MEP Fabrication Software Problems: The 5 That Cost Real Money, and How to Avoid Them

Why does scoping this as a shop only system happen so often?

The pain is loudest in the shop, so the fix gets written for the shop. Routing, stations, barcodes, work in progress, a dashboard for the shop manager. Everybody in that meeting works in the building where the problem is visible.

The diagnosis is incomplete rather than wrong. A mechanical, electrical and plumbing (MEP) fabrication shop measured on machine output will optimise nesting and batching, because that is what it is measured on and both are rational. Nesting groups similar parts regardless of where they go. Pipe spool fabrication batches by size and schedule to cut changeover. Both produce output in an order the field cannot install.

What is missing is a demand signal. The field publishes an install sequence by zone with required on site dates, derived from the project schedule and refined weekly by the superintendent who actually knows what has slipped. The shop then plans against those dates, and nesting still optimises but within a window bounded by need date. The shop keeps most of its efficiency and stops building four weeks ahead of a zone that is not ready.

The second half of the same omission is kitting. A kit is not a list of spools, it is everything a crew needs to install a zone, including hangers, rod, fittings, gaskets, fasteners and the drawing package. Shipping 95 percent of a kit delivers none of the value, because a crew missing one box of threaded rod is a crew standing still.

The scoping test is one question. For a given zone, can the proposed system say what is required, what is fabricated, what is kitted, what is loaded, what is received on site and what is installed. Six states. If it stops at fabricated, it is a shop system and your yard problem survives it.

What goes wrong when you migrate item numbering and model data?

The same elbow exists as a model element with a unique identifier, as a computer aided manufacturing item with a shop number, as a printed barcode label, and as a line on a cut list. Nothing in that set is authoritative, and migration is where you either fix that or entrench it.

The usual failure is importing whatever the current project used and calling it the item master. Numbering has drifted across three projects, two of them reused prefixes, and one detailer added a suffix convention nobody else follows. A scheme that has drifted cannot be reconciled after the fact, which is why this is the hardest thing in the category to retrofit and the cheapest thing to get right in week one.

Two specific traps sit underneath it. Labels get reprinted, on the floor, routinely, and a reprint that mints a new sequence creates a second identity for one physical part, so its scan history splits and the item appears both fabricated and not. And model element identifiers change when a family is swapped or an element is deleted and recreated, so a mapping keyed only to the model identifier breaks after the first design change, which arrives within weeks.

Three rules. Define the item master and the numbering scheme before importing anything, and do not attempt to reconcile historical projects into it. Carry the model element identifier, the shop item number and the label identity as three separate attributes with the item master key as the only thing anything else references. And make a label reprint resolve to the existing item rather than create a new one, every time, with no exception for the night shift.

Why do the model, machine and accounting integrations break after launch?

Three integration surfaces cause most post launch pain here, and all three fail quietly rather than loudly.

Model changes are first, and they are certain. A republish updates quantities, and if the import writes directly, a spool already cut can vanish from the requirement list with nobody telling the shop. Model imports have to be proposals rather than writes: a change produces a reconciliation report showing added, removed and modified items, the detailing manager approves it, and any item already cut becomes superseded with a disposition rather than disappearing.

Machine posting is second. Posting cut data to a plasma table, a coil line or a pipe cutter is specific to the controller on your floor, and the dangerous failure is a post that succeeds with the wrong kerf or the wrong material assumption. Nothing errors. You find out at the first fit up, having made several hundred parts. A first article check on every new post, before the run, is worth the four minutes it costs.

Accounting is third. Viewpoint Vista, Sage and Acumatica are each their own project, and the recurring failure is a posting into a closed period that gets rejected silently, after which the two systems drift for a month.

The fixes are ordinary. Treat every inbound file as untrusted with a schema check that fails the whole import rather than writing partial rows. Never let an automated process remove an item that has been cut. Never post to a closed period without an alert and a queue. And alarm on any feed that has not arrived since yesterday, because an absent file is invisible while a wrong file is at least present.

What happens when weld and material traceability are not covered?

