How to implement ERP in a fabrication or machining company

Job shops are harder to implement than repeat-production plants in one way and easier in another. Harder, because every job is a little different and the routing cannot be written once and forgotten. Easier, because the shop is usually small enough that the owner, the planner and the supervisors can sit in one room and agree how it works. This guide uses that advantage.

Before you start: decide what the system must answer

Write down the questions that cost the shop the most when nobody can answer them. For most job shops the list is short: which jobs will be late this week, how loaded each machine is, how many parts are at each vendor, which heat a part came from, and what a finished job actually cost. Those questions set the first scope.

• Name one owner from the shop, usually the owner or the production head, not an outside consultant alone.
• Pick the first machine group: the machines that decide most delivery dates, with a supervisor who wants it to work.
• Agree what stops when that group goes live, such as the whiteboard or the job register for those machines.

What data do you need to prepare?

Setup and run times do not have to be exact on day one. A reasonable figure is enough to start; the times operators log will show the real numbers within a few weeks, and the routings can be corrected from them.

The stages of an implementation

1. Configure the shop as it runs

Machines, shifts, rates, routings, reason codes and inspection plans are set up in the shop's own words. Opening stock is loaded by heat number, vendors are added, and the accounting sync is connected. By the end of this stage one real job should run on screen from release to dispatch, including a vendor step.

2. Run the first machine group on real jobs

Operators, setters, supervisors and QC on the first group are trained at their machines and start recording real jobs while the rest of the shop carries on as before. The gaps appear quickly: a machine whose hours were set too high, a missing downtime reason, a routing that needs another operation. Fix them the same day.

3. Switch the first group over

The first group now runs on the system alone. Its jobs come off the whiteboard, and the planner loads those machines from the board instead of from memory.

4. Extend group by group

The remaining machines, the vendor operations and the costing follow in the same pattern. Each group goes faster, because routings, reason codes and the training approach are already proven.

How long does it take?

That depends on the scope, the state of the data and the system chosen. An enterprise ERP covering finance, purchasing and several plants is a much larger project than putting jobs, machines and vendors live for one machine group. FleekERP's implementation plan is built to put the first line live in 18 days. A shop migrating years of history from another system, or one whose routings have to be written from scratch first, will take longer.

Common ERP implementation mistakes in fabrication and machining

1. Configuring software before understanding the process

Setting up screens before anyone has followed a job round the shop gives you a system that fits a template. Walk one real job first: where the drawing comes from, who writes the routing, where bar is cut, how parts leave for the heat treater and how they come back. Configure to that, and run it for a few weeks before asking for changes.

2. Poor master-data preparation

Machines with unrealistic available hours, routings without setup times, opening stock without heat numbers: each one makes the load board or the first job costs wrong, and people stop trusting them early. Clean the data for the first machine group before go-live. The opposite mistake is just as common: trying to load every old job and transaction. Open jobs, current stock and active routings are enough to start.

3. Ignoring shop-floor users

Machinists and setters make most of the records, and they are often the last to be asked. Train them at their own machine on a real job, and keep their entries to the few they need: setup, count, scrap with a reason, downtime with a cause. In FleekERP these are made on shop floor apps on ordinary Android phones, without barcode scanning, so picking the job from a short list has to be quick. Show supervisors their machines' load and progress from day one, so the entries are used. The piece on why WhatsApp ends up running the floor covers the habits that have to change.

4. Trying to digitise everything at once

Every machine, every vendor, costing, quality and dispatch on the same day spreads the team too thin to fix problems as they appear. One machine group running properly teaches the shop more than the whole shop half set up.

5. Unclear BOM and routing ownership

In a job shop, routings are written all the time. If nobody owns them, every planner writes them differently, setup times are left blank, and drawing revisions on the order drift from the routing. Name who writes and approves routings and BOMs, and agree how a revised drawing is handled for jobs already released.

6. No pilot, or a pilot with no end

Going live across the shop without a pilot means every problem appears everywhere at once. A pilot group finds them on a small scale. Set the date the whiteboard comes down for that group, though. A pilot that runs alongside the whiteboard for months becomes a second copy nobody trusts.

7. Measuring software usage instead of operational outcomes

Logins and entries show whether people are using the system. They do not show whether jobs leave on time or whether the shop now knows what they cost. Measure what the project was meant to change.

After go-live: what to measure

• Jobs delivered on or before the promised date, and how early a late job was seen coming.
• Machine load against available hours, week by week, for the machines that decide delivery.
• Short and rejected returns from vendors caught on the job rather than at month end.
• Scrap and rework by operation and reason, and whether they are falling.
• Actual job cost against the quotation, for repeat parts and for new ones.

The share of jobs recorded at the machine is still worth watching in the first weeks, as an early sign the outcomes will follow. For how the whole flow runs once the shop is live, see FleekERP for engineering and machining units.