From clean plans, often yes; from photos alone, no. Tools such as Togal.AI, iBeam and BuildVision AI measure from PDF plans quickly, but the accuracy figures are the vendors' own claims. Run five finished jobs through a tool, compare with what you actually laid, and quote from it only if the errors stay small.
Two caveats shape everything below. A takeoff gives net area, while a flooring quote needs the area you order, which depends on pattern, room shape and roll widths. And no plan shows subfloor moisture, flatness or what is being pulled up, so the site measure never goes away. The test tells you how far to trust the first number.
What each tool reads, what it claims and what it costs
Six tools come up for flooring, and they are not the same kind of product. The claims below are the vendors' own, quoted from their pages in October 2026. None of them is an independent test.
| Tool | What it reads | Claim on its page | Price signal (USD) |
|---|---|---|---|
| Togal.AI | Drawings, detecting and measuring automatically | "up to 98% accuracy"; "98% accuracy on floor plans" | Growth $299 per user a month billed yearly; Business quoted for teams of four or more |
| iBeam | PDF plans and specifications; its home page says it "measures floor areas and material requirements across all flooring types" | "99%+ accurate" takeoffs in 24 to 72 hours, each QA-checked by its staff before delivery | Done-for-you annual licence from $25,000 a year, set by trade and volume |
| BuildVision AI | PDF (vector or scanned), DWG, DXF, JPEG and PNG | Confidence scores and review flags on every takeoff; no general accuracy guarantee | From $299 a month; 14-day full access |
| Metres.ai | Uploaded PDFs and image files, measured on screen | "Measure plans, count materials, and build accurate quotes"; I found no mention of AI on the page | $20 a month or $220 a year; two-week trial |
| MeasureSquare | Flooring-specific takeoff for carpet, sheet vinyl, tile and hardwood; a mobile app for measuring on site | Calls itself "the market's #1 flooring takeoff estimation software"; the mobile app generates seam diagrams and tile layouts | Not shown on the pages I read |
| magicplan | A phone or tablet scan of the room (LiDAR on compatible devices) or manual measuring, turned into a floor plan | No accuracy claim on its pricing page | $250 to $400 a month depending on commitment, with 10 projects a month by default |
Three things stand out. First, every accuracy figure was measured on someone else's drawings, and yours may be scanned, skewed or missing a scale bar. Second, Metres.ai ranks for this phrase, yet its page describes digital takeoff on uploaded plans and I found no mention of AI there. A $20 a month measuring tool may be all a small flooring firm needs. Third, BuildVision AI's FAQ lists pre-trained models for electrical, HVAC, plumbing, roofing, concrete, demolition and general civil work, and offers to fine-tune on your own drawings for other trades, which it says typically takes a five-day onboarding. Flooring is not on the pre-trained list, so ask how it handles finishes.
Photographed or scanned plans are a separate risk from photos of rooms. BuildVision lists scanned PDFs and image uploads, and Metres.ai accepts image files. A photo of a drawing is only as good as its scale calibration, which is where most surprises start. For quoting lines once you have the areas, the tutorial on a flooring quote template with area, waste and underlay picks up where this one stops.
It also helps to ask what "98% accurate" means. A figure like that could describe the share of rooms found, the share of objects counted correctly, or an average area error, and the headline claims I read do not say which. A tool that finds 98% of rooms can still be 10% out on the one with the bay window. Ask the vendor what was measured, on what kind of drawings, and against what truth. Your five-plan test, below, is the answer that counts.
What a plan cannot tell you: the site measure list
Even a perfect area leaves gaps that only a visit fills. Quote from the plan as a budget price, and treat the measured quote as a separate step once the customer proceeds. This is what the visit has to capture.
| Item | Why the plan cannot show it | What to record on site |
|---|---|---|
| Subfloor type, moisture and flatness | Plans rarely state it | Moisture readings, gaps under a straightedge, a repair allowance |
| What is being removed | The plan shows the design, not the old floor | Floor type, adhesive, layers, volume to dispose of |
| Thresholds and height changes | Not drawn | Height at each doorway, the transition type needed |
| Cabinets, islands, appliances, baths | Drawn as symbols with no depth or fixing detail | Whether flooring runs under them or stops at them |
| Stairs and landings | A stair symbol, not a count | Treads, nosing type, riser heights, winders |
| Walls that are not square | Plans are idealised | Diagonals; out-of-square rooms raise waste |
| Pattern start and layout lines | Design intent only | Starting wall, focal line, plank direction |
| Access and furniture | Not shown | Who moves what, how boxes reach the room |
The first two rows decide whether the job is a floor or a floor plus a repair. The tutorial on flooring contract clauses for subfloor and moisture covers how to word the risk once you have found it.
