Drawing revisions are handled badly by treating them as small updates instead of new scope. The reliable workflow is: log every revision the day it arrives, compare it sheet-by-sheet against the set you actually priced, re-take-off any sheet that changed structurally, and state in your bid exactly which revision of each sheet your number is based on. Estimators who skip the comparison step and "eyeball the clouds" are the ones who find 14 extra tons of steel after contract award.
Most bid busts I've seen weren't takeoff errors on the original set. They were revision errors — pricing Rev 2 while the GC awarded on Rev 4, or catching the clouded change on S-201 but missing the unclouded one on S-502. The original takeoff gets your full attention for two days; Addendum 3 arrives at 4pm the day before bid and gets twenty minutes. That asymmetry is where the money leaks out.
Why Revisions Are Where Estimates Quietly Go Wrong
Three things make revisions more dangerous than the original takeoff:
You're anchored to your first number. Once you've priced a job at 82 tons, every revision gets unconsciously evaluated as "probably doesn't change much." An engineer upsizing the roof beams from W12x19 to W16x31 across 40 pieces adds roughly 7 tons — about $18,000 in material alone at $1.20/lb installed-cost thinking — and it looks like nothing on the page. The section marks just got a little different.
Revisions arrive under deadline pressure. Addenda cluster in the last 72 hours before bid. You don't have two days; you have an evening. Corners get cut in exactly the place accuracy matters most.
Changes aren't always where the clouds are. A column size change on the framing plan quietly changes base plate thickness on the detail sheet, anchor bolt diameter in the schedule, and connection design on three sections — and the engineer may only cloud one of those sheets. If your review is "look at the clouds," you're pricing a subset of the change.
The failure mode is almost never a wild miss. It's 3-8% of tonnage, silently, which on a job you won at 4% margin is the whole margin.
A Revision-Tracking System That Fits on One Sheet
You don't need document-control software. You need one table per bid, kept in the bid folder, updated the day anything arrives. If you already keep a bid log like the one in our guide to organizing steel bids, this lives right next to it.
| Column | What goes in it | Why it matters |
|---|---|---|
| Sheet no. | S-101, S-201, etc. | One row per structural sheet |
| Rev priced | The revision your takeoff used | This is your baseline — no ambiguity later |
| Latest rev | Updated every time a set arrives | Instant visual: any mismatch = work to do |
| Date received | When the new rev landed | Proves timeline if scope disputes surface |
| Change summary | One line: "Roof beams upsized, gridlines 4-7" | Forces you to actually read the sheet |
| Tonnage impact | +/- tons, even a rough number | Running total tells you when to re-price vs. absorb |
| Status | Not reviewed / Reviewed / Re-priced | The only three states that matter |
Two rules make this work:
- Log it the day it arrives, review it within 24 hours. A revision sitting unread in your inbox is a liability with a timestamp. Even a 10-minute skim that fills in "change summary" beats a full review you never get to.
- The "rev priced" column never changes silently. When you re-price a sheet, you update the column and the date. Your bid letter references this table's state. If the GC later claims you should have caught something in Rev 5, you can show your bid was explicitly based on Rev 4, received after your number was submitted.
Fifteen rows for a typical job, five minutes per revision to maintain, and it eliminates the "which set did we price?" conversation entirely.
Re-Takeoff Strategy: Full Recount vs. Targeted Delta Check
The question with every revision: do you re-take-off the whole set or just check what changed? Neither answer is always right. Here's the decision logic I use:
| Situation | Approach | Time cost |
|---|---|---|
| Narrative addendum only (spec language, no drawing changes) | Read it, log it, no takeoff work | 15-30 min |
| 1-2 sheets revised, changes clearly clouded, details unaffected | Targeted delta: recount only affected sheets, spot-check adjacent details | 1-2 hrs |
| Framing plan revised (member sizes or layout) | Recount that plan fully, plus every schedule and detail sheet it references | Half day |
| 4+ sheets revised, or "reissued for bid" stamp | Full re-takeoff. Treat it as a new set. | Same as original |
| Revision block says "general revisions" with no clouds | Full re-takeoff. That phrase means "we changed things and didn't mark them all." | Same as original |
The trap in the targeted approach is trusting the clouds to be complete. They frequently aren't — clouding is done by a drafter under the same deadline pressure you're under. So even on a targeted delta check, I add two safety passes:
- Recount member schedules end-to-end on any revised sheet, not just clouded rows. Schedules are where unclouded quantity changes hide.
- Compare sheet totals against your baseline. If your original takeoff said S-201 carried 22 tons and your delta check says 22.4, fine. If it says 26, the clouds didn't tell you the whole story.
