How to Check a Steel Takeoff Before You Bid
To check a steel takeoff, run five checks in order: every structural sheet was covered, plan counts match the schedules, total tonnage per square foot fits the building type, nothing was counted twice on enlarged plans, and every "typical" note was expanded. It takes about 15 minutes and catches most errors that cost bids.
Why does every takeoff need a second look?
Because the person who did the takeoff is the worst person to check it. You read the plans the way you read them the first time, and you skip the same things. A second look, with a fixed routine, finds the errors that a re-read misses.
The errors that hurt aren't usually arithmetic. They're coverage and interpretation:
| Error type | Typical cause | Effect on the bid |
|---|---|---|
| Missed sheet | Partial set, late addendum, roof framing on an architectural sheet | Short by a whole system |
| Double count | Member shown on overall plan and enlarged plan | Overbid on every affected area |
| "Typical" not expanded | "W16x26 TYP" counted once instead of per bay | Badly short |
| Wrong size read | W16x26 vs W16x36, L4x4x1/4 vs L4x4x1/2 | Tonnage drift that's hard to spot |
| Schedule ignored | Columns or lintels only listed in a schedule | Missing pieces |
| Existing steel counted | "(E)" or "EX." members not removed | Overbid |
Common steel takeoff mistakes and manual takeoff errors go into each one in detail. This post is the routine to catch them before the number goes out.
If you'd like a takeoff where every counted member already points to its spot on the drawing, run your next set through SteelFlo. Your QA pass becomes checking boxes, not recounting. The first 3 takeoffs are free.
Step 1: Did you cover every sheet?
Time: 3 minutes.
Open the drawing index and the takeoff side by side. For every sheet, mark one of three things: counted, checked and no steel, or not applicable.
- S-sheets. Every foundation plan (for embeds and anchor rods), framing plan, roof plan, elevation, section and detail sheet should be counted or explicitly checked.
- Schedules. Column schedules, lintel schedules, beam schedules and brace schedules are often on their own sheet. They count as their own source.
- A-sheets with steel. Canopies, stairs, rails, ladders, roof screens, and bent plate edge angles often live only on architectural sheets. If misc metals are in your scope, check them.
- Addenda and ASIs. Confirm the takeoff is on the latest revision of each sheet. Check the revision block, not the file name.
If your index says S-101 through S-501 and your takeoff doesn't mention S-302, find out why before you move on. The commercial takeoff checklist has a sheet-by-sheet list for bigger jobs.
Step 2: How do you check counts against schedules and elevations?
Time: 4 minutes.
Schedules and elevations are independent sources for the same members. Use them to check the plan count.
Columns. Count columns on the foundation plan (one per anchor rod group). Compare to the column schedule, by mark. Then compare to the takeoff. If the foundation plan has 42 columns and your takeoff has 38, you've missed some, or some are on a different plan.
Braces. Count brace frames on the plans, then check each frame on the bracing elevations. A chevron frame is two braces per level, an X is two, a single diagonal is one. Multiply by levels.
Lintels. If there's a lintel schedule, count openings on the architectural plans and elevations and compare. Lintels are one of the most commonly missed items because they live on A-sheets.
Beams per bay. Pick two or three typical bays, count beams in each by hand, then check that the takeoff's count for those bays matches. If the typical bay has 4 infill beams and there are 24 typical bays, you should see roughly 96 of that size before exceptions.
A quick reference for what to compare:
| Member | Primary source | Cross-check against |
|---|---|---|
| Columns | Framing plan | Foundation plan, column schedule |
| Beams and girders | Framing plans | Beam schedule, typical bay count |
| Braces | Framing plans | Bracing elevations |
| Lintels | Lintel schedule | Architectural elevations, openings |
| Embeds and anchor rods | Foundation plan | Column count × rods per base plate |
| Misc metals | Architectural sheets | Specs section 05 50 00 |
Step 3: Does the tonnage make sense for the building type?
Time: 2 minutes.
Divide total structural steel weight by the gross floor area. Compare to typical ranges for the building type. These are rules of thumb for checking, not design values; the engineer of record sizes the members.
| Building type | Typical structural steel (lb per sf) |
|---|---|
| Single-story warehouse | 5 to 8 |
| Single-story retail or flex | 6 to 9 |
| Low-rise office, 2 to 5 stories | 8 to 12 |
| Mid-rise office | 10 to 14 |
| School or institutional | 7 to 11 |
| Hospital or lab | 12 to 18 |
| Heavy industrial | 15 to 25 or more |
Example: a 3-story office, 60,000 sf gross, with a takeoff total of 195 tons. 195 × 2,000 ÷ 60,000 = 6.5 lb/sf. That's below the 8 to 12 range for a low-rise office. Something is probably missing: a framing level, a schedule, or a lateral system. Go back to Step 1.
If it lands at 16 lb/sf, look for double counts (Step 4) or heavy transfer girders that explain it.
Out-of-range doesn't always mean wrong. Long spans, heavy roof loads, vibration criteria or a lot of misc steel can push it up. But it always means "explain it before you bid." Our steel per square foot rules of thumb has the full table, including bay size effects.
Also check average piece weight: total weight divided by piece count. For typical commercial framing you'd expect roughly 500 to 1,500 lb per piece. A much lower average often means a lot of small misc pieces or a double counted schedule of light members. A much higher one can mean missed infill.
