Study FMS when a known flow problem keeps the spindle waiting. A small shop must first prove the loss with shop data. It must then show that the part, fixture, machine and job data are fit for use.
This is a go/no-go test. It does not predict savings, promise a CNC link, or pick a supplier. It lists what a buyer should know before asking for a quote.
Approved hardware overview
Use the clip to identify physical flow
This 45-second overview shows CNC machines, robotic handling and physical cell flow. It demonstrates hardware and process context only. It does not define software order scope or prove machine compatibility, cycle time, output or a customer result.
A 45-second overview of CNC machines, robotic handling and physical cell flow. It does not define software order scope or prove compatibility, cycle time, output or a customer result.
Where is the lost time?
Do not start with a cell layout. Start with a time study of the real part mix and normal bad days. Keep cutting time apart from wait time. A weekly mean can hide short jobs, rare faults and changeover days.
| Measure | How to record it | What it can reveal |
|---|---|---|
| Spindle waiting. | Minutes by cause, machine, part family and shift. | Shows if loading, setup or the job queue is the real limit. |
| Manual touch work. | Load, clean, gauge, set the fixture and enter data. | Shows which tasks could move off the machine. |
| Stops and faults. | Log the alarm, time, cell state and restart owner. | Shows what must be found and fixed before a low-staff run. |
| Good-part output. | Count passed parts, rework and scrap by one check rule. | Gives a fact base for a later gain check. |
| Queue and stock delays. | Log jobs held by programs, tools, blanks, pallets or checks. | Shows if a control platform could address the delay. |
One CNC may just need an automatic pallet changer. Study FMS when shared jobs, pallets, tools or checks cause wait time across more than one machine.
The five readiness gates
Each gate needs a record, drawing or named owner. A guess is not a pass.
- Stable part family. Blanks, datums, grip faces, programs and check rules are under control. If staff still change the method during a run, wait.
- Known workholding route. Each part has a set fixture or pallet, a clean step and a clamp-state check. One part value does not prove the cut part.
- Machine link facts. List the exact CNC control, code level, options and signals. A brand or plug does not prove a link.
- Usable job data. Control the part, program, tool, fixture and pallet IDs. Name the job-list source and the owner of each field.
- Test owners. Name what the shop, quality team, IT staff and cell team must give, test, sign and reset.
Safety is a separate full-cell duty. A normal process signal is not a safety-rated function by default. The final design needs a site risk check, safe access and tested stop and restart states.
Bring a data pack to the supplier review
A good data pack lets the vendor find gaps before a quote becomes a promise. Include:
- machine model, controller model, software level, option list and available interface documents;
- representative part drawings, blank condition, annual and batch demand, routing and inspection plan;
- fixture and pallet drawings, payload, centre of gravity, clamp method and cleaning needs;
- program, tool, fixture and pallet naming rules, plus the current source of job priorities;
- shift pattern, operator tasks, common alarms, tool-life limits and restart steps;
- network, user access, backup, cybersecurity and remote-support rules;
- the baseline measures above and the buyer's proposed acceptance method.
Mark unknown items as unknown. This is better than a quote filled with broad link or output claims.
What should the smallest pilot prove?
The smallest useful pilot has no set machine count. Use the least scope that can test the real delay, data flow and control steps. One shop may need one path through one CNC. Another may need a shared job list.
| Question | Test evidence | Limit to agree |
|---|---|---|
| Can the job start with the right data? | Trace order, part, program, fixture and pallet IDs. | Set rules for missing, copied and changed data. |
| Can the CNC and cell finish each step? | Test signals, state changes, waits, alarms and reset rights. | List each blocked and timed-out state. |
| Can the part route repeat? | Check parts with the buyer's own test plan. | Name the drawing points and sample rule. |
| Can the team fix faults? | Run fault tests and log the safe restart. | Name who may reset and when work may start. |
| Did the pilot cut the measured loss? | Use the same time and output rules before and after. | Set the target, test span and items left out. |
Do not turn an estimate into a promise. The quote should split pre-ship tests from site tests. It should name the sample, tools, test state and sign-off owner.
An ROI sheet built from the shop's own data
Payback is an output, not an input. Use costs the shop can prove. Keep the base case apart from the best case.
Inputs to place on one sheet
- accepted project cost, including integration, guarding, training, spares and internal labour;
- current and pilot-verified spindle waiting by cause;
- current and pilot-verified manual touch hours;
- loaded labour cost and the value assigned to a recovered spindle hour;
- good-part, rework and scrap costs under one inspection rule;
- new software, service, energy, network, maintenance and tooling costs;
- planned demand and the share that the tested part family can actually cover.
Use one clear model. Add the agreed labour change, spindle gain and quality-cost change. Then take out the new monthly run cost. Divide the full project cost by that checked net gain. If the gain is not yet known, the payback date is not known.
When is a small factory not ready?
- The bottleneck moves every week. Stabilise demand, routing or the process before automating a moving target.
- Part and fixture IDs are informal. A control platform cannot schedule data that the shop cannot identify.
- Operators correct the process by feel. Write and test the normal method and its limits first.
- Tool life, chips or inspection stop the run first. Solve the shorter operating window before adding more queued work.
- The machine interface is assumed. Get the exact manuals, signal list and ownership boundary.
- No one owns faults or safe restart. Name the people, permissions and test before asking for less supervision.
- The ROI begins with a vendor percentage. Replace it with the shop's own baseline and a pilot measurement.
A “not yet” choice can save money. It also gives the shop a clear work list.
Turn the baseline into a go/no-go brief
Send NEXTAS the time study, part list, CNC control details, fixture and pallet data, fault log, shift plan and test goals. We can review the gaps and discuss a pilot. This is not a result claim. The quote must state the order scope, links and tests.
Request an FMS readiness review →
Questions small manufacturers ask before FMS
Is an FMS always cheaper than buying another CNC machine?
No. An extra CNC adds cutting capacity, while an FMS may reduce waiting or coordinate shared work. Compare both options with the same measured demand, labour, downtime, quality and support costs.
What data should a small manufacturer collect before an FMS study?
Record cutting time, waiting time, load and unload work, setup work, common faults, good-part output, scrap and rework, shift pattern, part families, tool limits and the exact machine and controller versions.
Can existing CNC machines be connected to an FMS?
Possibly, but the machine brand alone does not prove it. Review the exact controller, installed options, available signals, door and cycle sequence, network rules, safety design and the party responsible for each interface.
How large should the first FMS pilot be?
Use the smallest scope that can test the real bottleneck and the required data and control handshakes. The machine count, part family, test duration and acceptance limits are project-specific.
What evidence should support a go decision?
Use a signed baseline, a stable part and process pack, an interface register, a fault and restart plan, a risk review, a pilot acceptance sheet and an ROI model built from the shop's own costs. A rendering or supplier estimate is not proof of an operating result.
Continue the decision
Read scope before selecting architecture
Do you have a measured FMS problem?
Share the baseline and the unknowns. Ask for a gap list before asking for a performance estimate.
Send the readiness inputs →