Waste and pattern allowances sit outside the takeoff
Most takeoff tools return net area per room. The allowance is yours to add. The ranges below come from two flooring retailers' ordering guides, which themselves defer to the manufacturer's coverage figures and the installer, so use them as a sense-check against your own job records rather than as a standard.
| Layout or product | Range quoted | Note |
|---|---|---|
| Luxury vinyl plank or tile, straight lay | 5% to 10% | Simple rooms sit at the low end |
| Laminate, straight lay | 5% to 8% | |
| Engineered wood, straight lay | 5% to 10% | Simple rooms at the low end |
| Solid hardwood, straight lay | 8% to 10% | |
| Large-format tile, straight lay | About 10% | 12 in or bigger |
| Planks laid diagonally | 10% to 15% | Up to 17% for solid hardwood |
| Herringbone or chevron | 10% to 20% | One guide gives 15% to 20% for every product |
| Natural stone | 15% to 20% | Small mosaics 12% to 17% |
| Carpet tile / broadloom | About 5% / about 10% | Broadloom depends on roll width |
| Hallways, closets, bays, L-shaped rooms | Higher end of the range | Very small rooms with many angles: 12% to 15% |
Roll goods behave differently. Carpet and sheet vinyl waste depends on roll width and where the seams fall, which a net-area figure cannot show. MeasureSquare's page says its mobile app generates seam diagrams, which is the kind of output to ask any vendor for.
One bedroom shows why. At 13 ft 6 in by 17 ft it is 229.5 sq ft net. On a 12 ft roll run along the 17 ft wall, one drop covers 12 ft of the width and leaves a strip 1 ft 6 in wide. If the pile must run the same way, that strip comes out of a second 17 ft length unless an offcut from another room on the same job is long enough, and two 17 ft drops are 408 sq ft bought for 229.5 laid, 78% over net. Turn the roll the other way and two drops of 13 ft 6 in cover 24 ft of the 17 ft needed: 324 sq ft bought, 41% over, one seam, and a 7 ft offcut that might carpet a cupboard or a landing. Where the range also comes in a 15 ft width, a single 17 ft drop is 255 sq ft, 11% over, with no seam. Same room, three different orders. A flat percentage on net area cannot see any of that, and a seam plan can.
Take a 640 sq ft open-plan room in engineered wood, sold in boxes that cover 20 sq ft. Laid straight at 8%, 640 x 1.08 is 691 sq ft, or 34.6 boxes, so order 35. Laid diagonally at 15%, 640 x 1.15 is 736 sq ft, or 36.8 boxes, so order 37. Two boxes apart in material, but the labour and the cutting are a different job, so pattern belongs on the quote as its own line and not only as a percentage. One of the guides also warns that boards from different production runs can differ slightly in colour tone, so order the whole job in one go.
A five-plan test you can run this week
The test needs five finished jobs where you know what was actually ordered and laid: an invoice, a cut list or a supplier order. Choose a mix. A simple three-room flat, an open-plan floor, a job with stairs, a room with angles or a curve, and a commercial space.
- Pull the plans you were given at the time, including any that were scanned or had no scale bar. Those are the ones that matter.
- Before you run the tool, write down your own net area for each room from the job record.
- Run each plan. Export the room areas and perimeters, and note how many minutes it took.
- Record the area for every room, the total, the percentage gap, any room missed or merged, and the minutes you spent fixing the output.
- Decide the pass mark in advance. Mine would be within 2% on simple plans, within 4% on complex ones, and no missing rooms. Move those to suit your margin: if your net margin is 10%, a 4% area error is 40% of your profit on that job.
A completed sheet, using invented figures, reads as follows.
| Job | Actual net area (sq ft) | Tool result | Gap | What went wrong |
|---|---|---|---|---|
| 1. Three-room flat | 612 | 618 | +1.0% | Merged a cupboard into the hall |
| 2. Open-plan ground floor | 1,140 | 1,128 | -1.1% | Clipped the area under a bay |
| 3. Hall, stairs and landing | 436 | 391 | -10.3% | Stairs not counted; the plan shows a symbol only |
| 4. Angled and curved rooms | 520 | 541 | +4.0% | Curve approximated with straight edges |
| 5. Retail unit | 2,340 | 2,311 | -1.2% | No scale bar; set from a known door width |
| All five | 5,048 | 4,989 | -1.2% | Errors partly cancel in the total |
The last row is the trap. The total is within 1.2% while job 3 is 10% short, because plus and minus errors cancel when you add them up. Judge every job separately. The reading from this sheet: the tool passes on open rectangles and fails on stairs, so use it for room areas and measure stairs by hand. If you want a wider comparison across several products on one plan set, the tutorial on trialling three takeoff tools on one real plan set sets that up.
