In this guide 6 sections
Buy an off-the-shelf ERP for accounts, stores and payroll, where your needs match every other factory. Build the part that holds your variants: the configuration of spindle, bed and tool changer into one sellable machine. Test any package against three real orders before you sign.
Where machine tool makers break a standard ERP
A standard ERP treats a product as an item with a code. A lathe is not that. It is a bed length, a spindle bore, a turret or tool changer, a coolant tank and a controller, all chosen by the buyer.
Each choice changes the drawing, the bought-in list and the assembly hours, so one order can create a variant nobody has built before. A package with flat item codes forces you to invent a code for every combination. Ugly.
That is the first sign to watch. If the vendor suggests a new item for every variant, your item master will grow faster than anyone can keep tidy.
What a package already does well
Ledgers, payables, receivables, tax returns, payroll and stores are solved problems, and a mature package has been tested by thousands of finance teams whose auditors already know its reports. Writing them again is waste.
Purchase orders to vendors of castings, ball screws and linear guides also fit well. So do goods receipt notes and stock valuation.
Buy these parts and do not apologise for it. Your edge is the machine, not the general ledger.
- General ledger, payables and receivables
- Payroll and statutory returns
- Stores, goods receipt and stock valuation
- Vendor purchase orders and invoice matching
- Dispatch documents and e-invoicing where the law asks
Where a built system earns its cost
Build the configurator and the machine record. A configurator takes the buyer’s choices, checks that they fit together, and produces a bill of materials, a price and a delivery estimate. A spindle with too much torque for the chosen headstock gets rejected before the quote leaves.
The machine record follows each serial number through casting, scraping, assembly, run-in, inspection and dispatch. It holds the geometric test sheet, the ballbar result and the final inspection report.
Both are where off-the-shelf tools are weakest. Both are also what your buyers feel when the machine arrives on time and works.
- Rules that block incompatible options
- Bill of materials generated from the choices
- Test sheets stored against the serial
- Spares history for each installed machine
Test any package against three real orders
Take the last three machines you shipped: one standard model, one with a rare option and one that needed a special fixture. Have the vendor key them in live during the demo, from enquiry to dispatch, with no prepared script.
Watch for workarounds. Spreadsheets pasted into a notes field, free-text option lists and hand-typed part numbers all mean the package does not model your product. Count every one.
Then ask how a change to the bed casting drawing reaches orders already in assembly. The reply will say more about fit than any feature list.
A split that usually works
Many machine tool makers end up with a bought finance and stores core, plus a built layer for configuration, test records and spares. The two share a part master and a customer master, and a small integration passes order numbers and costs across.
Keep the boundary clear on paper before any work starts. Write down which system owns the part number, the price list and the stock count. Arguments later usually trace back to two systems owning the same field.
Your next step is small. List every field that changes between your best-selling model and its three commonest variants, and show that list to each vendor.
Build or buy scoring sheet
One sheet lists every business area, from variant configuration to spares quoting, giving each its own row. Columns hold the area and its users. Others record how often it changes, whether a package covered it in the demo, the workaround seen, and a final choice of buy, build or keep as is.
Open a blank worksheet to printQuestions owners ask
Can a small machine tool maker use a standard ERP at all?
Yes, for finance, stores and purchase. Trouble starts when every option on the machine needs its own item code. Test that before buying.
Is a built system always slower to start?
Not if it starts small. A configurator for one model family can go live while the rest of the business stays on the package.
Who should sit in the vendor demo?
Bring your design engineer, the assembly supervisor and whoever quotes. They will spot a missing option rule faster than the owner.
What if we already run a package badly?
Find out whether the pain is in the package or in how it was set up. Many cases are poor master data, not a wrong product.