The Laser Scanning Deliverable Spec Template

Without a deliverable spec, a 6mm registration offset across floors can go undetected until ductwork is already ordered for fabrication - by which point rework costs can reach tens of thousands of dollars. This article gives you the spec that prevents that outcome, plus the technical framework behind every line. For a broader view of our 3D laser scanning services and deliverable options, start there first, then come back to lock in your project-specific numbers.

---

Why a Deliverable Specification Belongs in Every Scope of Work

“Provide as-built laser scanning of the existing building” is not a deliverable specification. It tells no one the accuracy tier, point density, coordinate datum, file format, or what constitutes acceptance. Every vendor quoting that language is pricing a different product - and you have no contractual recourse when the data does not support your downstream workflow.

Why a missing accuracy report is a project risk. When a registered point cloud is delivered without a per-station accuracy report, a high-RMS station-pair link can go completely undetected. Consider a scenario where a mechanical room compresses the scan path: two stations positioned on opposite sides of large equipment, sharing only two spherical targets at a shallow approach angle, can produce a link RMS well above the ±2mm threshold required for MEP coordination. Without a per-station report, the BIM manager has no way to identify that weak link. The result can be corridor centerlines on one floor offset from the floor above, discovered weeks into MEP coordination - after ductwork dimensions have already been submitted for fabrication. A spec requiring per-station RMS ≤ 2mm and a printed accuracy report listing every station-pair link at handover surfaces that problem before the model leaves the scanner’s office.

What this template covers: point cloud density and coverage parameters, LOD and LOA targets, registration accuracy thresholds, file format requirements, QA documentation, and a written acceptance window with measurable rejection triggers. For context on how file deliverables feed into downstream workflows, see our complete breakdown of laser scanning file formats and deliverables.

---

Section 1 - Scan Coverage & Point Cloud Density Requirements

Before specifying a number, define the geographic and physical scope. “All accessible areas” is better than nothing but still leaves gaps. A tight spec names floors, zones, and explicit exclusions with room IDs: locked server rooms (e.g., “FL2-101”), roof plant areas without safe access, occupied tenant spaces with restricted-hour windows, and areas behind fixed equipment where scanning is physically blocked. Spell them out with room designations - ambiguity is where scope disputes originate.

Point Density by Use Case

Point spacing is measured at the scan range. At standard resolution, terrestrial scanners such as the Trimble X7 capture approximately 6mm point spacing at 10m - that is the default for most architectural work. Tighten to 3mm for MEP coordination where pipe hangers and duct flanges need to resolve clearly. Go to 2mm or better at the element face for curtain wall replacement where you are dimensioning mullion lines.

Station Spacing and Overlap

Density is not just a resolution dial - it is a function of station spacing. We typically space scan stations every 400-600 sq ft in open floor plates, tightening to one station per 150-200 sq ft in mechanical rooms or congested interiors where sightlines are short.

A concrete example: in a 1,200 sq ft hospital mechanical room with a double-height AHU bank on the north wall, 7 stations are needed to maintain 50% overlap around the AHU and keep every station within 5m of the primary pipe runs. At open spacing (one station per 400-600 sq ft), that same room would get 2-3 stations - adequate for architectural walls, completely inadequate for MEP routing behind equipment.

The Trimble X7 at its standard capture rate (full dome) achieves 6mm point spacing at 10m. Switching to a higher-resolution mode - which increases angular density at the cost of longer scan time per station - achieves 3mm spacing at the same 10m range. Per the Trimble X7 scan settings documentation, standard mode options run 1, 4, or 7 minutes per station depending on the selected profile; confirm the exact scan duration profile with your vendor when specifying resolution, as scan time per station varies significantly by setting. Specify which density profile you require; “high resolution” means nothing contractually.

Require minimum 30% point cloud overlap between adjacent scan stations for reliable cloud-to-cloud registration. In tight mechanical rooms or below-grade utility spaces, specify 50% or higher: repetitive geometry (parallel pipe runs, flat CMU walls) creates false minima in ICP convergence - the algorithm finds a locally optimal alignment that is actually a systematic offset of 4-8mm across a 20m run, and without strong feature variation from dense overlap, there is no geometry to pull it back. That 4-8mm drift is invisible in the registration software’s overall summary statistic but shows up as a clash miss in coordination. A spec requiring 50% overlap and at least 3 physical targets per station pair forces the crew to deploy additional stations in exactly those locations.

