Context
Duty drawback lets importers recover duties paid on goods that are later exported or destroyed — valuable, but only if a business can actually prove the underlying transactions match up. For clients handling high volumes of trade activity, assembling that proof was a slow, manual, spreadsheet-driven exercise, and the delay directly limited how much duty recovery a client could realistically pursue in a given period.
Challenge
The process touched several disconnected data sources — import records, export documentation, inventory data — each with its own format and quality issues. Matching them required judgment calls that a spreadsheet-based process handled inconsistently from analyst to analyst. And because the output feeds a regulatory claim, any automation had to preserve a clear, defensible evidence trail — it wasn't enough to just go faster; it had to hold up under scrutiny.
My Role
I led the design and rollout of the automated framework, working across the trade compliance and technology teams to move the process from manual to systematic.
Approach
Rather than automating the existing manual process step-for-step, we redesigned it around a rules-based matching engine — using Python and SQL to standardize inputs from the different source systems, apply consistent matching logic, and flag exceptions that genuinely needed human review instead of routing everything through manual checks by default. That last part mattered: the goal was to point human attention only where it added value, not to remove it entirely.
Evidence management was built into the workflow itself rather than bolted on afterward — every matched transaction retained a traceable link back to its source records, so the audit trail regulators expect came out of the process automatically rather than requiring separate reconciliation work later.
Outcome
Processing turnaround dropped from roughly three to four weeks down to two to three days, with a more consistent and defensible output than the manual process had produced. That speed meant clients could pursue duty recovery on a much shorter cycle, directly increasing the value they could realistically capture.
Automating a regulated, evidence-heavy process isn't just a speed problem — the real design challenge is building the audit trail into the workflow itself, so faster doesn't come at the cost of defensibility.