Step 4: How do you catch duplicates, enlarged plans and "typical" notes?
Time: 4 minutes.
This step catches the most expensive errors in both directions.
Enlarged plans. Stair cores, elevator pits, mechanical areas and entrance canopies often appear on the overall framing plan and again on an enlarged plan at a bigger scale. Count them once. Look for a clouded or hatched area on the overall plan with "SEE ENLARGED PLAN" or a reference to another sheet.
Partial plans and repeated floors. A "TYPICAL FLOOR FRAMING PLAN" may cover levels 2 through 5. Make sure the takeoff multiplies it by the right number of floors, and that level-specific plans (Level 2 with a different atrium, for example) aren't also counted on top of the typical.
Typical notes. Search the general notes and plan notes for "TYP," "TYPICAL," "U.N.O.," and "SIMILAR." Each one is a rule that applies to many places. "L4x4x3/8 EDGE ANGLE TYP AT ALL SLAB EDGES" is a perimeter measurement, not one piece.
Existing steel. On renovation work, members marked "(E)," "EX.," or "EXISTING" usually aren't in your scope. Make sure they're out of the count.
Sections and details. A beam shown in plan and cut again in a section is still one beam. If your takeoff counts labels in details, check that detail callouts aren't adding pieces.
Reviewing an AI takeoff: what should you look at first?
The same five steps apply. What changes is where you spend time. With a hand takeoff, you're checking the counting. With an AI takeoff, the counting is done; you're checking what it found and what it decided.
Look at these first:
- Anything the tool flags. Good tools tell you where they're less sure: unusual sizes, members found on detail sheets, repeated sheets. Start there.
- Sheets with zero members. If a framing plan came back empty, open it. Either it really has no steel, or the tool couldn't read it. Tools tend to do best on vector, CAD-exported PDFs.
- Your biggest line items. Spot check the five sizes that make up most of the tonnage. Click into a few boxes on each and confirm they're on real members.
- Repeated plans and existing steel. Confirm the tool treated enlarged plans and "(E)" members the way you would.
- Tonnage sanity. Run Step 3 on the AI total, same as a hand takeoff.
This is where review speed comes from. In SteelFlo, each member is a box on the drawing and flagged items come first, so the QA pass is clicking through the questionable ones and confirming or rejecting them. For more on how AI fits into the estimating process, see how to reduce estimating errors with AI.
Printable QA checklist
| # | Check | Time | Pass? |
|---|---|---|---|
| 1 | Every S-sheet marked counted or checked-no-steel | 2 min | |
| 2 | Schedules, A-sheet misc steel and latest addenda included | 1 min | |
| 3 | Column count matches foundation plan and schedule | 1 min | |
| 4 | Brace count matches bracing elevations | 1 min | |
| 5 | Lintels match openings and schedule | 1 min | |
| 6 | Two typical bays hand-counted and match | 1 min | |
| 7 | lb/sf within range for building type (or explained) | 1 min | |
| 8 | Average piece weight looks reasonable | 1 min | |
| 9 | Enlarged and partial plans not double counted | 2 min | |
| 10 | Typical floors multiplied correctly | 1 min | |
| 11 | All "TYP" and "U.N.O." notes expanded | 1 min | |
| 12 | Existing steel removed | 1 min | |
| Total | about 15 min |
Print it, or paste it into your estimate template. The value comes from doing it every time, including on the jobs you're sure about.
Contract scope matters too. AISC 303, the Code of Standard Practice, defines what's normally included as structural steel versus other items, which affects what belongs in your count at all. For section weights, use the current AISC Shapes Database rather than memory.
A 15-minute check is cheap. The BLS reports a median cost estimator wage in the high $70,000s (May 2025 data; check the current figure). At roughly 2,080 hours a year, that's about $38 an hour, or under $10 in wages for a 15-minute check, against an error that can be worth thousands on one bid.
Frequently Asked Questions
How long should a takeoff review take?
A structured QA pass on a typical commercial job takes about 15 minutes. Larger jobs with many schedules or levels may take 30 to 45 minutes. It should be a fixed routine, not a re-do of the takeoff.
What is a good lb per square foot for a steel building?
It depends on building type. Single-story warehouses often run 5 to 8 lb/sf, low-rise offices 8 to 12, and hospitals 12 to 18. Use the range as a sanity check, and explain any result outside it before you bid. Engineering requirements set actual tonnage.
How do I avoid double counting on enlarged plans?
Look for areas on the overall plan that reference another sheet, such as "SEE ENLARGED PLAN" or a hatched area. Count those members only on the enlarged plan. Also check that typical floor plans and level-specific plans aren't both counted for the same level.
Should the person who did the takeoff check it?
Ideally, no. A second estimator or a chief estimator catches more. In a small shop where that isn't possible, use a fixed checklist and do the review after a break, so you aren't reading the plans the same way you just did.
Do AI takeoffs still need to be checked?
Yes. AI speeds up counting, but someone still has to confirm what was found and decide scope questions like existing steel and repeated plans. A good AI tool makes this faster by showing where every member came from and flagging uncertain items first.
Bottom line
Coverage, counts, tonnage, duplicates, typicals. Run those five checks on every takeoff and most bid-killing errors get caught before the number goes out. SteelFlo is built around that review step: every member is a box you can see on the drawing, with flagged items up front, so the check is clicks instead of a recount.