Mechanical contractors welding pressure piping carry obligations that a general tracking tool ignores entirely: weld maps by spool, welder identification against qualification records, procedure references, heat numbers from material test reports, and nondestructive examination results with their reports.

The failure is not that the records do not exist, it is that they exist somewhere else. Kept in a binder or a separate spreadsheet, they are not linked to the item identity the shop and the field already use, so assembling a turnover package at the end of a job takes weeks and uses the people who should already be on the next project. That is a real labour cost that appears nowhere in the software business case.

The second failure is welder qualification continuity. If qualification expiry is not linked to weld assignment, welds get made by someone whose continuity has lapsed, and it is discovered at turnover when re-testing is expensive and the schedule is gone.

What belongs in scope, once the six states are working: quality records attached to the same item identity as everything else, welder qualification expiry blocking assignment automatically rather than being tracked on a wall chart, and a turnover package that is a single export rather than a project. It is a strong reason to build, and it is not a first release feature, which is the mistake most quality managers push for.

Should you build custom or configure what you already own?

If you run a single trade shop, model in Revit, and want conventional spool and duct tracking with kitting, buy. STRATUS is well established with mechanical and sheet metal contractors and sits naturally alongside the Autodesk tooling your detailers already use. MSUITE covers similar ground and belongs on the same shortlist. If your process fits either, use it and put the energy into detailing standards, which is where prefabrication actually succeeds or fails.

The signals that building is the honest answer are specific. Your numbering and zoning methodology is genuinely proprietary and is the reason your install rates beat your competitors. One shop serves several trades, and sheet metal, pipe and plumbing have different work in progress models rather than different settings. You need the demand signal to come from a field sequence your superintendents control weekly. You require weld and material traceability tied to the same item identity as everything else. Or you run more than one shop and need to load balance between them, which no product handles well because it depends on your geography and your fleet.

When two or more are true, build. When none are, a product will serve you better and faster.

How do hidden costs get into the quote?

Four places, all visible at scoping if somebody asks.

Machine integration. Posting to the specific equipment on your floor is the single largest variable in the category, and a quote written without naming the machine and the controller is a guess.

Labels. Industrial labelling sounds trivial until the label has to survive galvanising, a site winter and six months in a yard. Material trials and printer selection are real work, and a label that falls off converts a tracked item into an untracked one.

Multi trade. Pretending sheet metal, pipe and plumbing share a work in progress model produces a system that suits none of them. Each trade is closer to a parallel implementation than a configuration.

Offline capability in the field app. This is architecture rather than a feature, and it affects every screen, because plant rooms and basements have no signal and that is exactly where install scanning happens.

Then the recurring line nobody quotes. Maintenance runs 10 to 20 percent of build cost per year, so roughly $15,000 to $30,000 on a $150,000 first release, covering hosting, monitoring, model tooling updates and the annual drift in machine and accounting formats. Ask for the estimate broken out per trade and per machine.

What separates a build that works from one that fails here?

Rolling out one trade, on one active project, with the six states and nothing else. The most common way these projects stall is adding weld traceability and earned value to release one because the quality manager and the project manager both pushed hard in the same meeting. Both are worth building. Neither belongs first.

Then making the field scan trivially easy. A phone, a QR code, one tap, working with no signal and syncing later. Anything heavier is abandoned in the second week, and once the install feedback loop is broken it does not get repaired, because nobody trusts a partial data set.

Then expecting a fortnight of adjustment on the floor. Station layout, label placement and scanner positioning all need tuning, and adoption depends entirely on scanning being faster than the whiteboard it replaces. Budget the time rather than treating it as resistance.

Then an owner inside the business, and it should be the detailing manager who keeps the spreadsheet, working alongside a superintendent who will actually publish the install sequence. One without the other rebuilds the same shop only system with a better interface.

Then the contract. The repository, the infrastructure accounts and the data are yours before kickoff. This system encodes your numbering conventions and your install methodology, which is the competitive knowledge you are paying to formalise, and renting it back from a vendor makes no sense.