Every Manufacturer's Data Record review finds something. After seventeen years of compiling, reviewing and defending data records across oil & gas, subsea and energy projects, the surprising part isn't that findings happen — it's how predictable they are. The same ten issues appear on almost every package, regardless of fabricator, country or product.
They're predictable because they're structural: each one lives in the gap between documents, where no single author owns the error. A certificate is fine, a weld map is fine — it's the reconciliation between them that fails. Here they are, roughly in order of how often they hold up payment.
1. Witness points signed by the wrong party — or not at all
The ITP says a hold point needed the client's witness. The record shows the fabricator's QC signed it, or the box is simply empty. Either way the activity is contractually unverified, and there's no fixing it retrospectively — the test happened months ago. This is the finding that turns into commercial letters.
2. Heat number traceability breaks
The material certificate says one heat number, the as-built or marking record says another, and somewhere in between a transfer wasn't recorded. The steel is almost certainly fine; the evidence chain isn't. Reviewers pull this thread hard because if traceability fails once, they stop trusting the rest of the record.
3. The weld map and the NDT records don't reconcile
Welds on the map with no corresponding NDT report; NDT reports for weld numbers that don't exist on the map; repairs tested but the original defect report missing. Reconciling these two documents is the single most time-consuming part of a manual review — which is exactly why it's done last, worst, or not at all.
4. Expired calibration behind a test record
The pressure test passed — but the gauge's calibration certificate expired three weeks before the test date, or the cert for that gauge ID isn't in the record at all. One date comparison invalidates a milestone activity.
5. Superseded revisions in the final dossier
The record contains Rev B of a drawing or procedure; the project closed out on Rev D. Compilation pulled from a folder that was current six months earlier. It reads as carelessness, and it makes the reviewer question every other revision in the book.
6. Certificates that don't match the specified grade or type
The spec called for EN 10204 type 3.1 certification to a particular grade; the record contains a 2.2 declaration, or a 3.1 for a different grade than what the purchase order required. The material question is usually resolvable — the documentation finding still stalls acceptance.
7. Test records missing acceptance criteria or instrument IDs
A test sheet with a result but no recorded acceptance criterion, or no gauge/instrument identification, can't be independently verified. "Passed" is an opinion until the record shows passed against what, measured with what.
8. Punch items closed on paper, open in reality
The punch list says cleared, but the supporting evidence for the clearance — the re-test, the replaced certificate, the updated drawing — never made it into the record. The item is administratively closed and evidentially open.
9. The index promises documents the record doesn't contain
The table of contents was written from the contract requirements; the record was compiled from what existed. The gap between the two is a ready-made findings list, and it's the first thing an experienced reviewer checks — because it takes ten minutes and never comes up empty.
10. Sign-off chronology that doesn't make sense
The final inspection is dated before the repair it supposedly verified. The coating report pre-dates the weld it covers. Date-order errors are usually innocent transcription — but they read as fabrication, and they invite a level of scrutiny no record survives comfortably.
Why good reviewers still miss these
None of these findings require judgement to detect. They require cross-referencing — holding two documents in mind at once, hundreds of times, at the end of a project when everyone is exhausted and the milestone is overdue. That's precisely the work humans are worst at and rules are best at.
It's why we built MDRverify: a rule library of 1,300+ checks that runs this entire list — and the long tail behind it — across the whole record, with every finding traced to the exact document and clause. Your reviewers verify findings instead of hunting for them, and the record goes out clean the first time.
Got a record heading for review? Ask about early access — or just argue with our list. We've heard good cases for at least five more.
Run this list automatically.
MDRverify checks all ten — and 1,300 more — across the full record, before the client does.