ERP for Job Shops: Why High-Mix Work Breaks Standard Systems
Most ERPs assume repeat production. Job shops and machine shops quote one-offs and reroute them every time. Why that breaks, and what to look for instead.
Most ERPs are built on an assumption that job shops break on day one: that you make the same thing more than once.
If you searched for machine shop software, you are in the right place. A machine shop running one-offs and short runs is a job shop by another name, and everything below applies the same way.
Repeat production is what gives a standard system its structure. A part number has a bill of materials, a routing, and a standard cost, and those get refined over runs. A job shop quotes something it has never made, routes it around whatever is free, and may never make it again. The data model expects a library and gets a stream of one-offs.
That mismatch is not a configuration problem. It shows up as an implementation that stalls, or as a system that goes live and gets worked around, which is the same outcome arriving more slowly.
The four places standard ERPs break on high-mix work
1. Quoting from history that does not exist. Standard costing assumes a prior run to cost against. Your estimator is pricing from experience, a similar job from two years ago, and judgement about the material. Systems that require a costed BOM before a quote force estimators back into the spreadsheet they were supposed to leave, which is where most job shop ERP adoption actually dies.
2. Routings that are decided at run time. In repeat production the routing is a property of the part. In a job shop it is a decision made the morning of, based on what is free, who is in, and what else is on the machine. A system that treats the routing as fixed makes every real-world reroute either a data-entry chore or an untracked deviation.
3. Part numbers that multiply. Every one-off wants an identifier. After three years you have tens of thousands of part numbers used once each, and searching them is useless because nobody remembers what the customer called it. The system is technically correct and practically unusable.
4. Scheduling built for lines, not for contention. Finite scheduling assumes stable cycle times. Yours vary by an order of magnitude between jobs. The schedule the system produces looks authoritative and nobody on the floor believes it, so a whiteboard reappears and the two drift apart.
What actually matters in a job shop system
The priorities invert compared with a repeat manufacturer. What matters most is not planning depth. It is estimating, tracking actuals against the estimate, and knowing what a job really cost when it ships.
| Capability | Repeat production | Job shop |
|---|---|---|
| Standard costing | Central | Nearly useless |
| Estimating from similar past work | Minor | The core of the business |
| Actual cost captured per job | Useful | Decides whether you were right |
| Fixed routings | Central | A liability |
| Flexible operation tracking | Minor | Central |
| MRP and demand planning | Central | Mostly noise |
| Quote to actual comparison | Nice to have | The most valuable report you own |
The bottom row is the one worth dwelling on. If you cannot compare what you quoted against what the job actually consumed, you are estimating on instinct forever and you will never find out which kinds of work quietly lose money. Most job shops we meet suspect a category of job is unprofitable and cannot prove it.
Why the MES answer is usually wrong here
Shops that struggle to see the floor often get pointed at an MES. For high-mix work that is generally the wrong direction, because MES pays back through repetition and standardised routings, and a job shop has neither. The implementation ends up modelling a process that will not hold still.
What a job shop usually needs from the floor is narrower: which job is at which station, how long it has been there, and what got consumed. That is capture, not execution management, and it is a much smaller build. The MES and ERP comparison covers where that line falls.
The contrarian part: your spreadsheet is not the problem
The estimating spreadsheet in a job shop is usually the most sophisticated asset in the building. It encodes years of judgement about material, setup, and which customers change their minds. Replacing it with a rigid quoting module is a downgrade dressed as an upgrade, and estimators know it, which is why they keep using the spreadsheet after go-live.
The problem is not that the estimate lives in a spreadsheet. It is that the estimate never gets compared against what happened. Fix the loop and the spreadsheet can stay for a long time. The wider version of this argument applies to the whole floor.
A five-question test
1. Can you tell, for last month, which jobs came in over the estimate and by how much?
2. When a job gets rerouted on the floor, does the system know?
3. Can your estimator find a similar past job in under a minute?
4. Do you know your true cost per job, including the setup and the rework?
5. Is anyone maintaining a schedule outside the system, on a whiteboard or in a spreadsheet?
A yes on 5 is the diagnostic one. A parallel schedule means the floor does not trust the system's, and no amount of configuration fixes a trust problem.
Questions we get
What is the best ERP system for machine shops? There is no single best one, and any list that names one is ranking marketing budgets. The shops we see succeed pick on fit against their own weirdest job, not on feature count. Test the candidate with a one-off that reroutes halfway through, and watch whether the system fights you. That single test separates the categories faster than a six-month evaluation.
What is the best machine shop management software? Same answer, different words, and worth saying plainly because the two phrases return different results. The category includes full ERPs, shop management systems, and scheduling tools that do one slice well. Which is right depends on whether your problem is quoting, the floor, or the money, and most shops are clearer about that than they think once someone asks.
What are the top 3 ERP systems for manufacturing? We will not rank them, and we would be suspicious of anyone who does without asking what you make. The honest version: the systems that come up repeatedly for high-mix work are the shop-focused ones rather than the tier-one suites, because the tier-one suites are built around repeat production. Ask the vendor how many of their reference customers run one-offs, and ask to talk to one.
Is ProShop an ERP system? Yes, and it is a credible one for shops that fit its model, particularly AS9100 work where its quality module is the draw. It is opinionated about how a shop should run, which is a strength if you agree with it and a cost if you do not. That is the real evaluation question with any of them, not whether the label says ERP.
What is the best software for machine shop scheduling? Scheduling is the one to be most careful with, because finite scheduling assumes stable cycle times and yours vary by an order of magnitude between jobs. A schedule the floor does not believe gets replaced by a whiteboard within a month. Before buying a scheduling module, check whether the problem is the algorithm or the fact that nobody trusts the data going into it. It is usually the second.
Are there ERPs built for job shops? Yes, several, and for a shop whose work fits their model they are a reasonable buy. The evaluation question is the same as anywhere: whether you adapt to the model or the model fits how you already run. Worth testing with your weirdest job rather than your most typical one.
Is machine shop ERP different from job shop ERP? Same category, different words. Machine shop describes the equipment on the floor, job shop describes the production model, and most shops we work with are both. Vendors use the terms interchangeably. The question that actually sorts systems is whether the data model assumes you will make the part again, which is the test in this post whatever you call your operation.
We do some repeat work and some one-offs. What then? Very common, and it is the case standard systems handle worst, because they are built around one mode. The practical answer is usually to let the repeat work use the structured path and stop forcing the one-offs through it.
Is custom overkill for a 30-person shop? Not necessarily, and the calculus changed with what it now costs to build. The honest comparison is against what you would pay in licences and adaptation over five years, plus the cost of the workarounds. The build, buy, and partner-built comparison runs those side by side.
What should we fix first? Quote to actual. It is the smallest useful loop, it needs no floor hardware, and it usually surfaces one category of work that has been losing money quietly. That finding tends to pay for the rest.
Our estimators do not want to change tools. They are usually right to resist, and their spreadsheet is not the bottleneck. Start by capturing actuals and feeding them back, which makes the estimator better without taking their tool away.
How long does something like this take? The first workflow should be in real use in weeks. If a plan says a year before anyone touches anything, it has no feedback loop, and that is the most reliable predictor of a stall.
What is next
Pick one question you cannot currently answer and make it answerable. In most job shops that is quote versus actual, and it is a smaller project than an ERP evaluation by an order of magnitude.
The readiness assessment takes 2 minutes and will place you in a tier. If you already know the answer is that nothing connects, tell us which job you wish you could cost properly and we will start there.