Industry
Track and trace that holds up when the ship is late
Visibility, customs filing and cost-to-serve systems for operators who move real freight.
Context
Technology in Logistics & Supply Chain
We build the systems that keep freight moving and paperwork correct: EDI and API integration with carriers and customers, customs filing, yard and depot operations, driver applications that work without signal, and the settlement logic that turns a completed move into an invoice. Most of the value is in the boring parts — reference data, partner-specific message quirks, and exception handling when a leg is subcontracted three times.
Estimated arrival times get most of the attention and deserve less of it. On lanes where variance comes from port congestion or customs holds you cannot observe, a better model buys you very little; buying berth and terminal data changes more than another gradient-boosting run. We will tell you which of your lanes are worth modelling and which need a data source instead of an algorithm.
- 11 weeks to 4 days
- partner EDI onboarding time at Corvus Logistics
- 1.9%
- customs filing rejection rate, down from 8.4%, at Pilbara Freight
- ±42 min
- median arrival prediction error on modelled short-sea lanes
Challenges
What actually keeps this sector awake
Not generic disruption — the specific constraints that shape every technology decision here.
An EDI estate with a dialect per partner
EDIFACT and X12 messages that each large customer has bent to their own shape, mapped in a tool one contractor understood. Onboarding a new partner takes eleven weeks, and a silent mapping failure surfaces as a missed delivery, not an alert.
Customs regimes with hard filing windows
Pre-arrival declarations under ICS2 and equivalent regimes must be filed before loading, with data owned by parties who supply it late. A rejected filing at 02:00 stops a container and nobody is watching a queue at that hour.
Subcontracted legs you cannot see
Your customer expects a status every hour; the actual driver works for a partner of a partner with no telematics integration. Coverage gaps get filled with phone calls and optimistic assumptions that later become claims.
Cost to serve unknown until month end
Fuel, demurrage, detention, waiting time and reweighs land weeks after the move, so pricing decisions are made on averages. Loss-making lanes stay in the book for two quarters because nobody can attribute cost to shipment.
Approach
How we address them
EDI normalised into a shared event stream
Partner messages translated into one internal event model with schema versioning, so a new customer format is a mapping change rather than a code change. Every translation failure raises an operational alert with the original payload attached.
Customs filing with validation before submission
Data completeness checked against the regime's rules at booking, chased from the responsible party while there is still time, and filed with automatic retry and escalation. Rejections route to a monitored queue with the specific failing field named.
Telemetry fusion with published error bars
Carrier feeds, AIS, terminal events and driver application pings combined into one position record, with confidence stated. Arrival predictions ship with an interval, and we report model accuracy per lane so nobody trusts a number that has not earned it.
Shipment-level cost and margin
Accruals raised at execution time for the charges that always arrive late, reconciled against supplier invoices automatically, so margin per lane and per customer is visible within days rather than after the close.
Services
Relevant practices
Engineering
Enterprise Integration
Contracts, idempotency and replay — the unglamorous parts that keep 40 systems in step.
Engineering
API Development
Contract-first interfaces with versioning, quotas and deprecation rules agreed up front.
AI & Data
Data Engineering
Pipelines with contracts, tests and an on-call rota, not a folder of scheduled scripts.
Engineering
Microservices
Service boundaries drawn around team ownership and failure domains, not fashion.
Engineering
Mobile App Development
Native and cross-platform apps built for field conditions, not conference demos.
AI & Data
Machine Learning
Forecasting, ranking and detection models that keep working after the data shifts.
Products
Platforms used in this sector
Industry
Atmora Fleet
Telematics, trip costing, maintenance and drivers for owned, leased and attached vehicles.
Operations
Inventory Management
Multi-warehouse stock, batch traceability and counting on handheld scanners.
Operations
Vendor Portal
Purchase orders, dispatch notes and invoices handled by suppliers themselves, matched before they reach accounts payable.
Data & AI
Atmora Lens
A governed semantic layer with dashboards attached, so every report agrees on the numbers.
Operations
Workflow Automation
Approvals, exceptions and hand-offs modelled as processes you can version, measure and change without a release.