Most revision mistakes aren't caused by a change nobody could see. They're caused by a change nobody noticed. The revision note said "updated to comments", the clouds were around the obvious bits and the moved setting-out point wasn't picked up.
So, comparing two revisions properly isn't about the tool. It's about covering the whole sheet, checking the things an overlay can't show and knowing what each method misses. This guide goes through the manual method, then what software adds, then how to read a result.
Start with the right pair
Before you compare anything, make sure you're comparing the right two issues. The pair that matters is the new revision and the one the site has actually been working from. That isn't always the previous issue. If P04 went out for comment and site is still building from C01, compare the new C02 with C01.
Check the drawing number and title match, and that both sheets are at the same scale and paper size. A sheet that's been re-scaled or moved to a different sheet size will throw every automated method off.
The manual method
This works with paper, a PDF viewer or anything in between.
- Read the revision note and the revision table. What does the designer say changed? Write it down. It's a list to work from, don't assume it's a summary of the change.
- Find the clouds. Check each one against the note. A cloud with no matching note, or a note with no cloud, should be queried.
- Flick between the two. Put the sheets on screen at the same zoom and switch between them, or just lay them out on a table. Movement often catches your attention in a way side by side doesn't.
- Check the numbers. Dimensions, levels, gridline references and setting-out coordinates. Don't assume a dimension is right because the line it describes looks the same.
- Read the text. Notes, specification references, schedules, door and window tags, material callouts. This is where an overlay is weakest and where cost often hides.
- Follow the cross-references. If the sheet points at a detail or section on another drawing, check that one hasn't changed too.
It's slow. On a busy sheet it can take an hour. But it's the baseline everything else should be measured against.
What overlay does well and what it misses
An overlay puts the two revisions on top of each other in different colours, so anything that moved shows as a colour fringe. Many PDF tools used in construction can do it showing each sheet in a separate colour and can stack more than two.
A wall that moved 100 mm, a drain run that changed route, a door that swapped hands: they all show up immediately.
It's much weaker at:
- Text. A changed note, a new specification clause or an edited schedule shows as a small coloured smudge, if it shows at all. It still needs to be read.
- Sheets that don't line up. If the title block moved, the drawing was re-scaled or the viewport shifted, everything looks changed. Align the sheets on fixed points (gridline intersections are best) to align the information that is important.
- Redrawn linework. A designer who redraws an area without changing it can produce a cloud of colour that means nothing. That creates noise needs to be checked.
- Scans. A scanned print never lines up perfectly with another scan, so an overlay of two scans is mostly noise.
What automated comparison adds
Compare tools go a step further and mark the differences for you. They work in different ways and it's worth knowing which kind you're using.
Some compare the images, pixel by pixel drawing clouds around the areas that differ, letting you align sheets automatically or by picking three points. Because it works on an image, it finds any visible change but doesn't know what the change means.
Some compare the text. They compare text only, which is useful for specifications and notes but ignores the drawing itself. For scanned files it compares images of the pages instead.
AI-based comparison, like the free tool on this site, reads both sheets and describes each change in plain words, with where it is and what it might affect. That's useful for a first pass over a busy sheet and for explaining a change to someone who wasn't looking at the drawing. But even then it's a reading of the drawing, not a measurement of it. It can miss a change and it can describe one wrongly.
How to read a comparison result
Whatever produced the result, treat it as a list of places to look, not a list of changes. In practice:
- Check every finding against both sheets. Open the two revisions and look at each one yourself before acting on it.
- Look for what's missing. Run your own check of the text, the dimensions and the cross-references.
- Be wary of alignment noise. If a result flags most of the sheet, the problem is probably alignment or a re-scale, not a wholesale redesign.
- Keep the result. Save it with the date and both revisions. If the change is disputed later, it's part of your evidence of what changed and when you knew.
Then take each real change through the revision review checklist: what it affects, whether it's been instructed and who needs to know.