Foundry Production Management Software: How Do You See a Rising Scrap Rate Before the Customer Rejects the Shipment?

The rejection that arrives six weeks after the cause

A customer quality engineer calls on a Thursday. Twelve castings from a shipment failed at their machining line with porosity that opened up after the first cut. He wants to know what else from that lot is at risk, and he wants an answer today because his line is down.

Your answer takes two days to assemble. The shipping record gives you the lot. The lot came from a week of production. Which heats poured those moulds is on the pour log, a clipboard by the furnace, and the entries after Wednesday afternoon are in a different hand because the melt supervisor was off. The spectrometer results for that week are in a folder on the metallurgist's PC named by date, not by heat. The grinding department scrapped nine pieces of the same part number that week, but the tally sheet says only scrap total, not defect code, so nobody can tell you whether the porosity was already showing up in house.

By the time you reconstruct it, the answer is that the moulds in question came from the second half of a heat where the pour temperature dropped while the ladle waited for a crane. Everyone on the floor half remembers that. Nothing recorded it. And that is the recurring pattern in casting: the information that explains a defect exists for about four hours, in three separate places, and then it stops existing.

Why general ERP (Enterprise Resource Planning) breaks on a foundry floor

Manufacturing software assumes units in and units out. A foundry does not work like that. You buy scrap and pig by the ton, you melt in heats, you pour moulds that contain a number of cavities per pattern, and the casting that leaves the mould weighs a fraction of what you poured because the gating and risers come back as revert. Your yield is metal to good castings, and it is the number your entire margin sits on. A standard bill of materials cannot express any of that without abuse.

Then there is tooling. In a large share of jobbing and production foundries the pattern belongs to the customer, sits in your rack, and carries obligations: you maintain it, you report its condition, you are responsible if it wears out of tolerance, and you cannot use it for anyone else. That is not an inventory item, it is an asset with a contract attached, a cavity count, a shot count, a maintenance history and an owner. Generic ERP has nowhere to put it, so it lives in a spreadsheet and a rack tag.

Then quality. Automotive and pressure containing work drags in certificates, material test reports, chemistry per heat and part submission requirements. The chemistry exists as a printout from the spectrometer. The certificate gets typed. Anyone who has watched that typing happen understands why the traceability chain breaks exactly where the customer will pull on it.

What B&L Odyssey actually leaves you doing

B&L Information Systems built Odyssey ERP specifically for metalcasters, and that focus is real. It understands heats, patterns and casting units in a way that a general manufacturing package does not, and for a conventional production foundry it is a credible answer. If your operation looks like the operations it was designed around, use it.

The places we get called into are the edges of that fit. Foundries with heavy downstream machining and outside processing, where the value added after shakeout rivals the casting itself and the routing spans two or three outside vendors before the part ships. Foundries with unusual melt practice or alloy portfolios where the chemistry model and treatment steps do not match the standard flow. Operations that want live capture from the spectrometer, the pour furnace and the moulding line rather than keyed entry, which is where most of the actual defect intelligence comes from. And shops with a specific commercial model, such as consignment stocking programs or tooling amortisation schedules per customer, that the package expresses awkwardly.

The honest framing is that this is a category with one strong specialist product and a lot of foundries running QuickBooks plus five spreadsheets. If the specialist fits you, buy it. If your process has a shape it does not model, you will end up building around it anyway, and building around a package is usually more expensive than building the layer that matters.

What a custom foundry build has to include

The heat is the spine. A heat record with charge makeup by material and weight, furnace, times, treatment steps, and chemistry readings pulled from the spectrometer against the target window for that alloy, with out of window readings flagged before the metal is poured rather than after. Every mould poured references its heat and its pattern. That single link, enforced at the point of pour rather than reconstructed later, is what turns a two day traceability exercise into a query.

Then pattern and tooling as first class assets. Owner, cavity count, current condition, shot count since last maintenance, storage location, contractual obligations and the parts it can produce. Shot counts accumulate automatically from production, so pattern maintenance becomes a scheduled event driven by use rather than a discovery made when a casting comes out of tolerance.

Then scrap by defect code, captured at every station where scrap is found: shakeout, grinding, heat treat, machining, final inspection. Two fields matter and both get skipped in paper systems: the defect code and the station where it was found. A porosity defect found at final inspection after machining costs many times what the same defect costs at shakeout, and until you record where it was caught you cannot see that cost at all. Once you can, the defect Pareto by part, by pattern and by heat becomes the daily production meeting rather than a quarterly quality report.

Then yield and cost as computed facts. Metal melted, metal poured, good castings shipped, revert returned, per part and per heat. Actual cost per casting from real charge cost, real energy where you can meter it, real labour at routing rates and real scrap. This is where foundries discover that the part they have quoted the same way for six years is losing money because its yield fell when the gating changed.

Then the downstream: routing through cleaning, heat treat, machining and outside processing with a real record of parts sent out and received back, because that is where work in progress goes missing. Then certificates generated from captured chemistry rather than typed, and shipping tied to the lots and heats that a recall question will ask about.

Cost, timeline and what moves them

A first release runs $70,000 to $150,000 and ships in 12 to 18 weeks. That is the heat and pour model, pattern and tooling assets, scrap by defect code and station, and yield reporting. It is deliberately floor first, because the floor data is what everything else depends on and it is the part that has to be reliable before anyone trusts a report.

The full platform with quoting, order management, machining and outside processing, certificates, shipping and accounting integration runs $180,000 to $420,000 across 6 to 12 months. What pushes it up: live equipment integration with the spectrometer, moulding line controls or the pour furnace, which is worthwhile but needs someone who has driven serial and OPC interfaces in a plant environment. Multiple sites. Automotive customer requirements with formal part submission packages. What holds it down: starting with one moulding line and your top 30 part numbers by volume, which covers most of the money and all of the process learning.

When you should not build

A jobbing shop under about 15 people pouring a small pattern set does not need this. A good travel sheet, a disciplined pour log and QuickBooks will do, and the owner already carries the information in his head accurately. The build case begins when the number of active patterns and part numbers exceeds what one person tracks, when customer owned tooling brings obligations you can be held to, or when scrap is a number you see monthly instead of a signal you act on daily.

How to choose a developer

Ask them to model a heat and a mould on a whiteboard before anything else. If they draw a work order that consumes materials and produces units, they are building a machine shop system and your revert, your cavities and your treatment steps have nowhere to go.

Ask how scrap gets recorded at the grinding station. The right answer is a station terminal or a scanner with defect codes as large buttons and a total interaction of a few seconds, because a grinder with a glove on will not use a form. Any design that assumes typing on the floor produces empty data.

Ask what plant equipment they have integrated. A spectrometer, a furnace controller and a moulding line each have their own interface reality, and a developer who has pulled data off one of them will tell you specifically which one. Ask for the make and the protocol, not a general claim about integrations.

Ask who owns the code, the repository and the cloud accounts, and settle it before kickoff. At Digital Heroes the client owns all of it from the first commit. Your alloy windows, defect taxonomy and yield history are the accumulated knowledge of your melt deck, and they should never sit in an account you cannot open.