ARKit + LiDAR Business Applications: The 2026 Industry Map

What are businesses actually doing with iPhone LiDAR in 2026? Measuring things. Documenting things. Replacing tape measures, ladders, truck rolls, plaster casts, and $15,000-80,000 fixed scanning stations with a sensor that field crews already carry. The market that works is not "AR experiences" - it is spatial productivity: capture reality fast, turn it into structured data, push it into the system of record.

This post is the map: each vertical below covers the problem being solved, the published evidence, and who is proving it. It is also the hub for our LiDAR series - deep dives will link from here as they publish. For what the sensor can and cannot measure, start with the peer-reviewed accuracy data.

Construction and AEC: scan-to-CAD is the mature market

The problem: the gap between design drawings and as-built field reality. Manual measurement burns hours, misses details, and forces repeat site visits.

The evidence: Twindo (formerly Canvas) sells scan-to-CAD conversion at $0.14-0.22 per square foot and reports over 12,000 customers and 100M+ square feet modeled. Its published customer quotes are the concrete kind worth reading: an architecture firm reporting a 2,500 sq ft building that used to take 2 hours of measuring plus 1.5 hours of data entry now captured in about 30 minutes of scanning; a construction group estimating roughly $1,500 saved per home on measuring and modeling. Polycam reports 2.5B+ square feet digitized in 2024. Matterport - acquired by CoStar Group for about $1.6B in 2025, with $104.2M ARR and 1.2M subscribers in its final independent filings - proves the willingness to pay at portfolio scale. All company-reported figures except the Matterport financials, which are from its public filings.

Insurance and restoration: the strongest commercial fit right now

The problem: claim cycle time, leakage, unnecessary reinspections, dangerous roof climbs, and adjusters re-typing field notes into estimating software.

The evidence: Hover reports around 285 measurements extracted per property from about eight smartphone photos, $300-500 saved per claim, 10-15% more claims handled per adjuster, and adoption across most of the top 10 US property carriers - company-reported, but consistent across years of material. In 2026 Polycam added direct ESX export to Xactimate, the insurance industry's estimating standard, priced at $11.99 per export on top of its subscription. That pricing move is the tell: the deliverable is worth more than the scan. Restoration contractors publish the same pattern - magicplan customer cases report on-site estimates going from 2-3 hours plus an office return trip to about 20 minutes of sketching and 10 minutes of estimating.

Real estate: floor plans at listing volume

The problem: floor plans are proven to help listings, but commissioning them the traditional way is slow and costly per listing.

The evidence: CubiCasa has produced over 5 million floor plans from 5-10 minute phone scans, with cloud processing and human QA behind them, and distributes directly into listing platforms. Matterport reports a ~12% uplift in booking conversion for listings with digital twins (company-reported). This is the vertical where capture is most commoditized - the differentiator is turnaround time, QA, and distribution into the platforms agents already use.

Logistics and warehousing: dimensioning is an early market with hard ROI

The problem: dimensional-weight billing errors and freight reclassification. The US NMFC freight classification system moved to density-based classes in mid-2025, which made accurate dimensions a direct billing issue - and fixed dimensioning stations cost $15,000-80,000 each.

The evidence: vMeasure reports ±0.2 inch accuracy on regular parcels; Optioryx reports ±2-3 cm on pallets, with scans syncing to warehouse systems in under two seconds (vendor figures). One honest caveat a serious product must carry: phone LiDAR is not legal-for-trade - certified billing instruments require NTEP/MID certification the phone does not have. The positioning that works is operational accuracy: catch misclassified freight before the carrier's re-weigh does.

Manufacturing and field service: remote assistance has the one audited ROI number

The problem: expert scarcity. Senior engineers cannot be on every site, and travel plus downtime is the cost.

The evidence: the strongest third-party ROI figure in the entire AR space is a Forrester Total Economic Impact study of Microsoft Dynamics 365 Remote Assist (2020, commissioned by Microsoft): 138% ROI with a 12-month payback. Around it sit vendor-reported cases - DHL Supply Chain reporting training cut from weeks to hours, JigSpace customers reporting fewer parts-identification errors. A quieter workflow is growing underneath: field engineers scan mechanical assets with LiDAR and import the mesh into CAD as a template, designing retrofits without manual measurement.