That baseline comparison only works if you kept per-sheet totals from the original takeoff — one more reason to structure takeoffs by sheet from day one. If you're marking up PDFs by hand, the PDF setup habits that make comparison possible — consistent naming, per-sheet extraction, saved markup layers — pay for themselves the first time a Rev 3 lands.
Clouds, Deltas, and Revision Blocks: Reading What Changed
Quick decoder for what the drawing set is telling you:
- Revision clouds — the scalloped bubbles around changed areas. Useful as a starting point, never as a complete inventory. Clouds from earlier revisions are often left on the sheet, so confirm the delta number inside the cloud matches the current rev.
- Delta triangles — the numbered triangle next to a cloud ties it to a specific revision. A cloud with a Δ2 on a Rev 4 sheet is old news; you priced that already. Only the Δ4s are new work.
- Revision block — the table in the title block listing each rev, date, and description. Read the description with suspicion: "Revised per RFI 034" tells you nothing about tonnage. Pull RFI 034 and read what was actually asked.
- "Issued for" status — IFB (issued for bid), IFP (permit), IFC (construction). If you priced an IFP set and the IFC set arrives, expect changes whether or not anything is clouded. Permit-to-construction is where engineers finish the design.
One habit worth building: check the revision block on every structural sheet when a new set arrives, not just the sheets the transmittal says changed. Transmittals are wrong often enough to matter.
Pricing Addenda Under Deadline: A Triage Checklist
Addendum 3 lands at 4pm, bid is due at 10am. In order:
- Read the narrative first, drawings second. Spec changes (a shift from standard primer to a full paint system, added galvanizing, sequence restrictions) can swing price more than a beam size change, and they're buried in text.
- Scan every revision block, not just the transmittal list. Five minutes to confirm what actually changed.
- Bucket each change: tonnage, labor, or terms. Tonnage changes get recounted. Labor changes (added moment connections, field welding) get hours. Terms changes (schedule compression, retainage) get a management decision, not just a number.
- Recount what you can, plug what you can't — and mark the plugs. If you genuinely can't recount a revised sheet in time, carry an allowance and note it internally. An honest plug you track beats a guess you forget.
- Update the bid letter's document list before you update the price. The revision references in your proposal are your protection; get them right first.
- Compare new total to old total and gut-check the delta. If Addendum 3 clouded half the roof plan and your price moved 0.4%, one of those two facts is wrong.
If an addendum drops less than 24 hours before bid and materially changes structural scope, you also have the option of asking for an extension — GCs grant them more often than estimators ask. And if the revision churn on a job is heavy enough before award, that's a data point for your bid/no-bid decision: a design team issuing four addenda in two weeks will keep behaving that way after contract, and you'll be pricing change orders for the duration.
Protecting Yourself: Qualifications and Exclusions Language
Your bid letter should make the revision basis explicit. Minimum language:
"This proposal is based on structural drawings S-101 through S-501 dated [date], Revision [X], and Addenda 1-3. Changes issued after this date, including revisions to drawings listed above, are excluded and will be priced as a change to the contract."
Three additions that have earned their keep:
- List sheets individually with their rev numbers (or attach your tracking table) rather than saying "latest revisions." "Latest" is an argument waiting to happen.
- Exclude scope shown only on non-structural sheets. Architectural and mechanical sets grow steel — lintels, equipment frames, elevator hoist beams — and revisions to those sets rarely get transmitted to you.
- State that acceptance of post-bid revisions requires written re-pricing. This converts "you should have caught that in Rev 5" from a dispute into a change order, which is where it belongs. The pricing mechanics are covered in how to price steel change orders.
None of this is adversarial. It's the same clarity a good GC wants — they'd rather know your basis now than discover the gap in month four.
How a Fresh AI Takeoff Per Revision Makes Deltas Visible
The reason estimators skip full re-takeoffs is honest: on a manual workflow, a recount costs the same as the original — hours or days you don't have at addendum time. So the targeted delta check, with all its blind spots, wins by default.
That math changes when a takeoff is cheap to repeat. Running a complete fresh takeoff on each revised set — Rev 2, Rev 3, IFC — gives you per-sheet counts you can put side by side against your baseline. The unclouded schedule change and the quiet member upsize show up as a number that moved, whether or not anyone drew a cloud around it. You review the differences instead of hunting for them.
This is the workflow SteelFlo was built for: upload the revised set, get every steel member detected and boxed on every sheet with 95-99% accuracy (AI detection plus your verification), and compare counts against what you priced — in minutes, not another all-nighter. With 4,500+ sections across 6 standards, it handles whatever the revised set throws at it. Addendum lands at 4pm? Re-run the revised set in minutes — your first 3 takeoffs are free.