The calc sheet changed. Nobody diffed it.

The relief valve sizing workbook lives in Excel. Always has. Process updates a flow case — column D on the HMB tab, a few cells downstream. Saves. Exports a PDF snapshot for the submittal. The P&ID still shows the old valve size because the piping lead has not picked up the change yet. Both files go out on the same transmittal.

Reviewers compare PDF to PDF in Bluebeam. The calc PDF looks fine — neat tables, familiar layout. Nobody sees that cell D14 went from 12,400 to 8,900 lb/hr. The diff is a number in a grid, not a cloud on a drawing.

Why spreadsheets slip through

Engineering review culture is drawing-heavy. Redlines on P&IDs are legible; redlines on a calc workbook are a specialty skill. Excel's track-changes is brittle across versions. Many teams export to PDF precisely because the customer cannot edit the file — but that export also hides the edit history.

• Formula cells change value without changing font color — nothing "looks" different on print.
• Multiple tabs; reviewers open the summary sheet, not the inputs tab where the error lives.
• Linked workbooks break silently when someone moves a file in SharePoint.
• Version suffix in the filename says Rev B; the author forgot to bump it after a Friday edit.

Where it hurts

Downstream documents cite calc outputs by reference: "CV per Calc HV-202 Rev B, sheet 3." If Rev B's PDF is stale but the native file moved to Rev C, datasheets and narratives drift from the numbers that actually govern the design. Failures show up late — procurement, fabrication, or field fit-up — when changing a valve size is expensive and explaining the gap is worse.

If the calc is the source of truth, the calc gets diffed like a drawing — not buried as a static PDF.

Review habits that catch spreadsheet drift

• Include native Excel in the review session when possible, not PDF alone.
• Require a short change log tab on every calc workbook — what changed, why, which downstream docs need update.
• When a driving input changes, grep the package for the old number (same as pressure spec checks).
• Assign a calc reviewer who is not the author — different discipline if you can spare one.
• On release, lock the workbook hash alongside the PDF snapshot so Rev B cannot mean two things.

Visual diff for Excel sounds mundane until the first time you watch a cell-level highlight catch a flow rate that would have sailed through a PDF compare. Kord treats spreadsheets as first-class deliverables in the same review session as the P&ID and the spec — because on most projects, the spreadsheet is where the mistake starts.