two clocks touch time and lead time. It's the same idea.
If you've read about value stream mapping, you may have found it hard to use in a job shop, where every job takes a different route. A one-job trace is the lightweight version: no product families and no symbols, just one job's real route and where it stopped.
An Example: 60 Pump Housings (Hypothetical)
This is a made-up job for illustration. The pattern is common, but trace your own jobs. Don't copy these numbers.
| Step | Working | Waiting | Waiting for |
|---|---|---|---|
| Order accepted, drawing checked, released | - | 1 day | Drawing query |
| Raw material | - | 3 days | Material arriving |
| Setup and turning | 6 hours | - | |
| Queue for VMC | - | 2 days | Machine busy |
| Setup and milling | 9 hours | - | |
| Queue for deburring | - | 1 day | Nobody assigned |
| Deburring | 2 hours | - | |
| Queue for inspection | - | 2 days | Inspector busy |
| Inspection | 1 hour | - | |
| Anodising at supplier | - | 6 days | Outside work |
| Packing and dispatch | 0.5 hour | 1 day | Waiting for invoice |
| Total | 18.5 hours | 16 working days |
The job was worked on for 18.5 hours (about two and a half working days) inside a journey of about eighteen working days. Flow efficiency: about 13%. For close to seven days in eight, nobody was touching it.
Now split the 16 days of waiting by cause:
| Cause | Days | Share of waiting |
|---|---|---|
| Outside work (anodising) | 6 | Over a third |
| Material arriving | 3 | About a fifth |
| Queue for machine | 2 | One in eight |
| Queue for inspection | 2 | One in eight |
| Drawing query, deburr queue, invoice | 1 each | Rest |
Outside work and material together are 9 of the 16 days, more than half. Compare that with what a machine upgrade would do. If milling and turning ran a fifth faster, this job would save two to three hours. If the anodising wait were halved, it would save three days. The first is a purchase decision. The second is a phone call and a promised date.
Where Jobs Usually Wait
Once you trace a few jobs, the same causes repeat. These are the common ones, and what usually fixes each:
| Waiting for | What you hear | First fix |
|---|---|---|
| A decision | "Waiting for sir to confirm" | Write down which decisions the supervisor or senior operator can make without asking. See why the shop can't run without the owner. |
| Material, drawing or program | "Job was loaded and then we found the material wasn't here" | A ready check two days before the job starts: material in, drawing latest, program and tools ready. |
| A machine | "VMC 2 is full for a week" | Find out whether it's the true bottleneck with the five checks for bottlenecks, then protect it. |
| Setup, tools or fixtures | "Machine stood while we looked for the fixture" | Get the next job's tools, program and material ready beside the machine while the current one runs. |
| Inspection | "Parts are waiting for QC" | Agree an inspection slot and an order, so parts don't queue by accident. |
| Outside work | "Plater hasn't returned them" | One list of everything sent out: supplier, date sent, date promised. Call the day before it's due. Have the challan and transport ready on the day the parts are ready, so paperwork doesn't hold them. |
| The customer | "Waiting for the revised drawing" or "waiting for sample approval" | Ask for everything you need at order acceptance. Write down who you're waiting on and the date you'll chase. |
Most shops have one or two of these dominating. A rush job that takes over a machine adds one more cause: it moves waiting from one job to several others, which is how every job becomes urgent.
Why You Haven't Seen This Before
- Waiting has no owner. Every machine has an operator. A job sitting between two machines belongs to nobody.
- Systems record work, not waiting. Most spreadsheets and ERP screens show "in progress" from the day a job is released. The days it sat in a crate are invisible. See why nobody knows where the job is.
- Busy looks like progress. A full floor feels productive. It's also what rising WIP looks like.
- Quotes are built from machine time. If you add a rough buffer to machining hours, you never learn what the real route costs in days.
Turning One Trace Into a Decision
One job tells you a story. Five tell you a pattern. Here is how to get from a trace to a decision:
- Trace four or five jobs, including at least one late and one on-time.
- Total the waiting by cause across all of them.
- Take the biggest cause and ask what would halve it. Don't try to fix everything.
- Change one thing for the next ten jobs: the send-out list, the ready check, the inspection slot.
- Trace again in a month. If the biggest cause has shrunk, pick the next one.
It is slow, small and dull. It also works better than most improvement projects, because it starts from where the time is, not where people assume it is. Once you know the real waiting time, you can promise dates with more confidence; the Delivery Date Check builds a date from waiting, machining, outside work, inspection and buffer, and the WIP calculator shows what the work sitting on your floor costs you. For the wider set of levers, see how to improve on-time delivery in a CNC job shop.
Where to Start This Week
- Pick the last job that shipped late.
- Sit with the supervisor for an hour and list its steps, start and end of each.
- Mark each gap as working or waiting, and write what it waited for.
- Add up working and waiting, and work out the percentage.
- Look at the biggest cause. Decide who owns it this week.
You don't need a system for this. You need to look at one job's whole journey instead of one machine's day.
How We Solve This
At KeshRaghTech we start with an Operational Workflow Review: we trace real jobs through your shop, split their time between working and waiting, and name what they waited for. From that we agree with your team the few rules that cut the biggest wait, such as a send-out list, a ready check or inspection slots. Then we deliver them as a workflow application built around how you run, so every wait has a reason and an owner, and you can see where your jobs lose time as it happens. We coach your team until working this way is routine.