One more check belongs on job 5. A scale error is squared when it becomes an area error: a drawing 3% out in length is 6% out in area. A PDF that was printed or exported as "fit to page" no longer matches the ratio on its scale bar, and every room is then wrong by the same factor. Calibrate against a dimension you know, such as a door leaf, before you trust any area on a sheet.
Photo and scan measuring: check it against a tape
A photo has no scale. Apps that produce dimensions from a phone use LiDAR scanning where the device supports it, augmented reality, or measurements you type in, and magicplan's page says it offers both LiDAR scanning and manual measuring. The page gives no accuracy figure, so create your own. Scan three rooms you can also measure with a tape, and compare.
| Room | Tape | Scan | Area gap |
|---|---|---|---|
| Living room | 14 ft 2 in x 11 ft 9 in (166.5 sq ft) | 14 ft 4 in x 11 ft 8 in (167.2 sq ft) | +0.5% |
| Bedroom | 12 ft 0 in x 10 ft 6 in (126.0 sq ft) | 12 ft 1 in x 10 ft 7 in (127.9 sq ft) | +1.5% |
| Hall | 22 ft 6 in x 3 ft 4 in (75.0 sq ft) | 22 ft 9 in x 3 ft 7 in (81.5 sq ft) | +8.7% |
In this invented example every scan is within three inches on each side, which sounds fine, and the hall is still 8.7% out because three inches on a three-foot-four width is a large share of the width. A tolerance that is acceptable for a living room fails for a corridor, so test the narrowest room you are likely to quote. If a scan is used for ordering, also measure any room under about four feet wide by tape.
The plan that was not at scale
Imagine a firm quoting a refit from a PDF the client emailed. The client had printed the drawing to A4 and scanned it back, so the scale bar no longer matches the sheet. The tool reads the stated ratio and returns believable room areas, every one of them about 11% too high. The quote loses on price, and nobody questions the numbers because they look plausible.
A twenty-second check of one doorway against its known width would expose it. Put that check on every plan that comes from a third party, and include one plan that has been through a printer or scanner when you run the five-plan test.
Let an assistant compare the takeoff with the room schedule
Once you have the tool's room export, an AI assistant can compare it with the architect's room schedule faster than you can by eye. Keep it to comparing lists. Do not let it do the totals.
Below are two lists for the same flat. LIST 1 is the architect's room schedule. LIST 2 is the room export from our takeoff tool. Areas are in square feet.
LIST 1
Hall 62 | Living room 214 | Kitchen 118 | Bedroom 1 158 | Bedroom 2 121 | Bathroom 54 | Store 24
LIST 2
Hall 71 | Living/Kitchen 331 | Bedroom 1 160 | Bedroom 2 121 | Bathroom 54
Task: (1) List every room in LIST 1 with no match in LIST 2, and every room in LIST 2 that appears to merge two rooms. (2) For each matched room, show the difference in square feet and per cent. (3) Flag any difference over 3%. (4) Do not add up totals; I will do that in a spreadsheet. (5) If a name is ambiguous, say so instead of guessing.An illustrative reply: Store (24) has no match. "Living/Kitchen" (331) appears to merge Living room (214) and Kitchen (118), which total 332, a gap of 1 sq ft (0.3%). Hall 62 against 71 is +9 sq ft (+14.5%), flagged. Bedroom 1 is +2 sq ft (+1.3%). Bedroom 2 and Bathroom match.
What you do with it: check the percentages in a spreadsheet, because language models slip on arithmetic even when the logic is right. Then use your judgement on the findings. The missing store may be correct, since a store room often gets no flooring. The hall is the real question: did the tool include a cupboard, a threshold or part of the living room? That one needs a look at the plan, which is why the assistant flags and you decide.
What a wrong number costs against what the software costs
The subscription tools in the price column run from $220 a year to between $3,000 and $4,800 a year, and iBeam's done-for-you service starts at $25,000 a year. The right question is what one bad number costs you. Take an illustrative 1,800 sq ft job where material costs $4.50 a square foot and installation labour $3.00, so direct cost is $7.50 a square foot.
- Each 1% of area is 18 sq ft, or $135 of direct cost.
- A 4% miss is $540 before any return visit, a second delivery or a colour mismatch from a second batch.
