Weld Tracking Software Problems: The 7 That Cost Real Money, and How to Avoid Them

Why does weld identity get treated as a numbering convention?

Because on paper it is one. A drawing gives weld 14 all the context it needs, so nobody in the shop has ever experienced the number as ambiguous. The ambiguity only appears when the records leave the drawing, which is exactly what happens the moment you put them in a database, and by then the schema has usually been designed by somebody who was shown a traveller and asked to digitise it.

The consequence is not obvious for the first few thousand welds. It appears when a project runs long enough to carry drawing revisions, when the same spool numbers recur on a repeat order, or when two jobs run concurrently for different clients. Then queries start returning welds from the wrong drawing revision, repair records attach to the wrong joint, and the quality team quietly stops trusting the system and reverts to the marked up isometric they trust.

The fix is to generate a durable identity at the moment the weld map is created, combining job, drawing, revision, spool or component, and joint number, rather than accepting a number typed by whoever fills in the traveller. Everything then hangs off that identity: procedure, welder, filler lot, parent heat numbers on both sides, fit up, visual, examination request, result, repair and the repair's own examination. Get this right and the Monday morning question becomes a query that runs instantly. Get it wrong and every feature you add afterwards is decoration on a paper process.

What goes wrong migrating qualification records and weld history?

Welder qualifications are the dataset that decides whether the build delivers its main benefit, and they almost always arrive as scanned certificates. A scan is an image of a decision, not the decision itself. To evaluate whether a welder may make a given joint, the system needs the underlying attributes as structured data: process, filler metal grouping, base material grouping, thickness range, diameter range, position, and the continuity date derived from when the welder last used that process.

Transcribing those attributes from certificates is real work and it exposes disagreements. A shop with forty qualified welders frequently finds coupons whose recorded range does not match what supervisors believed, welders whose continuity lapsed and was never noticed, and qualification records for people who left two years ago still sitting in the active binder. Historic weld records have a parallel problem, which is that a weld log kept in a spreadsheet rarely holds the drawing revision, so migrating it produces records that cannot be located precisely on any current drawing.

The fix is to transcribe qualifications properly, with a welding engineer reviewing the ranges rather than an administrator copying fields, and to treat the discrepancies found as a deliverable in their own right. For historic welds, migrate them for traceability and mark them clearly as legacy records with a lower precision of location, rather than pretending they carry the same identity structure as new ones. Budget two to three weeks and expect the qualification review alone to justify part of the project.

Why do arc data, examination and drawing integrations break after launch?

Three integrations matter here and each fails differently.

Arc data breaks on hardware diversity. Kemppi WeldEye is strong with connected Kemppi power sources and Fronius WeldCube is excellent with Fronius equipment, which is exactly the point: a shop that has accumulated machines over fifteen years across several brands, including units that will never be connected to anything, gets partial coverage by design. A build that assumes complete arc data will produce analytics with a hole in them that nobody flags, because absent data looks like good data.

Examination breaks on the subcontractor. Reports arrive as PDFs in whatever layout that firm uses, and the layout changes when they update their template or you change supplier. If somebody retypes them, the delay is measured in days and the error rate is real. If extraction is automated, it needs a confidence threshold and a review queue, because a wrong weld identifier posted against a result is a defect in your record rather than in your steel.

Drawings break on format. If isometrics arrive as PDFs rather than from a system that can export joint data, weld map creation is manual, and that single fact changes the plan and the price more than any other variable in this category.

The fix is to state coverage honestly. Mark which power sources are connected and report analytics with the covered proportion visible. Route low confidence extractions to a human and never let a model write to the register unreviewed. And establish the drawing source before quoting, not during the build.

What happens when sampling rules and continuity checks are missing?

These are the two gaps that turn a documentation tool into a liability, because both involve decisions people make under schedule pressure.

Progressive sampling is the sharper one. Random examination regimes carry an obligation: when a randomly selected weld is rejected, the applicable code requires further examination, typically of additional welds by the same welder on the same run, and if those fail the scope expands again. Applying that rule on a Thursday costs somebody the weekend, which is precisely why it gets applied inconsistently. A system that records results but does not enforce expansion has documented the inconsistency rather than removed it.

