The Part 107 Training Tool Standard: fifteen questions to ask before you buy — or assign one to a class

The market for Part 107 preparation is crowded, and almost every product describes itself the same way: hundreds of questions, realistic practice tests, pass guaranteed. Those descriptions cannot be compared, because they are not claims about anything you can check.

This is a standard instead. Fifteen questions to put to any Part 107 training tool — ours included — grouped by what they actually test for. It names no vendor and ranks nothing. At the end we score ourselves against it in full, including the places we come off worst, because a standard that its author will not be measured by is a brochure.

If you are a student, ask these before you pay. If you are an instructor, ask them before you assign anything to a class, because your students will hold you to the answers.

Currency — is the material built for the exam that exists today?

1. Which blueprint is the practice weighted to, and where did the vendor get it?

The FAA retired the printed testing blueprint in FAA-S-ACS-10B on 29 September 2025 and moved the weighting to PSI’s Applicant Information Bulletin. Regulations went from a band of roughly fifteen to twenty-five percent to nearly half the exam. FAA-S-ACS-10B is still posted and still reads April 2021. A tool weighted to the old table allocates a student’s study time to the wrong proportions, and it will not say so, because it does not know.

The answer you want names the bulletin, states the percentages, and says how they were obtained.

2. What happened in the tool when the rules changed?

Night operations stopped requiring a waiver on 21 April 2021 (14 CFR 107.29). Operations over people became four categories under Subpart D the same day. Remote ID became mandatory on 16 September 2023 (14 CFR Part 89). Ask for one example of a question the vendor changed because of a rule change, and the date they changed it. A tool that cannot produce one has not been maintained through the period when maintenance mattered most.

Provenance — can an answer be checked, or only trusted?

3. Does every question cite the FAA source its answer comes from?

Not the tool’s own explanation — the regulation, advisory circular, handbook chapter or chart legend the answer rests on. A citation turns an answer key into something a student or an instructor can verify against the primary text. Without it, a dispute between a student and the tool has no referee.

4. What does the vendor do when a source shows they were wrong?

Every bank has errors. The question is whether the vendor finds them, fixes them, and dates the fix. Ask for the most recent correction and what it was checked against. A vendor who has never found an error in their own bank has never looked for one.

Fidelity — does practice resemble the sitting?

5. Is the timed simulation the real format?

The Unmanned Aircraft General test is 60 scored questions plus 5 unscored validation items, 120 minutes, 70 percent to pass, three choices per question. A practice test with four choices, or forty questions, or no clock, is measuring something other than readiness for that room.

6. How many full simulations can a student take before questions repeat?

Divide the bank by the per-sim draw in each content area. With Regulations at 48 percent of a 60-question sim, that area draws 29 questions each time, so a bank of 100 Regulations items supports three sims before a student has seen every one. Ask for the number. “Unlimited practice tests” describes how many times a button can be pressed.

Integrity — can the bank be gamed without knowing the material?

7. Does the vendor measure their own bank for shape tells?

The three that matter: whether the longest option is disproportionately the correct one, whether the correct answer clusters in one position, and whether words like always and never appear only in wrong answers. A student who notices any of these will score well on the bank and learn nothing. The only honest answer is a measurement, and a guard that stops the pattern creeping back as questions are added.

8. Does the explanation say why each wrong option is wrong?

An explanation that restates the correct answer produces a score. One that names what each distractor got wrong produces understanding. On every question a student gets right, they should be able to say why the other two were wrong; if the tool cannot help them do that, the question is not finished, whatever the score says.

Measurement — does the tool know when a student is ready?

9. Is readiness reported per topic, with the sample size beside it?

An overall percentage hides the topic that will fail the exam. Per-topic scores with the number of answers behind each one let a student see that Airspace is at 61 percent on nineteen answers, which is actionable, rather than that they are “at 78 percent,” which is not. Ask what the tool shows when the sample is too small to judge; the honest answer is nothing, plainly labelled, rather than a number it cannot support.

10. Is there a gate, and what is it?

“Ready to book” should be a stated threshold — a per-topic floor plus a number of passed timed simulations — not a feeling. Ask what the threshold is. Then ask whether the vendor claims the gate predicts passing; the honest answer is that it does not, because no one has the correlation data, and a vendor who says otherwise is selling you a number.

Also ask: when a student misses a question, what does the tool do with it?

Not one of the fifteen, because it is a method question rather than a measurement one, but it decides how the student’s hours are spent. A bank that shows every question equally often spends time evenly across what the student knows and what they do not. Ask what the review schedule is, in days. A tool that says “adaptive” and cannot answer in days is a bank with a shuffle button. We describe ours, box by box, in How study works here.

Charts — will the skill transfer to a figure the student has never seen?

11. Are chart questions drawn on figures from the printed testing supplement, or on excerpts the student has never seen?

From 26 October 2026 the FAA is embedding current sectional and IFR Low imagery in test questions rather than drawing only from the printed supplement. A student whose chart skill is recognising a specific supplement figure has a skill that stops transferring on that date. Ask how many chart questions the tool has, where the figures come from, and whether it teaches a scanning procedure for an unfamiliar excerpt rather than the answer to a familiar one.

