Reading a revision cloud without redoing the takeoff

Rev D lands at 4 p.m. and the bid is due tomorrow. The takeoff you finished on Rev C took a day and a half, and now every sheet in the set has clouds on it. The question that decides how your evening goes is what actually changed, and the clouds by themselves won’t answer that.

What a cloud does and doesn’t tell you

A revision cloud marks where the drafter wanted your attention on that sheet, in that revision. That’s the whole guarantee. It doesn’t guarantee that every change got clouded, that whatever’s inside the cloud actually touches your scope, or that the schedule got updated to match whatever moved on the drawing.

• Changes with no cloud at all. A schedule regenerated straight from the architect’s model can change a size or a count with nothing drawn to mark it, because nobody clouded it by hand.
• Clouds that have nothing to do with you. On most architectural sets, half the clouds are electrical notes, finish call-outs, or dimension corrections. They cost you time to review but don’t change a single unit.
• Updates that only happened on one side. The elevation gets clouded and revised, but the schedule keeps its old line. Now the two documents disagree, and the cloud only tells you about one half of that disagreement.

Start with the narrative, not the cloud

Every revision should come with a description somewhere, either in the title block or a separate revision narrative: something like “Rev D: revised window sizes at units 204 through 210, per owner.” Read that before you touch a single cloud. It tells you what the revision was actually trying to do, which the clouds themselves only hint at. If the narrative says “structural only,” you should still check the fenestration sheets, just less thoroughly than if it says “revised openings at the north elevation.” And if there’s no narrative at all, that gap is worth its own RFI.

The overlay pass

Once you know the intent, compare the sheets directly: old revision on one side, new revision on the other, or overlaid if your software handles that.

1. Start with the schedule. Diff the window and door schedule line by line: marks added, marks removed, and any field that changed on the marks that survived. The schedule is dense, so this step only takes a few minutes, and it’s what catches the regenerated-with-no-cloud problem.
2. Then go through the clouded sheets. For every cloud, ask one question: does this touch an opening, a size, an operation, or a note that governs fenestration? If the answer is no, keep moving.
3. Finish with the elevations that weren’t clouded at all but still carry your marks. You’re looking for the one-sided update here, where the schedule moved and the drawing didn’t, or the other way around.

Patch or redo?

Patch the takeoff, meaning change only the affected units, when the narrative and your overlay agree the change is bounded, the marks haven’t moved between revisions, and the schedule and elevations still agree once you account for the change.

Redo the count from scratch when marks got renumbered, sheets were added or dropped, or the schedule got reorganized. Once the identities move, patching from memory is exactly how phantom units end up in a quote.

The traps

• Renumbered marks. W4 quietly becomes W5, and a patched takeoff ends up pricing the old W4’s glazing spec against the new W5’s size. Any time the mark scheme shifts, redo the count instead of patching it.
• “No change” clouds. Sometimes a cloud just circles a re-annotated dimension that has the same value as before. Verify the number, confirm nothing actually moved, and let it go. Don’t re-count an entire wall because someone repositioned a text label.
• Revision creep in the quote. Your quote gets built against Rev C, but the order ends up getting cut from Rev E. State which revision your price is based on, right on the quote, every single time.

Twenty minutes vs. a day and a half

It becomes a fast habit: read the narrative, diff the schedule, walk the clouds that actually matter, then double check the sheets nobody bothered to cloud. A full redo only makes sense once the unit identities themselves have moved; most weeks, twenty careful minutes gets you further than a day and a half spent re-counting everything from scratch.