For large exterior facades, point density can relax to 10-15mm unless the scope involves glazing replacement, historical restoration, or condition mapping. In those cases, treat the facade as its own scan campaign with 2mm density specifications at the glass line. See Section 5 for the element-by-element checklist.

---

Section 2 - Registration Accuracy: What Numbers to Require and How to Read the Report

Registration stitches individual scan stations into a single coordinate model. Software - Leica Cyclone REGISTER 360, Trimble RealWorks, or FARO SCENE - finds common geometry or physical targets between overlapping scans and aligns them. The residual error after alignment is the registration error. This number determines whether your model is usable.

Acceptable RMS Thresholds

These are registered-model thresholds, not instrument specs. To put instrument specs in context: the Leica RTC360 has a 3D point accuracy of 1.9mm at 10m (Leica RTC360 datasheet) with a range noise of 0.4mm at 10m; the Trimble X7 - the scanner we deploy to the field - is rated at ±2.4mm 3D point accuracy at 10m per the official Trimble X7 specification; the FARO Focus Premium is 2mm at 10m. Each of these instruments is capable of supporting ±2-3mm registration - but only if station spacing, target placement, and environmental conditions are controlled. A vendor who quotes instrument spec as proof of registration accuracy is conflating two different metrics.

For a deeper explanation of how registration works and what failure modes look like, see our post on how point cloud registration works and what errors look like.

What a Registration Report Must Contain

Require the following as a contractual deliverable:

• Per-station RMS error values (not just an overall average)
• Overall mean error and maximum single-station error
• Number of common targets used per station-pair link
• Target type (spherical 145mm, checkerboard, cloud-to-cloud)
• Target distribution map (floor plan showing target locations and which pairs shared targets)
• Software name and version used for registration

Why Cloud-to-Cloud-Only Registration Is Not Acceptable in MEP-Dense Floors

Requiring physical targets is not bureaucratic overcaution - it addresses a specific failure mechanism. Cloud-to-cloud (C2C) registration relies on the ICP algorithm finding distinctive geometric features in the overlap zone. In open atriums with varied geometry, this works well. In MEP-dense mechanical floors - parallel pipe runs at 12-inch spacing, flat duct banks, repetitive CMU walls - the geometry is nearly periodic. The ICP optimizer finds a local minimum that satisfies the convergence criterion (typically 0.5-1mm iteration delta) but is actually offset from the true alignment by 4-8mm across a 20m run. The algorithm has locked onto the wrong pipe in the array. Physical spherical targets, placed at non-repeating locations, anchor the alignment to known coordinates and prevent this failure. A spec that permits C2C-only registration on MEP floors is a spec that permits this failure mode without any audit trail.

Red Flags - Reject If You See These

• Maximum single-station error above 6mm anywhere in the report
• Fewer than 3 common targets per station-pair link
• Registration performed cloud-to-cloud only with no physical target backup in MEP-dense areas (see above)
• Missing target distribution diagram - you cannot evaluate whether targets were clustered in one corner
• Report is a single summary number with no per-station breakdown

Verifying Accuracy After Delivery

You do not need a scanner to check registration quality. Request 5 independent check point (ICP) coordinates as a contractual deliverable - physical points (column centerlines, door jamb returns, plumbing stub-outs) measured on-site but not used in registration. Cross-check those coordinates against the registered cloud in Autodesk ReCap or Leica Cyclone Viewer. Also pull 8-10 field tape measurements of known distances - column bay spans, door opening widths, wall-to-wall dimensions - and compare against the cloud. If more than 2 of those 10 measurements deviate beyond your LOA tolerance, reject the deliverable and require reprocessing.

---

Section 3 - LOD and LOA: Choosing the Right Level for Your Project

These two acronyms measure different things and must both be specified.

LOD (Level of Development) is a BIMForum standard governing what elements are modeled and how much reliable information is attached. LOD 200 means generic shapes at approximate size and location. LOD 300 means elements modeled to actual dimensions with accurate openings. LOD 350 adds connections and interface geometry. LOD 400 is fabrication-ready. For a full comparison, see LOD 200 vs LOD 300 in scan-to-BIM projects and our scan-to-BIM LOD guide with accuracy tiers.

