Requirements First: What a Technology Consultant Uncovers Before US Customs Broker and Trade Compliance Leaders Commission Custom Brokerage Software

This article is part of our series on Custom Customs Brokerage Software Development for US Licensed Brokers: Building an ACE Filing, Classification and Duty Management Platform

Start With the Classifications You Already Filed

The first exercise in gathering customs software requirements takes a few days, needs no vendor, and is uncomfortable, which is why it is useful.

Take twenty classifications in use across your highest-volume clients. For each, establish who decided it, when, on what basis and under what authority.

Most brokerages find that for a meaningful proportion, somebody classified it years ago, nobody recorded why, and it has been copied ever since. That is the exposure, and it is not a software problem. It exists today, compounds with every entry, and is what an audit would examine first. No custom software development engagement fixes it; the platform only decides whether it keeps happening.

The second exercise follows: check whether any of those twenty carry additional duties or trade measures that did not apply when the decision was made.

Both are available to any brokerage willing to spend a week, and both produce action regardless of any build decision. What importers eventually see, through web application development, comes later.

The Two Decisions That Shape Everything

Before customs software requirements are written, two decisions should be settled, because every requirement depends on them.

The filing architecture: self-certification or a certified service provider. Certification brings months of process, specifications to implement per transaction type, and permanent maintenance. A provider brings a per-transaction fee and a dependence. For most brokerages the provider route is right, and it removes a large and uncertain component from the project.

The supervision model: how licensed review is structured, what triage separates routine entries from those warranting attention, who sets the escalation rules and what the record of review looks like.

That second decision belongs to the licensed broker rather than to a vendor or a developer, and it should be made before the workflow is designed rather than discovered inside it.

A consultant’s first contribution is forcing both decisions early, with the people qualified to make them, rather than letting a design imply answers to questions nobody asked. Both are settleable in a few weeks.

Requirements work is the decision layer of the platform described in Custom Customs Brokerage Software Development for US Licensed Brokers.

What Requirements Work Should Establish

The Classification Record Audit

The exercise above, across a representative sample, with gaps attributed: no recorded basis, outdated authority, inconsistent application, or decisions predating measures that now apply. This is the business case and the compliance finding in one.

The Agency Coverage Position

Which agency message sets the current client base requires, which the pipeline would add, and what maintaining each involves. This sizes a recurring cost and decides which clients the firm can serve, and it depends on knowing what each importer ships, which is why client records built as CRM system development matter before any build starts.

The Entry Process Observed

Watching entry writers work during a busy period: how documents arrive, what gets checked, where information is missing, and what licensed review actually consists of. Requirements written from observation differ from requirements written from description, and they are what makes UI/UX design services productive rather than decorative.

The Record Production Test

Pick three entries from different years and produce the complete record as a request would require it, then time it. This tests retention, assembly and whether production is a function or a project.

The Advance Filing Timeliness Review

Filing times against sailing across a period, with late filings attributed by cause, since causes are usually upstream and addressable without software.

The obligations these customs software requirements must account for are set out in CBP 19 CFR 111 Broker Licensing and Responsible Supervision, Reasonable Care and Five-Year Recordkeeping, ISF Timeliness Penalties, Country of Origin Rules and Forced Labor Enforcement.

What the Engagement Should Produce

A classification record audit with gaps attributed and the exposure characterized, flagged for counsel where it warrants it. An agency coverage assessment with maintenance costed as a recurring line. A filing architecture recommendation with certification and provider routes costed side by side.

A supervision model documented, authored by the licensed broker and reflected in the workflow design rather than implied by it. An observed entry process report. A record production test result with retention and assembly gaps identified. An advance filing timeliness review with causes attributed.

A platform test against the established brokerage products, with maintained filing specifications, agency sets and tariff content treated as the requirements they are.

A defined first release with exclusions written down if building proceeds, including whether importers need phone access at all, since custom mobile app development for status and approvals is a separate line rather than an assumption.

And a costed comparison of at least three paths, meaning adopt a platform, build a layer beside one, and full custom, with data subscriptions, agency set maintenance and interface maintenance shown as ongoing in each. The budget this protects is detailed in The Real Price of Custom Customs Brokerage Software.

Reading the Recommendation

Adopt an established platform when the test shows one covers the operation, which for most brokerages it does. The maintained filing specifications, agency message sets and tariff content are the argument, and that work never stops.

Build a layer beside a retained platform when the gap is the classification decision record or the client experience. That is a defensible narrow project, since the decision record is where most brokerages are genuinely weak and where the products are thinnest.

Build fully when scale, a distinctive service model, or a platform sold to clients justifies it. That last case is SaaS development and should be priced as a product from the start rather than discovered to be one later.

Reconfigure rather than replace when the assessment shows capability that was never set up, which happens and is the cheapest outcome.

One consideration whichever path: the classification record gaps are worth closing now with licensed broker time, since they describe an exposure that exists today and grows with every entry.

A consultant who reaches the platform recommendation and says so is demonstrating the independence that makes the rest usable.

What a Consultant Should Know About This Business

Four questions establish whether a consultant has worked with licensed brokers.

Ask how the system would handle classification. The answer should be candidates with their basis for a qualified person to decide, with the reasoning recorded. Anybody describing automated classification, a model that returns a code, or a lookup that resolves it has misunderstood what classification is. Where AI product and agent development belongs here is extraction for verification, candidate proposal with basis, matching to prior decisions and discrepancy surfacing, never classification, origin determination, valuation conclusions or transmission without licensed review.

Ask what happens to licensed review at volume. A good answer describes triage that directs attention and escalation rules set by the licensed broker. One that describes straight-through processing as the goal should end the conversation.

Ask about reasonable care. A consultant who has worked here raises discrepancy review unprompted and understands that client-supplied data is examined rather than passed through.

And ask how they would cost agency message sets. The right answer treats them as maintained content with a recurring line per agency, not as a one-time build.

Red Flags in the Conversation

Requirements work offered free and contingent on winning the build. A fixed price before the filing architecture is settled. Agency message maintenance absent from running costs. No test of the established platforms. No time observing entry writers. The supervision model left to the vendor to define.

And the ones that should end the conversation: automated classification presented as a feature, candidates sorted or filtered by duty rate, any capability framed as reducing duty through code selection, straight-through filing described as removing the need for licensed review, origin derived from shipping documents, and forced labor tooling framed as assessing detention risk rather than establishing facts.

The strongest positive signal is a partner who asks to sit with a licensed broker while they review a day’s entries.

Conclusion

Audit twenty classifications before deciding anything, since what that reveals is an exposure growing with every entry. Settle the filing architecture and the supervision model early, with the licensed broker authoring the second. And test the established platforms honestly, since maintained specifications, agency sets and tariff content are permanent work a brokerage otherwise carries alone.

Apply the same four questions to any consultant, the software companies in the United States who would like to build it included.

If you are weighing custom brokerage software, pulling twenty classifications and asking who decided each and on what basis is the week of work to do before any vendor conversation.