Instruction — what does an instructor actually get?

12. What leaves the student’s browser, and what does the instructor see?

For a class, the useful view is per student and per topic, with sample sizes, and a statement of which topic to reteach first, computed from the class rather than from one student’s bad week. Ask exactly which data is sent to the server and whether any of it reveals which questions a student saw. Then ask how seats work: whether a licence is reassignable and on what cycle, or consumed once, and when a consumed seat’s clock starts.

Verification — can the source be read, and is the check repeated?

13. Can the student read the FAA source from the explanation, or only a citation string?

A citation string tells you where to look. The text is what you check against. Ask to see the source behind one explanation. If the answer is a citation string, ask where the text is.

14. When the FAA amends a rule, how does the vendor know which questions are affected?

Rules change on dates. Ask whether each question records the date of the rule text it was written against, and what the vendor does when an amendment comes after that date.

15. Is the checking repeated when the bank changes, and what stops an unsupported change from shipping?

A bank checked once drifts as it is edited. Ask whether every change is checked against the source again, and what happens to a change that fails the check.

---

How we measure against it

We built Checkride107, so this is not a neutral scoring. It is an honest one, and the shortfalls are listed with the rest.

#	Question	Where we stand
1	Blueprint	Weighted 48 / 25 / 20 / 5 / 2 across Regulations, Operations, Airspace, Weather, Loading, transcribed by hand from PSI’s bulletin, page 4. The currency line on the site names the date the FAA retired the old table.
2	Rule changes	The rule tracker is a dated log. The bank carries fourteen items on 107.29 night operations as amended and seventeen on the four Subpart D categories.
3	Citations	Every one of 464 questions names the section its answer comes from. The build fails if a citation is missing. In October 2026 every question was reviewed against the text of the FAA document it cites, each verdict quoting the passage it rests on.
4	Corrections	Most recent: 4 October 2026, change sets 2026.10.5 to 2026.10.7 from the October source review. 42 existing questions were corrected (28 of them in the citation string only), 6 with the wording of the keyed answer changed, and one retired. The detail is in The source decides, on this blog. Before that: 13 September 2026, a Class E authorisation item corrected against FAA Order JO 7210.3 ¶19-6-3 — our answer was the over-cautious one and it was still wrong.
5	Format	60 scored plus 5 validation, 120 minutes, 70 percent, three choices.
6	Depth	5 full simulations with no repeated question, derived from the bank at build time rather than typed.
7	Shape tells	Length, position and absolutes are each measured and each is a build guard: the build fails if any drifts. We do not publish the percentages, because a published figure is a commitment to republish it monthly and we have not made that commitment yet.
8	Explanations	Shortfall. Every explanation cites its source, and the newest items name why each wrong option is wrong. Many older ones only explain the right answer. That rewrite is an open queue and it is the largest piece of content work we have.
9	Per-topic readiness	Per topic, sample size beside every score, greyed below five answers rather than shown.
10	The gate	85 percent on every topic plus two timed sims at 80 percent or better. We make no claim that it predicts passing; there is no correlation data and we say so on the page.
11	Charts	Both. Chart questions show the actual pages of the FAA testing supplement, FAA-CT-8080-2H, with figure and page numbers and a zoom, and every legend and figure is open in the figure library. A smaller set of questions uses excerpts drawn inside the product that reproduce no supplement plate. A scanning procedure and the four multi-step conversions are on the decoders page. In October, fifteen chart-reading questions were added, thirteen of them on the supplement's own pages, for readings that transfer to a chart the student has not seen.
—	Study loop	Five Leitner boxes. A missed card returns the same day; a card answered right five separate times is not seen again for two weeks. The intervals are rendered on the Study page from the engine, not typed.
12	Instructor view	Per student, per topic, sample sizes, reteach-first from the class. Only the summary leaves the browser — score, sample size, gate status; no answers, no question ids. Seats are either a monthly licence, reassignable, or prepaid, consumed once with the clock starting when the student activates.
13	Read the source	Partly. For the 274 questions that cite a regulation, Read the source shows the section text with the date it is current to. For a testing-supplement figure, it opens the FAA figure page. For other FAA documents it shows the title and edition, not yet the page text.
14	Amendments	Partly. Most questions record the date of the rule text they cite; some do not yet. The source audit lists every cited CFR section amended after a question's date, so the questions to read again are named rather than guessed.
15	Repeated check	The source audit runs on every change to the bank and fails when a question's content no longer matches a review quoted from its source. It reports to us; it does not yet stop a deploy on its own.

The second shortfall is verification by a person. In October 2026 every question was reviewed against the text of its cited source, and the audit repeats that on every change. A person has read 364 of the 464 against the source; the rest are queued, and no verification date is shown for any item yet. A tool that shows you a “verified” date should be asked what that date means.

Take the fifteen questions into any evaluation. If a tool answers all of them well, use it, whoever makes it.