LOA (Level of Accuracy) is a USIBD Level of Accuracy (LOA) Specification governing geometric fidelity - how closely modeled geometry matches real-world conditions. LOD and LOA are independent axes: a model can be LOD 300 (fully modeled) but LOA 20 (inaccurate) if the scanning or modeling was careless.

LOA tolerances per the USIBD LOA Specification.

Matching LOD to LOA for Scan-to-BIM

The Cost of Mis-Specifying LOD 400

The most common and costly mis-specification on renovation permit scopes is LOD 400. Here is what that mis-spec actually costs on a 50,000 sq ft office renovation, broken down:

The 40-60% cost increase on a LOD 400/LOA 40 spec versus LOD 300/LOA 30 is driven primarily by independent verification labor and the additional modeling time required to achieve ±2mm geometric fidelity on every element face. For a permit set, the architect of record cannot use LOD 400 geometry in any way that LOD 300 does not already support. The extra spend produces no additional permit-deliverable value.

LOA 40 also has a contractual obligation that LOA 30 does not: independent check shots - not just registration residuals - must confirm accuracy at every floor. A single-pass field scan without ICP verification cannot contractually guarantee LOA 40 regardless of the instrument used.

---

Section 4 - File Format Requirements: Specify Every Format You Will Actually Use

Format disputes at project closeout are almost always caused by a scope of work that said “point cloud data” and nothing else. Every deliverable format should be named, versioned, and tied to a specific downstream use.

Point Cloud Formats

Rule: Always require E57. It is instrument-agnostic, open-standard, and will remain readable in 10 years when the vendor’s proprietary software has cycled through three versions. A vendor who delivers only .fls or .tzf without E57 is locking you into their software ecosystem.

Coordinate System Specification

State the datum, units, and benchmark explicitly:

• Datum: State Plane (specify zone), NAD83, WGS84, or local project grid
• Units: US Survey Feet or International Feet or Meters - do not leave this ambiguous
• Benchmark: Provide easting, northing, and elevation of the project benchmark with a description of the physical monument

Leaving coordinate system unspecified means you may receive a point cloud in an arbitrary local coordinate system that does not align with your civil drawings, your Revit shared coordinates, or your GIS basemap.

Derived Deliverables

2D drawings and BIM models derived from the point cloud have their own format requirements that are frequently under-specified.

The section count (“minimum 3 building sections”) and elevation coverage (“all exposed faces”) above are contractual floors, not standards. Your actual spec should name section cut locations by grid line and designate which faces by compass orientation. Specifying “3 sections” without grid line designations lets the vendor cut through the three simplest, clearest floors and skip the congested mechanical levels. Tie every derived deliverable to a specific use: “sections at Grid Lines B, D, and F for structural engineer’s bearing-wall assessment.” That language is enforceable; “minimum 3 sections” is not.

Specify file naming convention in the contract: PROJECT-FLOOR-SCANDATE-VERSION.rcp. Multi-phase projects with uncontrolled file naming create version confusion when a floor is rescanned after construction. See our discussion of the scan-to-Revit workflow from point cloud to finished BIM model for how these formats move through the production pipeline.

---

Section 5 - Scope Matrix: Element-by-Element Checklist

Copy this table into your RFP. Mark each element, set LOD and LOA targets, and add notes for project-specific conditions. Leave nothing implied.

For clash detection tolerances and how element scope interacts with clash settings, see clash detection tolerances and settings for point cloud models.

---

Section 6 - QA/QC Documentation Package: What to Require at Handover

The point cloud and model are not the only deliverables. The QA package is equally important and almost always omitted from low-bid scopes. Require all of the following:

1. Registration Accuracy Report (covered in Section 2). Non-negotiable. Per-station RMS, overall mean and max, target log, target distribution map, software version.

2. Scan Station Location Plan. A floor plan or site plan with every scan station plotted, labeled with station ID and scan sequence. This lets you audit coverage gaps - if there is no station within 8m of a mechanical room, you know immediately why that corner is sparse and whether it is within spec. Require one plan per floor and one per scanned exterior face. A submission without station plans cannot be audited; you are taking the vendor’s word on coverage.

