Metal Processing · Implementation
How to implement ERP in a metal processing company
A practical plan for putting an ERP live in a processing plant: what to decide first, which data to prepare, how to start with one machine, and the mistakes that delay go-live.
By Nexfloe · · 6 min read
- Industry:
- Metal Processing
- Article type:
- Implementation
ERP projects in processing plants rarely stall because of the software. They stall because item masters are a mess of sizes, opening stock has no heat numbers, or the floor is asked to weigh everything without being shown why. This guide sets out an approach that keeps the project small enough to finish and close enough to the machines to be used.
Quick answer
Implement ERP in a metal processing plant in stages. Agree the records the plant most needs, such as batch yield and heat numbers on dispatch, then prepare item masters with units and conversions, processing routings, machines, stage checks, scrap reasons and opening stock by lot and heat number. Configure and test on real orders, put one machine live first with operators and QC trained at it, then add the rest.
Before you start: decide what the system must do on day one
Write down the questions the plant head and customers ask most often. For most processors the list is short: yesterday's yield on each machine, the heat number behind a delivery, where the scrap came from, what is at the pickling vendor, and what an order really cost per tonne. Those define the first scope.
- Name an owner from the plant, usually the plant or production head, not IT or accounts alone.
- Choose the first machine: one with steady volume, a willing supervisor, and orders customers ask questions about.
- Agree what stops when that machine goes live, such as the production register or the scrap book for that machine.
What data do you need to prepare?
| Data | What to check | Usually owned by |
|---|---|---|
| Item masters | Grade, thickness, width or size, units of weight and pieces, conversion on each item, no duplicate sizes | Planning or sales |
| Processing routings | Operations in sequence, machines, rates, vendor steps, expected yield | Production |
| Machines and work centres | Slitters, cut-to-length lines, presses, furnaces, shifts and available hours | Production |
| Stage inspection plans | Characteristics with nominal and tolerance, such as width, burr, flatness or hardness, and frequency | Quality |
| Certificate characteristics | Which chemistry and mechanical values are recorded at receipt | Quality |
| Suppliers and vendors | Mills, traders and job work vendors for pickling, galvanising or heat treatment | Purchase |
| Opening stock | Each coil, plate or bar by lot, heat number, weight, pieces and location, with certificate values where held | Stores and quality |
| Usable offcuts | Grade, size, weight and heat number, or a decision to scrap them before go-live | Stores |
| Open orders | Customer orders with specification, quantity, unit and date | Sales or planning |
| Reason codes | Short lists for scrap (edge trim, setting loss, rejection, offcut) and downtime | Production and quality |
Rates and expected yields do not need to be exact on day one. A working figure is enough, and the floor's own weights will show the real yield within a few weeks.
The stages of an implementation
1. Configure the plant as it runs
Machines, shifts, items with their units, routings, scrap reasons, inspection plans and certificate characteristics are set up in the plant's own terms. Opening stock is loaded by lot and heat number, and the accounting sync is connected. By the end of this stage a test coil should run from receipt to dispatch on the screens.
2. Run the first machine on real orders
Operators, the supervisor and QC on the first machine are trained where they work, one short session each, and record real batches while the rest of the plant carries on as before. Gaps show quickly: a missing scrap reason, an item with the wrong conversion, a check at the wrong frequency. Fix them the same day.
3. Switch the first machine over
The first machine now runs on the system alone, and its register stops. Planning, stock, yield and cost read what the floor records.
4. Extend machine by machine
The other machines, the vendor steps and dispatch follow the same pattern. Each goes faster than the last, because the items, routings and training approach are already proven.
How long does a metal processing ERP implementation take?
Three things decide it: how much is in scope, how clean the item and stock data is, and which system you choose. A broad enterprise ERP covering finance, procurement and several plants is a much larger project than putting production, stock by lot and quality live on one machine. FleekERP's implementation plan is built to put the first production line live in 18 days. A plant migrating years of history from another ERP, or with item masters and routings that need rework first, will take longer.
Common ERP implementation mistakes in metal processing
1. Configuring software before understanding the process
Setting up screens before anyone has walked the bays gives a system that matches a template, not the plant. Follow one coil first: where it is weighed, where it is issued, where trim goes, which checks happen at the slitter, what goes to vendors. Set the system up to match that coil's path, and run it unchanged for a few weeks before asking for new fields.
2. Poor master-data preparation
Duplicate sizes, wrong conversions between weight and pieces, opening stock without heat numbers: each one breaks the first yield figures and the first traces. Clean the items and stock you are starting with before go-live. The opposite mistake is as common: migrating every past transaction. Open orders, current stock by lot and active items are usually enough, and history can stay in the old system for reference.
3. Ignoring shop-floor users
The slitter operator and the QC inspector create most of the entries, but they are usually the last to see the system, and often not at the machine. Train them at the machine on real batches, and keep their entries to what they need: output weight or count, scrap with a reason, downtime with a cause, and check values. FleekERP takes these entries through shop floor apps running on ordinary Android phones. Show supervisors their machine's yield from the first day, so the weights are used and not just collected. The habits that need breaking, such as weights sent as photos to a group chat, are covered in why WhatsApp ends up running the production floor.
4. Trying to digitise everything at once
Receipt, every machine, vendor processing, dispatch and costing on the same day spreads the team too thin to fix problems as they appear. One machine live is worth more than five half configured.
5. Unclear ownership of routings and yield standards
If nobody owns the routings and expected yields, they drift: a new size is added with no yield figure, a knife setup changes and the routing still shows the old trim. Name an owner, usually the production head for routings and yields and quality for inspection plans, and agree how a change is made.
6. No pilot, or a pilot with no end
Switching every slitter, press and furnace at once means each gap shows up on every machine on the same day. A pilot machine finds them on a small scale. Set the date its register stops, though: a pilot that runs alongside the old book indefinitely becomes a second copy nobody trusts.
7. Measuring software usage instead of operational outcomes
How often people log in and enter data tells you the system is in use. They do not show whether the plant loses less material. Track the yield, scrap and trace times the project was started to fix.
After go-live: what to measure
- Yield per batch by machine, reviewed weekly, and whether the gap to the expected yield is closing.
- Scrap by reason and shift, and whether the largest reason is shrinking.
- The time taken to answer a customer's heat number or certificate query.
- Weight differences on vendor returns caught on the order, not at stock-taking.
- Physical stock by lot and heat number compared with the system at month end.
The share of batches recorded at the machine is still worth watching in the first weeks, as an early sign that the outcomes will follow.
For how the full flow runs once the plant is live, see FleekERP for metal processors.
Frequently asked questions
How long does ERP implementation take in a metal processing plant?
That turns on what is in scope, how clean the data is, and which system is chosen. A broad enterprise ERP across finance and several plants is a much larger project than putting production, stock by lot and quality live on one machine. FleekERP's plan puts the first line live in 18 days; migrating years of history or reworking item masters takes longer.
What data does a processor need before going live?
Item masters with grade, size and units of weight and pieces, processing routings with expected yield, machines and shifts, stage inspection plans, certificate characteristics, suppliers and job work vendors, opening stock by lot and heat number, usable offcuts, open orders and short scrap and downtime reason lists.
Do we need heat numbers for all opening stock?
For any stock you expect to trace or certify, yes. Coils, plates and bars loaded without a heat number break the first traces. Where a heat number is genuinely unknown, decide before go-live whether that stock is sold as untraced, used internally or scrapped.
Which machine should go live first?
One with steady volume, a willing supervisor and orders customers ask about, usually the slitter or the cut-to-length line. Set a date when its register stops, then extend machine by machine.