11
Languages on launch · 6 more in pilot
DepotIQ replaces the clipboard with a phone. Walk a container, and the AI identifies every dent, hole, and corrosion patch — coded to IICL repair specs and quoted against your depot's rate card. Your assessor confirms. Your back office books revenue. The carrier gets a signed report before the truck leaves the gate.
Every box that comes off a truck gets walked, photographed, marked up on a clipboard, transcribed into a spreadsheet, priced manually against the carrier's rate card, then re-typed into Containerchain. It's been done this way for thirty years.
The IICL manual is already the world's most prescriptive damage taxonomy — every defect type, location code, severity threshold, and acceptable repair method is codified. We turned it into a model.
The assessor opens DepotIQ Field, points the camera at the container plate. ISO 6346 is parsed in 200ms — type, carrier, year, and tare are pre-filled before they reach the box.
Six guided angles — left, front, right, doors, roof, floor. The on-device model ensures coverage in real time; if you miss a panel, it tells you before you put the phone down.
Cloud inference identifies every defect against the IICL 6.1 taxonomy — dents, holes, corrosion (surface vs penetrating), bent cross members, paint failure, seal damage, floor delamination. Each gets a code, a location, a measured size, and a confidence score.
Every detected damage line is multiplied against your depot's rate card — different rates per container type, per carrier, per repair method. The quote is built before the assessor opens it.
The human is the second eye, not the first one. Assessors swipe-accept high-confidence items, drag-measure anything novel, and add IICL codes the model hasn't seen. Every override teaches the model.
The report is digitally signed, watermarked with the inspection video, and pushed straight to the carrier — Containerchain, MSC AU portal, Maersk DamageTrack, or your own EDI feed. No re-keying. No spreadsheets. No disputes.
Inspection happens on the bay. Allocation and sign-off happen in the office. Repair work happens at the contractor. DepotIQ is the connective tissue across all three — one workflow, three audiences, one locked report at the end.
High-contrast, glove-friendly PWA. Works in direct sun, queues offline, syncs over depot WiFi.
The connective layer between the assessor's quote and the contractor's spanner. Internal teams and external panels live in one directory.
Live queue, rate-card management, six-lane repair kanban, assessor + repairer scorecards, carrier integrations, and the analytics the back office needs.
62 damage classifications. 14 location zones per box type. Every code mapped to a parent damage family, an acceptable repair method, and a labour-time SOR. The model doesn't guess what damage is — it speaks the carriers' language out of the box.
| Code | Description | Severity | Repair |
|---|---|---|---|
| D-04-CB | Dent — corrugation bend | Minor | Hammer + paint |
| D-11-SR | Dent — side rail | Major | Cold straighten |
| D-15-RF | Dent — roof bow | Major | Reform + paint |
| C-22-PEN | Corrosion — penetrating | Critical | Cut + insert |
| H-31-PNL | Hole — panel puncture | Critical | Cut + insert |
| FB-14 | Cracked floorboard | Major | Replace 1 board |
| P-02-EX | Paint failure — exterior | Minor | Spot paint |
| S-08-DR | Damaged door seal | Minor | Replace gasket |
Per-depot, per-year. Conservative assumptions. Built from time savings on assessment, recovery on missed damage, and the disputed-claim rate going from 14% to under 3%.
Time saved is freed-up assessor capacity, redirected to higher-margin work — survey, M&R supervision, or just more throughput on busy days. Recovery on missed damage is upside that has been on the table for years; nobody's been able to capture it.
Based on 80 inspections/day · A$420 average ticket · 250 op days
Locked reports route automatically to your carriers and your DMS. We don't replace Containerchain. We feed it.
The IICL manual already tells us what every damage is, where it sits, and how to fix it. The only reason it's still a clipboard job is that nobody built the workflow.
DepotIQ ships in 11 languages on day one. The damage codes stay IICL. The interface, voice prompts, and signed reports speak whatever the gatehouse speaks.
