Industry field guide
Footwear and leather repair condition data for AI
On a shoe or bag, surface marks, folds and worn edges can look similar in one photo.
- 2022 NAICS
- 811430
- Practice pack
- Example and answer key (JSON)
- Updated
What this guide tests
Which visible change is supported, and does the remaining fold or inspection motion establish failure or complete restoration?
See the specimen and fields · Work through the fictional records
Task and source context
On a shoe or bag, surface marks, folds and worn edges can look similar in one photo. Troveo’s public physical AI guide describes close-up manipulation footage and repeated demonstrations. A repair shop can pair an original clip with independent inspection labels and test what the camera actually shows. Troveo robotics data guide · FTC repair report
Use a shop-owned, unbranded practice item. Mark the affected edge on a reference image, film a safe hand inspection, then take a same-angle after-photo. Include one example where a fold remains after the repair. Ask which changed region is supported and where evidence is limited to appearance.
Exclude customer items, markings that identify the wearer and proprietary design material. Obtain separate permission for the camera operator and performer.
Industry research
The U.S. Census Bureau’s industry definitions establish which activities these codes cover. The workflow sources in this guide describe the task or record type. Fiund’s pilot below uses fictional material to test a specific decision within that scope.
811430 · Footwear and Leather Goods Repair
This industry comprises establishments primarily engaged in repairing footwear and/or repairing other leather or leather-like goods without retailing new footwear and leather or leather-like goods, such as handbags and briefcases.
Build a pilot sample
Set up this pilot
Shoot a marked practice leather edge before and after cosmetic work. Include an occluded fold and no strength test. Grade visible change without generalizing to durability.
Prepare the test packets
- Start with the worked case. Use the fictional record types below as a guide to the source files you need. Give each packet a stable ID and save its source versions with it.
- Change one fact. Make one packet with enough evidence to answer the question and one where the evidence points to a different answer. Keep the other fields fixed so a reviewer can tell what changed.
- Remove the decisive evidence. Give the reviewer a third packet with a plausible near-match in its place. Have them write what can still be established and what evidence would resolve the rest.
Fields the reviewer needs
- Item, edge and feature identity
- Separates the mark from the fold.
- Matched view and lighting
- Supports the appearance comparison.
- Occluded region and test availability
- Limits hidden-condition and performance claims.
Use What repair and service businesses can record for AI to turn the specimen into a brief with a defined task, source and review owner.
Fictional exercise
A remaining fold does not erase a documented edge appearance change
An owned unbranded practice item has an original reference image and approved hand-inspection clip. The task describes appearance without product-performance or customer-responsibility claims.
| Record ID | Source | What it records |
|---|---|---|
| LR1 | Reference image | Item L, edge region E marked; pale surface mark M and separate fold F visible. |
| LR2 | Matched after-image | Same item/region/light; M no longer visible, F still visible. Reviewer confirms view pairing. |
| LR3 | Hand-inspection clip | Hand points to E but covers part of the edge. No material-strength or durability test supplied. |
Question for the model. Which visible change is supported, and does the remaining fold or inspection motion establish failure or complete restoration?
Show the reference answer
LR1 and LR2 support that M's visible appearance changed in the matched edge region while fold F remained. LR3 documents an inspection gesture but not the covered region or material performance. The persistent fold does not by itself negate the separate M finding, and the images do not establish complete restoration, strength or durability. Report the two visible features independently with their comparison limits.
How to check the result
Ask for the answer, the record IDs that support it and any fact the evidence leaves unresolved. Have a reviewer compare the response against the reference before deciding whether the sample is ready to expand.
Errors to flag in this case
- Combining a fold and a surface mark into one undifferentiated defect label.
- Treating a hand inspection or changed appearance as a durability test.
Change the inputs and run it again
Change the after-view angle so F is hidden but M remains visible. Require an unobserved-fold finding rather than declaring it removed.
Score the decision and the evidence separately
Save the reviewer’s expected answer, supporting record IDs and unresolved fields in a separate answer key. Score the model on the decision, its citations and its handling of the deliberate gap. A correct answer citing a different case counts as an evidence error.
Before expanding collection, have a second reviewer run the same checks on the held-out packet. Keep related versions, views and exports of the same case together so they stay on one side of the training and evaluation split.
Check permission and scope
Use shop-owned cleared props and original footage. Exclude customer items, wearer-identifying marks and proprietary designs.
Remove customer names, serials that identify owners and real safety-critical service instructions.
Describe an authorized archive through Fiund’s existing-data assessment, or describe a new capture capability through the collection brief. The contributor agreement covers the terms for participation.
Sources and further reading
Resources for this workflow
Includes worksheet
What repair and service businesses can record for AI
Prepare an authorized repair example with an observed state and a checkable outcome.
Includes worksheet
First-person video capture checklist for an AI data pilot
Choose safe props, camera positions and accepted intervals before filming.
Company and program reviews
Use these reviews to compare sourcing, capture or evaluation approaches. Confirm the applicable program, accepted sources and terms with the provider.
Provider cited in this guide
Troveo review: video licensing, business data and creator payouts
The guide cites material from this provider. The review explains the offering and which details require a specific brief or agreement.
Compare the approach
micro1 review for experts, trade footage and company data
Read the distinction between trade recording approval and other kinds of expert work.
Compare the approach
Dexset review: robotics capture, samples and acceptance
Compare what a task-capture export should preserve about visible actions and annotated outcomes.
Related industry examples
Industry guide · NAICS 811121
Collision shop panel condition and repair evidence for AI
An estimate can name a repair zone that is hard to locate in a tightly cropped photo.
Industry guide · NAICS 811411
Garden equipment accessory fitment and repair video for AI
A repair counter can show the visible difference between two similar accessory fittings without filming a full repair.
Industry guide · NAICS 811310
Industrial machinery repair verification and service release data for AI
A bench repair ticket may be complete while the equipment owner’s return-to-service approval is still pending.