Duty Drawback Software Problems: The 7 That Shrink Claims and Fail Reviews

Why does the scope get set as a filing tool instead of an evidence system?

Because filing is the visible part. The finance sponsor sees claims going out and refunds coming back, so the brief describes claim preparation and submission, and the matching and evidence work sits underneath as an assumed input. It is the same mistake as scoping an accounting system around printing invoices.

Drawback punishes this framing harder than most categories, because a claim is not a number, it is a number plus a file. The entry summary lines, the export evidence, the production records, the destruction certification where relevant, and the trail showing how the match was reached all have to hold together for years after the money has been received and spent. A tool that produces claims elegantly and stores evidence loosely is a liability that pays you first and bills you later.

Set the scope the other way round. The first release should ingest import lines at full granularity, capture export and destruction evidence from wherever it actually lives, and assemble a complete claim package for one claim type, end to end. Filing can continue through your existing channel while that proves out, and many companies keep it there permanently. If a proposal leads with claim submission and treats data assembly as a mapping exercise, ask how the reviewer in three years reconstructs the match. The answer tells you whether they have built a claim that survived a review or only one that got sent.

What goes wrong with historical import and export data?

Claims reach back years, and the older the data the worse it is. Import lines sit with one or more brokers in formats that changed at least once. Export evidence is spread across a shipping system, commercial invoices, transport documents and a forwarder's records. Manufacturing consumption sits in an enterprise resource planning (ERP) system that deliberately stopped tracking material origin at receipt, and may have been through a cycle count that overwrote the detail you now need.

Two failures repeat. The first is a business unit that reclassified a product family two years ago without restating history, so classification based matching produces different answers either side of an invisible line. The second is exports shipped through a third party logistics provider under a different consignor name, which simply do not appear when someone pulls an export file by shipper. That second one is where a large share of missed eligibility usually hides, and nobody finds it by looking at the file they already have.

Handle it with a reconciliation pass rather than an import. Compare export volumes in your own systems against what the logistics providers actually shipped, resolve identifier mismatches deliberately, and record for every period how complete you believe the export evidence to be. A claim built on a file you have not tested for completeness is a claim scoped down to whatever happened to be easy, which is exactly the leakage the project was meant to close.

Why do broker, enterprise system and forwarder feeds break after launch?

Because none of them exist to serve drawback. A broker changes their extract when they change platform, and nobody tells the drawback team because the drawback team is not their customer relationship. An enterprise system upgrade renames a field that carried your part mapping. A forwarder adds a new lane through a different entity. Each of these is routine to the party making the change and quietly fatal to a claim.

The failures are also slow. A broken customs feed announces itself within a day. A broker extract that silently stops including one entry type produces claims that look normal and are incomplete, and the gap surfaces when someone notices a refund is smaller than the previous quarter.

Build the defences that suit slow failures. Every feed carries a source and a received timestamp, and the system reports volumes per period per source so a drop is visible against history rather than only against expectation. Line counts, duty totals and distinct classification counts per broker per month make an unannounced format change obvious in days. Where a source cannot be automated, track the request as an open item with an age, so a plant that has not sent production records for six weeks is a visible item rather than a forgotten one. And keep the mapping between your part numbering and each source's identifiers as maintained data, since that mapping is what breaks first and what nobody documents.

What happens when the allocation convention and retention are not covered?

Fungible inventory is the hardest honest problem in drawback and the one most often deferred. Your enterprise system knows what a work order consumed. It does not know which import entry line supplied that material, because materials are interchangeable in inventory and the system stopped caring about origin at receipt. Recreating the link means choosing an allocation convention over receipts and consumptions and applying it consistently.

Where that choice is not made explicitly, it gets made implicitly by whoever wrote the matching code, and it will not survive the question of why this receipt was matched to that export. Manufacturing claims add a second layer: the proportion of imported input embodied in an exported unit depends on your bill of materials, your actual rather than theoretical yield, and how you treat waste. A percentage applied to a total is not an answer.

Cover it by documenting the convention in the system rather than in a specification nobody reads, versioning it, and recording on every claim which version was applied. Then pair it with retention. A customs authority can review claims after payment, so the claim package, calculation, supporting lines, documents and rules version, must be stored immutably and remain readable independent of the software that produced it. Build that for the reviewer you will meet in three years rather than the analyst filing this week, because both your data and the rules will have changed by then.

Should you build custom or configure what you already own?

Configure if you already run a global trade platform and your data is clean. ONESOURCE Global Trade and Descartes are strong products that handle drawback within a broader trade suite, and if your flows are conventional and your imports and exports arrive mapped, switching on the module is the correct answer and a build would be waste.

Outsource if your recoverable duty is modest and your flows are simple. Specialist drawback firms work on contingency arrangements, they are genuinely good at the work, and paying a share of a recovery you were not otherwise going to make puts no capital at risk. Plenty of importers should stop here.

The limit of both is upstream. A packaged platform will accept an import file in its format. It will not tell you that your export file misses shipments made through a logistics provider under a different consignor name, or that a product family was reclassified without restatement. That is your data, your part numbering and your yields, and it is the project. Build when the constraint is evidence rather than filing: when claims are consistently scoped down to what is easy to prove, when manufacturing substitution is left untouched because nobody can face reconstructing yields, or when you are a specialist filer whose margin depends on absorbing messy client data faster than competitors can.

How do hidden costs get into the quote?

Manufacturing drawback is the single largest underestimate. It is materially harder than unused merchandise because it requires yield modelling per product family, waste treatment and a defensible per unit calculation, and it is frequently priced as a variant of the simpler claim type. If manufacturing claims are in scope, insist they are quoted separately and per product family rather than as one line.

Source count is the second. Two brokers with different formats is two ingestion profiles, and a quote covering "import data ingestion" is pricing whichever extract the developer was shown. Historical data quality is the third, and it is staff time as much as engineering time, because deciding how to treat a reclassification or a missing plant record is a trade and finance judgement rather than a technical one.

The fourth is plant record availability, which nobody checks before quoting and everybody should. Find out whether production records were retained at the granularity a manufacturing claim needs before the project is priced, not after. And if you are a specialist filer, multi client separation is architecture rather than a setting: data isolation, per client conventions and per client reporting all have to be designed in from the first release, because retrofitting isolation is expensive and rarely fully convincing to a client asking about it.

What separates a build that works from one that fails here?

The builds that work prove one chain completely before widening. One product family, one claim type, import line to filed claim, with the package assembled and reviewed by someone who has defended a claim. The recovery from that single family usually funds the rest, which turns a capital request into a straightforward internal conversation and gives you a working definition of done that a specification cannot.

They also ship the opportunity analysis early. Given import and export history, showing what appears claimable but is currently unevidenced, ranked by value and by how close the window is to closing, converts an abstract systems request into a specific number. Most organisations have never seen that figure and it is usually larger than the finance team expects.

The failures share a shape: a broad platform built to handle every claim type and every source at once, with the allocation convention left as a configuration decision for later. Later never comes with the same attention, and the first review exposes it. Two guards are worth insisting on. Refuse to let the engine produce a claim it cannot evidence, so an unsupportable claim is impossible rather than merely discouraged. And settle ownership before kickoff, the repository, the cloud accounts and the right to hire anyone else, because the system holds the evidence behind claims a customs authority may examine long after the money has been spent, and that evidence has to remain explainable independent of any supplier.