CubiCasa vs Magicplan vs FloorScan: 2026 comparison
"CubiCasa, Magicplan or FloorScan: which one?" It is one of the most asked questions by French and European engineering firms in 2026. The honest answer: these are three very different products, sharing the words "AI" and "plans" in their marketing but solving three distinct problems. CubiCasa scans a property with a smartphone and generates a dimensioned plan in minutes — an on-site capture tool for real estate agents and inspectors. Magicplan does roughly the same with a more tradesperson-oriented UI and an integrated quote module. FloorScan starts from a PDF of an existing plan and extracts structured data — a back-office processing tool for estimators and architects. This comparison details the six criteria that tilt the choice: input, accuracy, exports, pricing, language, compliance.
Criterion 1: input (mobile vs PDF)
CubiCasa and Magicplan share a mobile approach: you open an iOS or Android app on site, you capture walls and rooms (Magicplan uses camera and odometer, CubiCasa can combine photo and LiDAR on iPhone Pro), and you get a dimensioned plan out. It is the ideal solution when you have physical access to the property and no existing plan — typically, a real estate agent publishing a listing plan, or a tradesperson quoting on site.
FloorScan, conversely, assumes you already have a PDF of the plan: sent by a client, scanned from archives, exported from an old BIM. You do not travel, you work at the desk on a received file. The input is radically different: a PDF (vector or raster) instead of a video capture session. The two modes are complementary more than competing — a property expertise firm typically uses CubiCasa in the field AND FloorScan to analyse archive plans of the same building.
Criterion 2: real accuracy on complex plans
CubiCasa advertises ±2% surface accuracy on residential apartments up to 200 m². Internal tests confirm: on simple plans (rectangular, no alcoves), error stays under 3%. Above that — L-shaped plans, rooms with columns, mezzanines — error climbs to 5-8%. Consistent with their target: real estate listings where displayed sqm tolerates a modest gap.
Magicplan posts similar numbers in photo mode, slightly better in LiDAR mode (iPhone Pro/iPad Pro), but its main strength is elsewhere: on-the-spot correction UI is fast and the integrated quote module is unique on the market. Typical wall accuracy: ±3 cm in LiDAR, ±5 cm in photo only.
FloorScan does not measure a physical property but reads a PDF, so its accuracy is bounded by the source plan's. On a properly dimensioned architect plan, you get ±1 cm on openings and ±1% on surfaces. On a poor-quality scan, the manual calibration module (entering a known dimension) adds 30 seconds but cuts the error to 1-2%. Detection mAP is 95% on standard European conventions.
Criterion 3: exports (DXF, Excel, BIM)
This is the most discriminating criterion for AEC use. CubiCasa offers JPG, PNG and DXF exports, but the DXF lands on a single layer with no semantic hierarchy. For real estate agency use (publishing a plan in a listing), it is largely sufficient. For real CAD use, you will redraw afterwards. No native Excel quantities export.
Magicplan exports to PDF, JPG, DXF, PNG and — rarely — a specific format for their integrated quote module (price per m² per room). Also no separate layers in the DXF.
FloorScan is designed for structured export: DXF with separate layers (Concrete, Partitions, Doors, Windows, Footprint, Rooms) in real meters with $INSUNITS=6, multi-sheet Excel with one row per room and perimeter/surface calculations, annotated PDF for client delivery, and JSON for BIM or ERP integration. It is the difference between a tool built for downstream CAD/BIM operations and a tool built for real estate listings.
Criterion 4: pricing, licensing, European compliance
CubiCasa works in scan bundles: roughly $10 to $25 per vector plan after uploading the mobile session, with volume tiers. No mandatory monthly subscription — you pay per use. US hosting (Floored / CubiCasa was Finnish then acquired by Roper) with a DPA addendum for European clients.
Magicplan offers a monthly subscription from $9.99 (Sketch) to $24.99 (Pro) per user. No pay-per-use, unlimited scan access. European hosting (Germany) for EU clients.
FloorScan is positioned as a SaaS subscription: Free (3 plans/month, 1 export), Pro ($31/month, 50 plans), Business ($109/month, unlimited, priority support). European hosting (France/Belgium), native GDPR compliance, your plans never feed AI training. Total cost of ownership for a firm processing 30 plans/month: ~$820/year with FloorScan Pro, ~$330-820/year with CubiCasa depending on volumes, ~$400/year with Magicplan — but these numbers compare products that do not do the same thing.
The bottom line: which tool for which need
Pick CubiCasa if you are a real estate agency, insurer or inspector needing to produce a dimensioned plan quickly from a physical visit, with no pre-existing plan. Reasonable cost per scan, quality largely sufficient for listing use.
Pick Magicplan if you are a tradesperson (painter, plumber, electrician) who wants to survey on site AND quote on the spot. The integrated quote module is unique and changes the daily ROI — 30 minutes saved per site on admin work.
Pick FloorScan if you are a construction estimator, architect, surveyor or engineering office receiving PDF plans of existing buildings to analyse. Structured CAD/BIM exports, European GDPR compliance, predictable subscription pricing. Ideal for industrialising large-volume plan processing.
In all other cases (hybrid setups, multi-discipline teams), consider combining two tools: CubiCasa or Magicplan for the field, FloorScan for back-office. The table above summarises the typical case for each persona.
"CubiCasa vs Magicplan vs FloorScan" is a badly framed question — the real one is "on-site capture or back-office processing". Once that is clarified, the choice narrows to one or the other of the matching pair. For large firms covering both use cases, investing in two tools rather than one is solidly net positive over the year. None of the three is bad; they are optimised for different moments of the AEC chain. The worst mistake in 2026 is still picking a mobile tool to do back-office work (useless DXF for CAD) or a PDF tool to do on-site capture (you do not have the plan).