The challenge
The legacy ERP spreads the same patterns across separate menus for every mode: air export, air import, sea export, sea import, customs import and customs export each have their own shipment lists, sub-screens and tracking. Many modules, from tariffs and the CRM to the accounting screens, showed no use, while the work that matters — customs jobs, filings and billing — sat inside that sprawl.
A replacement has to cover the whole regulatory flow without copying the sprawl. It also has to respect rules that are easy to get wrong: costs paid on the client's behalf are billed at cost without GST while service charges carry GST, customers deduct TDS before paying, and multi-company, multi-branch data has to be separated from the first day.
The solution
We proposed building one job engine rather than 400 screens. Customs import and export, air and sea import and export, and general jobs become one Job with a type, with mode-specific fields shown only where they apply. Milestone templates per job type define the ordered steps, dependencies and cut-off or free-day rules; milestone events then drive the status, the dashboard alerts, the daily status report, turnaround reports and customer notifications. Documents, containers, customs declarations, charges, invoices and receipts all hang off a job, and every record carries its company and branch.
Staff would work from one job page with eight tabs — overview, cargo, milestones, documents, customs, charges, billing and activity — and a side panel of seven sections. Every list and report uses one configurable grid, and every printed document is an HTML template rendered to PDF. Each outside system sits behind an adapter with a request log, retries and a status users can see: a GST Suvidha Provider for e-invoices and e-way bills, Tally XML voucher export, container and flight tracking, and email and WhatsApp, with ICEGATE filing and AI document extraction planned for phase 1.5.
The impact
What exists is the discovery record: a module-by-module map of the legacy system and what is actually used, six regulatory milestone flows, a list of domain rules, a 22-entity phase-1 data model, a system architecture, a recommended stack and a nine-step build order. We also wrote ten questions the client must answer before a firm quote, starting with which entity and branches do the real work today.
Nothing has been built. The recommendations — a modular monolith on PostgreSQL with company and branch on every row, a background queue for filings and exports, and storage in an India region — are recommendations, not decisions.
