What 86,616 interview questions reveal about what companies actually ask
We analysed every question in our bank — 86,616 of them, across 29 subjects and 510 distinct topics. The distribution is not what most preparation advice assumes.
Last updated · Dataset: 86,616 questions · Free to cite with attribution
Finding 1 — Testing disciplines make up 30.6% of all questions
Automation, API, manual and performance testing together account for 26,538 questions — 30.6% of the bank. Preparation material overwhelmingly focuses on algorithms and language trivia, but in this dataset the single largest cluster of questions is about how you verify software, not how you write it.
Finding 2 — The bank is 45.6% Medium, not front-loaded with easy questions
Difficulty splits 25% Easy / 45.6% Medium / 29.4% Hard. Nearly three quarters of questions sit at Medium or above. Candidates who practise only introductory material are preparing for roughly a quarter of what they will be asked.
Finding 3 — Difficulty is not evenly spread across subjects
Ranked by the share of questions rated Hard, among subjects with enough volume to be stable:
| Subject | Questions | Hard |
|---|---|---|
| C++ | 2,617 | 36.8% |
| Java | 2,610 | 33.4% |
| Rust | 2,623 | 33.2% |
| Objective-C | 2,603 | 32.4% |
| Haskell | 2,593 | 32.3% |
| Scala | 2,559 | 32.3% |
| SQL | 2,701 | 32.1% |
| Ruby | 2,602 | 32.1% |
| TypeScript | 2,595 | 31.9% |
| C | 2,618 | 31.5% |
Finding 4 — Tooling dominates the most-asked topics
Across 510 topics, the highest-frequency ones are overwhelmingly specific tools rather than abstract concepts. Naming a tool on your CV is an invitation to be examined on it.
Methodology
- • Dataset. All 86,616 questions in the Interview Prep Academy bank as of 2026-08-14. No sampling — the analysis covers the full corpus.
- • Difficulty. Each question carries an Easy / Medium / Hard label assigned at authoring time; we report the labels as stored rather than re-scoring them.
- • Subjects and topics. Counted from the stored subject and topic fields. A question belongs to exactly one subject and at most one topic.
- • Stability threshold. Per-subject difficulty is reported only for subjects with 500+ questions, so small subjects cannot produce misleading percentages.
- • Reproducibility. Every figure is generated by a script against the live database; none is hand-entered.
Limitations
This is a corpus analysis, not a survey of hiring outcomes. It describes what is asked in this dataset — not what every company asks, and not what predicts getting hired.
- • company. Company attribution is present on only a minority of rows, so no company-level breakdown is published.
- • seniority. Seniority is unevenly populated and dominated by a single bucket, so seniority comparisons are not published.
- • Selection. The bank reflects the subjects we cover. Its heavy testing representation is a property of this corpus and should not be read as an industry-wide hiring ratio.
- • Labelling. Difficulty labels are human-assigned and carry the subjectivity that implies.
Cite this analysis
Free to cite and quote with attribution and a link. Suggested format:
Interview Prep Academy (2026). "What 86,616 interview questions reveal about what companies actually ask." Dataset: 86,616 questions, 29 subjects. https://interview-prep.academy/research