Continuity is the quieter one. Under ASME Section IX, qualification is maintained if the welder has used the process within the preceding six months. That is a date calculation, which software does effortlessly and a supervisor at seven in the morning does from memory while short two people. When it goes wrong, the damage is not one bad weld, it is that suspect welds are distributed across every job that welder touched during the lapse, and you cannot bound the problem quickly. The cost is rarely the rework, it is the re examination of far more than you needed to.

The fix is to make the system refuse the assignment rather than report on it afterwards. When a supervisor allocates a welder to a joint, check process, base metal grouping, thickness, diameter, position and continuity, then either allow it or name the failing parameter and list which other welders are clear. For sampling, generate the additional examination requests automatically, select the welds by your written rule, and block the affected items from progressing until results arrive. Nobody then has to make an unpopular decision under pressure, because it is already made and logged.

Should you build custom or configure what you already own?

Under about ten welders with one or two repeat customers and a stable product, do not build. Weldnote is a focused, sensible documentation tool for procedure specifications, qualification records and weld logs, and paired with a disciplined spreadsheet and a good quality coordinator it will serve you. A custom system at that size is a badge rather than a benefit.

If your shop is genuinely single brand on power sources and your main interest is arc parameter monitoring, configure WeldEye or WeldCube. Reproducing what those platforms do inside their own hardware ecosystem is not a good use of money, and their parameter deviation capability is real.

The gap that pushes shops to build is consistent, and it is not documentation. It is that none of the packaged tools models your production flow. Your weld record has to travel with the spool through cutting, fit up, welding, examination, post weld heat treatment, hydrotest and painting, gathering signatures at hold points your quality plan defines and your client's inspector agreed to. Your sampling rule, your repair authorisation route and your data book structure are yours. The tell that this gap is costing you is simple: if the paper traveller still exists alongside the software, the traveller is carrying the process and the software is only carrying certificates.

How do hidden costs get into the quote?

The weld register is the cheap part. These are the lines that move the number.

• Each code you work to. A shop running ASME Section VIII, B31.3 and AWS D1.1 has three different rule sets for qualification range and examination, and each is real work rather than a configuration setting.
• Each customer data book format. Every major client has its own index. The first is a build, the rest are templates, but the first one is not free.
• Weld map digitisation. Genuinely hard if isometrics arrive as PDFs. Establish the drawing source before anyone quotes a timeline.
• Arc data per brand. Each power source manufacturer is a separate integration, and partial coverage needs to be represented honestly in the analytics.
• Qualification transcription. Two to three weeks with a welding engineer reviewing ranges, not an administrator copying fields.
• Extraction review. Somebody has to work the low confidence queue for material test reports and examination reports, and that is an ongoing role rather than a one off.

Ask for these itemised. A quote with one line for integrations has not been scoped against your shop.

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

Ask a prospective developer to model weld identity on a whiteboard before you sign anything. If they do not immediately raise that weld 14 is not unique, and ask how your drawings, revisions and spools are numbered, they have not built this and your first ten thousand welds will teach them at your expense.

Second, ask how they will represent qualification range. You want structured attributes with rule evaluation, not a scanned certificate with an expiry date. A system that only stores the certificate cannot stop a wrong assignment, and stopping the wrong assignment is most of the value on offer.

Third, scope the first release narrowly and deliberately: one code, one customer document format, the weld register with drawing revision awareness, qualification checking, and examination request and result capture. That is where nearly all the risk sits, and a shop that proves those on one live project has a foundation for everything else. Shops that attempt full material traceability, arc data and automated data books in a single release usually deliver none of them well.

Fourth, put ownership in writing before kickoff, covering the repository, the hosting accounts and an unrestricted export. Your weld register is evidence in any future dispute or claim and may need to be produced years after the job closed. A record you cannot export, held in an account you do not control, is a liability rather than an inconvenience.