3. Target Log. A spreadsheet listing every spherical or checkerboard target: target ID, XYZ coordinate in project space, and which station pairs it linked. This is your audit trail for tracing a registration error to its source station-pair link - the same information that enables you to identify a weak link before the model is used for fabrication.

4. Data Collection Log. Date and time of each scan session, operator name, weather conditions (temperature, wind speed for exterior scans), instrument serial number, and last factory calibration date. Temperature can affect instrument performance, particularly on exterior scan sessions conducted near or below freezing - any exterior scan session in cold conditions should be flagged and the data evaluated for additional systematic offset before registration. A log without instrument serial number and calibration date cannot be cross-referenced against the calibration certificate.

5. Instrument Calibration Certificate. Leica, Trimble, and FARO all offer annual factory calibration programs. Require a certificate showing factory calibration within the past 12 months as a standard deliverable.

6. Model-to-Cloud Deviation Report (BIM deliverables). For any Revit or BIM deliverable, require a color-mapped cloud-to-mesh deviation report generated in Autodesk ReCap, Leica Cyclone, or a comparable tool. The heatmap must use: Green (within LOA tolerance) / Yellow (1-2× tolerance) / Red (exceeding 2× tolerance). Specify that 95% or more of sampled surface points must fall in the green band at the stated LOA tier. A LOA 30 project means 95%+ of sampled points within ±5mm of the Revit model geometry.

7. Software Version Log. Document which version of Revit, Cyclone REGISTER 360, ReCap Pro, FARO SCENE, or CloudCompare was used to process and deliver the data. A Revit 2024 model will not open in Revit 2020. Lock this down before delivery.

---

Section 7 - Acceptance Criteria and Rejection Triggers

A deliverable spec with no acceptance window is unenforceable. Define the review period and the measurable criteria in the contract.

Review window: Client has 10 business days from delivery to either (a) accept in writing, (b) submit a written revision request with documented deficiencies, or (c) reject with documented reasons. Silence does not constitute acceptance.

Measurable acceptance criteria:

1. Registration RMS within specified threshold (e.g., ±3mm for architectural, ±2mm for MEP)
2. No single-station registration error exceeding 6mm
3. No coverage gap larger than 1m × 1m in any occupied or in-scope area
4. Model-to-cloud deviation: ≥95% of sampled surface points within LOA tolerance on the heatmap
5. All QA/QC documents listed in Section 6 present and complete
6. All files delivered in specified formats, named per convention, and openable in specified software versions
7. Coordinate system verified by comparing at least 3 control points against client-provided benchmarks

Revision rounds: Specify two rounds of revision included in the base contract. Revision for vendor error (data fails acceptance criteria) is at vendor cost. Revision for scope change or client-directed modifications to element scope is a change order.

Independent verification without a scanner:

1. Open the RCP file in free Autodesk ReCap Pro - measure 8-10 point-to-point distances between identifiable features (column centerlines, door opening widths, wall-to-wall spans). Compare against field tape. More than 2 out of 10 deviating beyond LOA tolerance = rejection trigger.
2. Request 5 ICP coordinates. Have your GC or owner’s rep measure those same points on-site with a calibrated tape or total station. Compare. Deviation must be within LOA spec.
3. Review the model-to-cloud deviation heatmap. Red patches over structural elements, column faces, or MEP equipment indicate model errors, not point cloud errors - require remodeling of those elements.
4. In Revit, use the Spot Elevation or Spot Coordinate tool against the point cloud background - compare spot elevations on slabs at 5-6 locations per floor.
5. In CloudCompare (free, open source), load the E57 file and compare two adjacent floors by aligning a known vertical feature - a stairwell or elevator shaft - and measuring drift. More than 3mm of drift across floors on a LOA 30 project is a rejection trigger.

Escalation clause: If delivered data fails acceptance criteria on two successive submissions, the client reserves the right to engage an independent third-party QA review at the vendor’s cost, not to exceed $5,000.

---

Section 8 - Putting It All Together: The One-Page Spec Summary Block

Two filled-in spec blocks, ready to paste into your RFP. Adjust values for your project.

