What Actually Gets Tested on the Customs Broker Exam

Every question from 17 past customs broker exams, 2018 through April 2026, sorted by what it actually tests. 1,129 questions total.

Four topics are just over 59 percent of the exam.

The breakdown

What stands out

Entry / Entry Summary alone is 249 questions, more than one question in five. Add the next three and you have 676 of 1,129 questions, just over 59 percent of everything asked in eight years.

The ten smallest topics combined come to 202 questions. Classification alone is 202. So the entire bottom half of the syllabus is worth less than one of the big four.

That has a straightforward consequence for how you spend study time. Most candidates distribute effort roughly evenly across the syllabus because that is how the syllabus is presented. The exam does not distribute questions that way.

None of this says the small topics are safe to skip. You need 60 of 80 to pass, and a bad run through Drawback and Quota can be the difference. It does say that an hour spent on Entry / Entry Summary is worth several hours spent on Recordkeeping.

Which parts of 19 CFR actually get tested

784 of the 1,129 questions cite a part of Title 19 as their primary authority. Those questions fall across 49 different parts, and fifteen of them carry 614, just over three quarters.

The remaining 34 parts account for 170 questions between them.

Part 111 sitting third is worth noticing. 65 questions test the rules governing the broker's own license: who needs one, what responsible supervision and control means, and what records have to be kept and for how long.

Which HTS chapters get tested

169 questions cite the HTSUS as their primary authority. They spread across 52 chapters, and the largest is chapter 85 at 13 questions.

The remaining 37 chapters account for 65 questions.

The tariff does not concentrate the way the regulations do. Fifteen of the 49 CFR parts carry three quarters of the regulation questions, so working the high-frequency parts directly is a reasonable use of study time. Chapter 85 is the largest single chapter across eight years of exams and it comes to 13 questions, which is 1.2 percent of the total.

So there is no set of chapters you can memorize that covers the classification section. What gets tested is whether you can find the right heading under time pressure, in a book you have not opened to that page before.

Has the mix changed over time?

The categories CBP uses to organize the exam changed around May 2024. Comparing sittings across that line compares two different labelling schemes rather than two different exams, so anyone telling you entry questions have doubled since 2024 is reading a relabelling as a trend.

What the data does support is consistency. Across the 5 sittings from May 2024 onward, under the current scheme:

• Entry and Entry Summary: 32 to 37 percent
• Valuation: 18 to 20 percent
• Broker Compliance: 18 to 26 percent
• Classification: 12 to 20 percent

5 consecutive exams inside those bands is worth knowing when you are deciding how to spend the next three months.

How this was built

Every question comes from an exam CBP published, paired with the official answer key CBP published alongside it. The citation on each question is CBP's own, taken from the answer key, not an interpretation.

Each question is counted once, under its primary citation. Of the 1,129:

• 784 cite a part of 19 CFR
• 169 cite the HTSUS
• 176 cite something else: the ACE and RTME reference documents, CBP forms, the U.S. Code, or a worked calculation with no single controlling citation

Topic assignment is ours. Where a question could sit in two topics, for example a classification question that turns on a valuation issue, it is filed under what the answer key cites. That is a judgment call and reasonable people would file some of them differently.

Limitations, stated plainly

The bank holds 1,314 questions and 1,129 are counted here. The other 185 are excluded because they depended on an exhibit CBP never published, or because the law they tested has since changed enough that the published answer is no longer right.

"Other" is 47 questions, 4 percent, that did not fit cleanly into a named topic. It is a real bucket, not a rounding error, and it is listed rather than hidden.

This covers 17 exams. It describes what has been tested, which is the best available guide to what will be tested, and is not a prediction.

Practice against the real distribution

The simulator's weighted study mode is built on exactly this analysis, so the questions it puts in front of you follow the distribution above rather than an even spread. You can also sit any of the 17 exams as a timed mock.

Browse all 17 past exams →

Last verified against the question bank: September 16, 2026.