- A tool at $299 a month costs $3,588 a year, which is about seven 4% misses. If you quote 60 jobs a year, that is a miss in one job in nine.
That is an upper limit on what to pay, not a reason to pay it. A tool you have tested and trust on your drawings is worth the money. A tool that needs 40 minutes of fixing per plan may not save you anything. For other price bands across estimating tools, see the tutorial on what estimating software costs.
Keep score on the first ten jobs you quote from it
The five-plan test checks the past. Once you quote from the tool, check the jobs ahead as well: for each of the next ten, log the boxes or rolls ordered, the amount laid, what went back and any second delivery. Three rows from an invented log show the pattern to look for.
- Open-plan flat, engineered wood at 8%: 35 boxes ordered, 33 laid, 2 returned. The allowance held.
- Three bedrooms and a hall in laminate at 8%: 52 boxes ordered, 51 laid, one spare. Tight but fine.
- Angled extension in luxury vinyl at 8%: 41 boxes ordered, 43 needed, a second delivery and half a day lost waiting for it.
One short job in ten is noise; two short jobs that are both angled rooms is a rule. Raise the allowance for that room shape, or measure those rooms by hand, and leave the open-plan setting alone. A steady run of returns says the opposite: the allowance is generous, and the cost is cash tied up in boxes you carry back to the van.
Six questions for a takeoff vendor's demo
- How is scale set, and what does the tool do when a sheet has no scale bar?
- What did the accuracy figure measure, and on which kinds of drawings?
- How does it count stairs, and how does it treat curves and bays?
- Does a person check the output (iBeam says its QA staff do), and what does that do to the turnaround?
- Can I export room by room with perimeters, so the numbers drop into my own estimate sheet?
- What happens to my plans: how long are they stored, and are they used to train the product?
The last question matters most for tenders, where plans may carry a client's confidential details. If the answer is vague, upload a redacted copy for the trial.
Which tool for which kind of job
| Job type | Likely fit | Test first |
|---|---|---|
| Small residential refits, many a month | A cheap manual takeoff tool, or room scanning on site | The narrowest room you quote |
| Open-plan new builds and multi-unit blocks | An AI takeoff tool for the first pass, with you checking room by room | Job with stairs and a bay |
| Commercial tender sets with many sheets | A takeoff service with human QA, or flooring-specific software | Addenda and revised sheets |
| Carpet and sheet vinyl | A flooring-specific layout tool that draws seams | A roll-width case with awkward seams |
For commercial tenders the tutorial on bidding flooring jobs with commercial takeoffs covers the tender side. Whatever you choose, keep the process in this order: plan takeoff for the budget price, site measure for the firm quote, and an allowance set by pattern and product on top of net area.
Flooring takeoff questions, answered
How accurate is AI takeoff from a flooring plan?
Vendors quote figures such as 98% or more, but those come from their own drawings, and I found no independent test. Accuracy depends on the plan: clean vector PDFs with a scale bar do best, scans and hand-marked sheets do worse. Run five finished jobs through the tool, compare each with what you actually laid, and judge job by job rather than on the total.
Can AI measure flooring from a photo of a room?
Not from a photo alone, because a photo has no scale. Apps that give dimensions from a phone use scanning, augmented reality or measurements you enter, and each should state its tolerance. Test a few rooms against a tape. Narrow rooms such as halls show the biggest percentage errors, because a small error in width is a large share of the width.
Do takeoff tools add the waste allowance for me?
Most return net area per room, and the allowance is a rule you apply. Some let you set waste by product or pattern. Check what yours does and whether it handles roll widths and seams for carpet and sheet vinyl. Whatever the software says, the manufacturer's coverage information and your own job history should have the final word on how much to order.
Is a $20-a-month takeoff tool enough for a flooring firm?
For small residential jobs where you check every room anyway, a manual digital takeoff tool may be all you need. AI takeoff earns its price on multi-sheet plan sets, where finding and measuring every room by hand takes hours. Test before you pay for the dearer tier, and keep a site measure in the process either way.
Further reads
- Remodelling AI Estimator: Test It Against Your Own Past Jobs — Testing an AI estimator against your own past remodelling jobs.
- AI Drywall Estimating Software: Plan or Photo Takeoff for a Small Crew? — The same plan-or-photo question for a drywall crew.
- AI for Carpentry: A Site Measure Checklist From Notes and Photos — A site measure checklist built from notes and photos.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: vendor pages for Togal.AI, iBeam, BuildVision AI, Metres.ai, MeasureSquare and magicplan, plus two online flooring retailers' ordering guides (Word of Mouth Floors and Flooring Market). All read in October 2026.