---

Example A: 3-Story Office Renovation (Scan-to-BIM)

PROJECT TYPE:       3-story office building renovation, 42,000 sq ft COVERAGE:           All floors, occupied and unoccupied; excludes server room FL2-101                     and roof mechanical screen (no safe access) POINT CLOUD FORMAT: RCP/RCS (primary) + E57 (archive) - both required COORDINATE SYSTEM:  Maryland State Plane, NAD83, US Survey Feet                     Benchmark: SE corner Column B4, E=1,423,850.22, N=529,104.77, Elev=124.35 POINT DENSITY:      6mm @ 10m for architectural; 3mm @ 10m for mechanical rooms FL1-FL3 SCAN STATION SPACING: Max 500 sq ft per station (open areas); max 200 sq ft (mechanical) OVERLAP REQUIREMENT: Minimum 30% adjacent-station overlap; 50%+ mechanical rooms REGISTRATION ACCURACY: ±3mm RMS per station; max single error ≤6mm SOFTWARE:           Leica Cyclone REGISTER 360 v2024 or Trimble RealWorks v12+ LOA TARGET:         LOA 30 (±5mm) BIM DELIVERABLE:    Revit 2024 (.rvt); LOD 300 throughout ELEMENT SCOPE:      Structural, architectural envelope, interior partitions, exposed MEP                     (ductwork 8-inch+, piping 2-inch+, conduit 1-inch+), equipment pads                     Above-ceiling: included FL1 mechanical room only (ACT removed by GC) EXCLUSIONS:         Furniture, buried utilities, below-slab, proprietary equipment internals QA DOCUMENTS:       Registration report (per-station), scan station plan, target log,                     data collection log, calibration certificate, model-to-cloud heatmap DEVIATION STANDARD: ≥95% sampled surface points within ±5mm on heatmap ICP CHECK POINTS:   5 pre-measured ICPs provided with delivery FILE NAMING:        PROJ-[FLOOR]-[YYYYMMDD]-[VER].rcp (e.g., OFFICE-FL2-20240315-V1.rcp) DELIVERY SCHEDULE:  Phase 1 (FL1-FL2 point clouds + registration report): Day 14                     Phase 2 (FL3 + complete Revit model + QA package): Day 28 ACCEPTANCE WINDOW:  10 business days from each phase delivery BUDGET CONTEXT:     LOD 300/LOA 30 interior modeling typically $0.06-$0.10/sq ft on top                     of field scan cost; total project range $18,000-$28,000 for 42,000 sq ft

---

Example B: Curtain Wall / Glazing Replacement (Point Cloud Only)

PROJECT TYPE:       12-story curtain wall replacement, south and east facades COVERAGE:           South facade (420 lf) and east facade (180 lf), ground to roof                     All mullion lines, spandrel panels, window openings POINT CLOUD FORMAT: E57 (primary) + LAS (for GIS/civil coordination) - both required                     No proprietary format accepted as sole deliverable COORDINATE SYSTEM:  Virginia State Plane South, NAD83, US Survey Feet                     Benchmark: monument at SW building corner, brass benchmark disk POINT DENSITY:      2mm at facade face (capture mullion reveals, gasket lines) SCAN METHOD:        Terrestrial scanner from lifts at 3 height intervals + ground stations                     Trimble X7 for upper floors if lift access is limited REGISTRATION ACCURACY: ±1.5mm RMS per station; max single error ≤4mm INDEPENDENT CHECK SHOTS: Required - minimum 8 ICP coordinates on facade, measured by                     total station, provided with QA package LOA TARGET:         LOA 40 (±2mm) - fabrication-grade BIM DELIVERABLE:    None (point cloud only); optional 2D DWG elevations if requested EXCLUSIONS:         Interior, roof, mechanical penthouses, east entry canopy QA DOCUMENTS:       Registration report, scan station plan (elevation-view), target log,                     ICP comparison table (measured vs. cloud), calibration certificate DEVIATION STANDARD: All 8 ICPs within ±2mm of point cloud coordinates FILE NAMING:        FACADE-[FACE]-[ELEVATION_BAND]-[YYYYMMDD].e57 DELIVERY SCHEDULE:  All facades + QA package: 10 business days from field completion ACCEPTANCE WINDOW:  10 business days from delivery BUDGET CONTEXT:     Facade point cloud at LOA 40 typically $8,000-$18,000 depending on                     access complexity; lift mobilization billed separately if required

---

For context on how these budget figures break down and what drives cost at different accuracy tiers, the scan-to-BIM LOD guide with accuracy tiers covers the full cost-to-LOD relationship.

