Method

How a comparison is done

The same steps a careful reviewer takes by hand, applied to every document of both stages, with every result traceable to its source.

  1. Lay out the project

    Each stage's files are placed in one folder tree: issue, then kind (drawings, models, specifications, bills, addenda, others). Sorted by volume, never by file extension. Duplicates are recorded, not lost. Sources are read-only.

  2. Build the register of each stage

    Register rows and drawing files are listed per stage and discipline: register and file, register only, file only. A copy inside a stamped archive is listed apart with its container named and is not counted as a second document.

  3. Pair the documents

    Each document is paired across the stages by its number, then checked by revision and date. The register's values and the title block's printed values are both read; every mismatch is listed. Volumes without a register are matched by number or name.

  4. Classify the change

    Per document: matched (same revision and date), revision changed, date changed, one stage only, not compared. A file that changed without a revision is always listed. Every "not compared" and every "one stage only" says why and where the twin is.

  5. Compare the drawings

    Sheet against sheet at a resolution that reads text and dimensions. Each change is looked at, named by element, and placed with a pin. Several records of one change become one finding. A change found on one level is looked for on every level. Moves and resizes under the project's limit (200 mm by default) are listed apart, not counted.

  6. Grade the findings

    Major

    An element added or removed, a room layout change, a level change, a geometry change (moved, resized, re-divided).

    Medium

    Any other non-documentation change: finish or hatch, wall type on the same geometry, a printed dimension restated, a door swing.

    Minor

    Documentation only: tags, notes, stamps, title blocks, drafting lines.

    • 3grades
    • 13category codes
    • 200mm size limit
    • 1finding per change

    Graded by what is drawn, never by "possible". Each finding carries tags: category code, zone, disciplines touched, repeated on N levels, printed value changed, and the bill item it affects.

  7. Show the evidence

    One image per finding: first stage on the left, second stage on the right, at the same window and scale. It opens large in the page and travels into the Excel export.

  8. Compare the specifications

    Whole volumes, section by section, between the two issues. Within one issue, the specification is checked against the drawings and schedules: codes on drawings exist in the specification, specified items have a drawn use, disagreeing values are listed. Quotes as printed.

  9. Reconcile

    Per-discipline sums equal the totals. The totals are explained against the engine's own counts: pairs, identical, one stage only, identity unknown, not rankable. Each number can be reached from the page.

  10. Deliver

    One self-contained HTML file, offline, phone-friendly. The Excel workbook is exported from the page, from a template, with the active filters. A share version embeds everything and leaves out the locked part.

Honesty rules

What the report will not do

No invented numbers

No rate or quantity appears unless it comes from a bill row or a drawing measurement whose method is written down. Measured quantities are marked approximate.

No vague labels

Not "needs a look" but the actual finding, with its count and its reference. Not "possible" but what is drawn.

No silent choices

Equal copies that cannot be told apart are held and listed. Anything the rules leave open is asked, with numbers, before it is decided.

No changes to the sources

Project files are read where they sit. Nothing is renamed, moved, overwritten or deleted.

NABU Compare Contact