Healthcare: contactless measurement with clinical validation

The problem: plaster casting for prosthetics and orthotics is slow and messy; contact-based wound measurement is inconsistent and carries infection risk.

The evidence: a peer-reviewed study found LiDAR wound measurement correlated 0.995 with ruler methods on flat wounds, with better consistency between raters. The US Department of Veterans Affairs has described workflows where scanning replaces plaster casts and cuts turnaround by 3-14 days. The honest limit: LiDAR alone struggles with dynamic body contours, so production clinical tools pair it with photogrammetry and ML.

Surveying and geodesy: our own corner of the map

The problem: full survey-grade equipment is expensive and slow to deploy for small jobs, pre-project inspections, and obstructed urban sites.

The evidence: Pix4D publishes cases of pre-survey site inspection cut from about 2 hours to about 20 minutes, and pairing phone LiDAR with external RTK GNSS receivers brings georeferencing under 5 cm (company-reported). This is first-party territory for us: we built and ship the production iOS app for a surveying technology company that replaces dedicated scanning hardware with an iPhone paired to an RTK GNSS receiver - LiDAR capture, correction data streamed from an NTRIP base-station network over Bluetooth, per-frame quality gating, and sub-5 cm georeferenced scans, used by field crews across Europe. The engineering lesson from that project shapes how we read this whole market: the sensor is table stakes; the pipeline is the product.

Interior design and renovation: the bundle beats the scanner

The problem: error-prone existing-condition drafting and slow quoting.

The evidence: magicplan's customer cases are the best published set of field workflows - a restoration firm reporting 30-60 minutes saved per job, another reporting roughly 50% time savings (company-published cases). The lesson generalizes: the best-selling product in this vertical is rarely the best scanner. It is the best capture-plus-estimating-plus-reporting bundle. Workflow integration wins over mesh quality.

Retail and try-on: where camera-only AR still carries the weight

Furniture placement and virtual try-on drive real conversion numbers, but the published figures vary wildly by source and metric - Shopify has reported a ~94% average conversion lift for merchants adding 3D content, while individual brand studies report add-to-cart and order lifts in the 27-65% range. Treat every number as metric-specific; none of them average cleanly. Notably, much of this vertical needs plane detection and face tracking rather than LiDAR - it is the reminder that LiDAR is what moves an app from visualization into measurement, and not every product needs to make that move.

Accessibility: the gap nobody has productized

Apple ships People Detection and Door Detection in the built-in Magnifier, and consumer apps prove LiDAR-based obstacle detection works. What barely exists: B2B deployments for hospitals, senior living, transport hubs, and campuses, packaged the way institutions buy - software plus building profiles plus support plus analytics. We flag it because it is the clearest open gap on the map, and because it overlaps our own assistive-vision R&D.

What is the pattern behind every product that works?

Three rules repeat across every vertical above:

1. The deliverable is the business, not the scan. Hover charges per report, Polycam per ESX export, Twindo per square foot converted. Capture is commoditized by Apple's own APIs; the defensible layer is capture, then structured data, then integration into the system of record - Xactimate, Revit, the WMS, the CMMS.
2. Two-tier architecture is forced by hardware. LiDAR ships only on iPhone Pro models (since the 12 Pro in 2020) and iPad Pro. Healthy products are "LiDAR capture for field agents, universal review for everyone else" - never "LiDAR app for everyone." For B2B this is fine: enterprises standardize hardware anyway.
3. Honesty about limits is a feature. The products with staying power state their accuracy class and their failure modes. The ones that promise "survey-grade in your pocket" generate churn and disputes.

How to read this map for your own product

If you are a founder or CTO circling this space, the questions that matter: which deliverable does your buyer already pay for today, which system of record does it need to land in, and does centimeter-class accuracy at 5 m range actually cover the job - the accuracy data answers the last one.

We build this category of product - ARKit and LiDAR capture, Metal point-cloud rendering, RTK integration, and the unglamorous export pipelines that make the data land where it earns money. If you want a second opinion on scope, feasibility, or the accuracy budget for a specific use case, talk to us or start with our AR and spatial engineering services.