---

FAQ

What point cloud density do I need for scan-to-BIM?

Use this as your baseline:

Density is a function of both scanner resolution setting and station spacing - not just the instrument spec. The Trimble X7 at full capture rate (standard dome) achieves 6mm spacing at 10m; to achieve 3mm at the same range, switch to a higher-resolution scan mode, which increases scan time per station. Per the Trimble X7 scan settings documentation, standard mode scan duration options are 1, 4, or 7 minutes per station depending on the selected profile - confirm the specific profile and duration with your vendor when specifying density. Stating only “high resolution” in your spec is not sufficient; state the point spacing at range and the scanner profile required.

What should I look for in a point cloud registration accuracy report?

A legitimate report contains: per-station RMS error values (not just a global average), overall mean and maximum error, number and type of targets used per station-pair link, and a target distribution map showing spatial coverage. Acceptable thresholds: ±3mm RMS for architectural, ±2mm for MEP. Reject if: maximum error exceeds 6mm anywhere, fewer than 3 targets link any station pair, registration was performed cloud-to-cloud only in MEP-dense areas without physical target backup, or the distribution map is missing. Leica Cyclone REGISTER 360, Trimble RealWorks, and FARO SCENE all produce exportable PDF reports containing this information.

How do I verify point cloud accuracy after delivery without a scanner?

Three methods, no scanner required: (1) Open the RCP file in free Autodesk ReCap Pro and measure 8-10 distances between identifiable points - compare against field tape; more than 2 out of 10 deviating beyond LOA tolerance is a rejection trigger. (2) Request 5 ICP coordinates from the vendor and have your GC verify them on-site with a calibrated tape or total station - deviation must be within your stated LOA tolerance. (3) Review the model-to-cloud deviation heatmap - green must cover at least 95% of sampled surface area within LOA tolerance. All three should be standard contractual deliverables on any LOA 30+ project.

What file formats should I require in my laser scanning deliverable spec?

Always require E57 as your open-standard archive - instrument-agnostic and readable long-term. For Revit-based workflows, also require RCP/RCS. For civil or GIS integration, add LAS or LAZ. Never accept only a proprietary format (.fls for FARO, .tzf for Trimble) as the sole deliverable - it locks you into that vendor’s software ecosystem. For BIM models, specify the Revit version explicitly - a model saved in Revit 2025 format may not open in an older license.

What is the difference between LOD and LOA in scan-to-BIM specifications?

LOD (Level of Development, BIMForum) governs modeling completeness - what elements are in the model and how much reliable information is attached. LOA (Level of Accuracy, USIBD LOA Specification) governs geometric fidelity - how closely modeled geometry matches real-world conditions. They are independent axes and both must be specified. A model can be LOD 300 (fully modeled to actual dimensions) but LOA 20 (inaccurate) if the scanning or modeling was performed carelessly. For 90% of renovation and tenant-improvement work, LOD 300 + LOA 30 is the correct combination. Requiring LOD 400 on a permit-set project adds $24,000-$32,000 in field and modeling cost on a 50,000 sq ft scope for no additional permit-deliverable value - it is the single most common and costly mis-specification in renovation permit scopes.

Why do I need physical scan targets - can’t the software register scans automatically?

Cloud-to-cloud registration works well in environments with geometric variety - open atriums, complex facades, stair towers. It fails predictably in MEP-dense mechanical floors where repetitive pipe runs and flat duct banks create nearly periodic geometry. The ICP algorithm converges on a local minimum - it finds what looks like a good alignment but is actually offset 4-8mm across a 20m run because it locked onto the wrong pipe in the array. Physical spherical targets at non-repeating locations anchor the alignment to known coordinates and prevent this. The cost difference is minimal (a standard target set for a floor costs 20-30 minutes of crew time to deploy); the difference in reliability on MEP floors is significant. Require physical targets with a minimum of 3 per station-pair link on any MEP coordination scope.

---

Ready to Lock In Your Spec?

Send us your drawings or RFP and we will pre-fill the deliverable specification for your exact scope - element-by-element, with your LOD/LOA targets, coordinate system, file formats, and QA requirements - and quote against it within 24 hours. You get the same specificity as Examples A and B above, calibrated to your building type, floor count, and downstream workflow, before you sign anything.

Get a quote from Capture - deliverable spec included