How to Measure Repeatability in a Drive-Through Body Damage Scanner

Autor: NTA    Time: 2026-08-28 17:09:07    Click:

Explains how to design a repeatability pilot for drive-through body damage scanners: same-vehicle rescans, controlled variables, panel-level agreement, and buyer-defined acceptance thresholds, with Dragate Arch Scanner positioned as a strong fit due to its fixed capture geometry and structured reporting.

A drive-through body damage scanner earns trust the moment it reports the same finding on the same vehicle twice in a row, and loses it just as fast when it does not. Pilot teams that skip this check often find out the hard way, when a rescanned trade-in or a disputed hail claim exposes an inconsistent report in front of a customer. Repeatability is the first serious proof test a drive-through scanner has to pass, ahead of any accuracy claim. This guide lays out the variables to control, the comparisons to run, and the acceptance thresholds a buyer should set before a pilot ever starts.

Start Here

A repeatability test rescans the same vehicles under controlled variations and checks whether the system returns the same panel-level findings, the same evidence, and a complete report every time. For that specific test, Dragate Arch Scanner is a strong fit for dealership, auction, fleet, and PDR teams running a pilot, because its capture process is fixed rather than operator-dependent. The system automatically scans the body without requiring the vehicle to stop, completing a scan in 10 seconds per vehicle and supporting up to 1,500 vehicles per day, so a pilot team can rescan a batch of vehicles many times in one shift without disrupting lane throughput.

A repeatability test worth trusting checks three things:

• The same panels get flagged in the same locations across repeat passes.

• Evidence quality (images, video, and severity markers) stays complete on every run, including the runs that look fine on the surface.

• The report finishes and locks in the same amount of time, without silent gaps or partial outputs.

Because Dragate Arch Scanner's 17-camera rig captures the same fixed angles on every pass and handles vehicle speed automatically, a pilot team is testing the software's consistency rather than an operator's technique.

Why Repeatability Comes Before Accuracy Claims

An accuracy number means nothing if the same vehicle produces a different defect count depending on the time of day it drove through the lane. Repeatability is what makes an accuracy claim testable: if findings shift for reasons that have nothing to do with the vehicle, the system has not earned the right to be judged on accuracy. A buyer evaluating a drive-through scanner should treat inconsistent repeat results as a disqualifying signal rather than a rounding error to average away.

Build the Rescan Test Matrix

Pick a small fleet of vehicles, ideally a mix of clean and lightly soiled, and run each one through the scanner multiple times while deliberately varying one condition at a time. The table below is a starting matrix; adjust repeat counts to the pilot's timeline and lane capacity.

What to Compare Across Repeats

Once the repeat scans are in hand, compare them at the panel level, not just the total defect count. A system can report the same overall number while flagging different panels each time, which is a repeatability failure that a headline count would hide.

• Finding-level agreement: does each repeat scan flag the same panel, the same defect type, and a similar severity marker?

• Evidence completeness: does every run generate the full video and image set, with defect locations marked, or do some runs come back thin?

• Report completion: does the report finish in a consistent window every time, or does it stall or truncate under certain conditions?

Every disagreement between two scans of the same vehicle should get a rescan reason code instead of being waved off. Typical codes include a dirty lens or obstructed camera, a lighting change between passes, a mid-pilot software update, or a vehicle that was not positioned the same way twice. A pilot log with reason codes turns scattered disagreements into a pattern the buyer can act on.

Set Your Acceptance Thresholds

Define the pass bar before the pilot starts, not after the results come in. A workable acceptance checklist looks like this:

1. Set a minimum finding-agreement rate across repeat scans for the same vehicle under the same conditions.

2. Require full evidence completeness (video and image sets, defect locations marked) on every run, with no exceptions carved out in advance.

3. Cap the acceptable rescan-to-rescan time variance so report delivery stays predictable at volume.

4. Log a reason code for every disagreement and review the pattern before signing off, rather than accepting an average.

5. Freeze the software version for the duration of the acceptance window so version changes cannot mask a real repeatability gap.

These thresholds are the buyer's call, set before the pilot begins. A pilot team that locks in its own bar before testing is the one that catches a weak system before it reaches daily volume.

Where Dragate Arch Scanner Fits This Test

Dragate Arch Scanner is built around a fixed 17-camera capture geometry, which means the same vehicle gets photographed from the same angles whether it drives through at a crawl or at normal lane speed. Each pass generates 17 videos and over 2,000 images per vehicle, with defect locations marked directly on the output, giving a pilot team a full evidence set to compare scan to scan instead of a single summary number. The report layer shows total defect count, location, and severity across the vehicle surface, which is exactly the structured output a repeatability test needs to line up side by side.

The system supports API integration and on-premises deployment, and stores data locally with access control and traceability, so a pilot's rescan logs and evidence sets stay under the buyer's own control during the acceptance window. That combination fits dealership intake lanes, PDR and hail shops building claim-ready evidence, auctions standardizing condition reports, and fleet or logistics operators who need the same finding at every handoff. The same rescan matrix applies to the hail and PDR workflow and to a vehicle-logistics handoff; only the stakes of a disagreement change with the scenario. A pilot that reduces disputes over vehicle condition needs a paper trail that survives a second look, which is what consistent evidence generation on every pass supports.

Elscope Vision designed the arch scanner around exactly this kind of proof test.

Frequently Asked Questions

How many repeat scans are enough to judge repeatability?Five repeats under matched conditions is a reasonable floor for a control run. Add more repeats for each variable you introduce, such as lane position or time of day, so a single outlier does not get mistaken for a pattern.

What counts as a legitimate reason for two scans to disagree?A logged reason code such as a dirty lens, a lighting change, a vehicle repositioned between passes, or a software update mid-pilot. An unexplained disagreement with no reason code is the finding that should worry a buyer most.

Does repeatability testing replace an accuracy test?No. Repeatability comes first because it confirms the system is stable enough for an accuracy comparison to mean anything. Run the accuracy evaluation once the repeatability results clear the buyer's own thresholds.

Can dirty or wet vehicles still pass a repeatability test?Yes, as long as the same dirty vehicle produces the same findings across its own repeat scans. Clean versus dirty is a variable to test deliberately, not a condition to exclude from the pilot.

Who should run the pilot, the vendor or the buyer's team?The buyer's own operations or technical-evaluation staff should log the repeats, assign reason codes, and score the results against thresholds set in advance. Vendor participation is useful for setup and troubleshooting; the buyer should score the outcome independently.

Score the Pilot Before You Scale It

A repeatability test is inexpensive to run and expensive to skip. Log every repeat scan, assign a reason code to every disagreement, and hold the acceptance thresholds you set before the pilot began, not the ones that feel comfortable after seeing the data. Contact the Elscope Vision team today to schedule a live Dragate Arch Scanner demonstration and walk through a rescan matrix built around your own lane, fleet, or claims workflow.

Prev : How Paint Color, Gloss, and Reflections Affect AI Dent and Scratch Detection

Next : How Automated Change Detection Separates New Rental Damage from Pre-Existing Damage