Voice prompts during walk-through, in-app guidance, error states, and the customer-facing damage report PDF all localise. Damage codes (D-04-CB, C-22-PEN…) and ISO container numbers stay universal — that's what carriers settle on.
The public site shows Option A — flat per-assessor seat — because it's easier to sell to a depot ops manager and easier to forecast. Option B captures the volume upside but adds a per-inspection line item. Both are modelled below. The acquiring owner can pick one, hybrid, or replace entirely.
| Dimension | Option A · Flat seat | Option B · Seat + usage |
|---|---|---|
| Revenue per mid-size depot / year | ~A$15K | ~A$59K |
| Revenue forecast difficulty | Low (deterministic) | Medium (volume-dependent) |
| Procurement / approval friction | Low — one line item | Medium — two line items |
| Upside on high-volume depots | Capped | Linear |
| Buyer pushback risk | "Are we paying for unused seats?" | "Are we being penalised for being busy?" |
| Best for | Single-yard, 5–25 assessor depots | High-throughput, carrier-aligned ops |
This is a working prototype. The field app (PWA) and the depot portal are real, clickable software. The backend API (Node.js + Express + PostgreSQL, Docker-deployable) is built. The AI workflow runs against a real Claude API integration. What's not in the box yet: a custom-trained YOLO-NAS detection model — that requires labelled depot footage from pilot partners, which is Phase 0 of the rollout. Until then, detection accuracy and the per-line confidence scores you see in the demo are simulated.
The detection model is not yet trained — that's Phase 0 of any pilot. The architecture is YOLO-NAS, fine-tuned on the IICL damage taxonomy. The target is >90% line-item detection rate by pilot month 3, against an academic baseline of 91.2% mAP from comparable YOLO-NAS container-damage research. Final accuracy is a function of training-data volume and depot footage diversity — and gets re-measured monthly in the portal once live. We don't quote a single headline number for a model that hasn't been trained against your boxes yet.
Per-assessor seat, billed annually. A seat is one named human assessor — managers and read-only users in the back-office portal are free. Add or remove seats month-to-month within your annual commitment; we true-up at renewal. The depot owns the data; if you cancel, we hand over a complete export.
No. The whole point is that the model handles the 90% of damage that's standardised — corrugation dents, paint failure, surface rust — so your assessors have time to find the things that actually matter: structural cracks, hidden floor damage, suspect repairs. Override rate (how often the assessor disagrees with the model) is the metric we both watch. Lower means more trust; we don't want it to be zero.
IICL is the base. We've also mapped Cargo-IMC, ANZBC and the major carrier-specific damage codes (MSC, Maersk, Hapag-Lloyd, ONE) onto the same underlying detection layer. You configure the output schema per-carrier in the portal.
Yes. The field app is offline-first and pairs to gate-in events from Containerchain CDMS, 1-Stop, or whatever your TOS pushes out. Reports flow back through the same channel. Most depots can run their first inspection inside two hours of receiving the iPad.
It's a Progressive Web App — installable today via "Add to Home Screen" on iOS and Android. No App Store wait, no MDM packaging required, and updates ship instantly. A native React Native build is on the roadmap if your fleet management needs it, but the PWA approach is what's running in the prototype and is more than enough for pilot.
Inspection video and photos are stored in AWS Sydney by default; ISO 27001 + SOC 2 Type II are roadmap items (formal audits scheduled post-pilot). You can route long-term archive to your own S3 / Azure Blob. The depot owns the footage; we'd hold a non-exclusive licence to use de-identified frames for model training under a pilot agreement, and you can opt out of training contribution at any time.
Two-week target for a pilot depot, end-to-end. Week 1: rate-card import, carrier integrations, assessor accounts, repairer roster. Week 2: shadow-mode running alongside your existing flow while the CV model trains on your footage. Cutover into live inspection at day 14. The CV detection model continues to improve through pilot months 1–3.
30 minutes with our team. We'll walk you through the field app, the portal, and exactly what your first month looks like.