Online Assessment Time Pressure Index 2027: Questions Per Hour by Company

Online assessment time pressure is best measured as a pacing signal, not a difficulty score. Citadel reports two questions in 75 minutes, while J.P. Morgan, Meta, Morgan Stanley, Booking.com, and Rubrik sit near two questions per hour in the comparable records reviewed here.

That ranking has a strict limitation: these are candidate-reported formats, not universal company rules. A “question” may be a short algorithm problem, a multi-level project, a code-review task, or one scenario with several hidden requirements. Use this index to set a practice pace and to spot where a format deserves more buffer. Do not use it to claim that one employer’s assessment is objectively harder.

The linked problems are general preparation resources. They are not presented as a verified copy of any employer’s private assessment.

Quick answer: which reported formats have the highest question density?

The calculation is count ÷ minutes × 60. “Count” means the number of questions or tasks the candidate explicitly described; it does not convert hidden test cases, subparts, or progressive levels into extra questions.

Reported employer	Candidate-reported format	Time	Normalized density	How to read it
Upstart	5 coding questions	60 min	5.0 q/h	Fast question switching; report says one hour and five questions.
J.P. Morgan	2 coding questions	60 min	2.0 q/h	A conventional two-question timed screen in this report.
Meta	2 coding questions	60 min	2.0 q/h	Candidate report says two questions and 60 minutes.
Morgan Stanley	3 coding questions	90 min	2.0 q/h	Three medium-difficulty questions in the cited report.
Booking.com	3 programming questions	90 min	2.0 q/h	Includes an application-style hotel-review scoring problem.
Rubrik	3 coding questions	90 min	2.0 q/h	Candidate described the set as medium-plus and time constrained.
Citadel	2 coding questions	75 min	1.6 q/h	Fewer questions per hour does not imply lower difficulty.
Salesforce	2 tasks: coding plus project	90 min	1.3 tasks/h	Mixed format; a project should not be ranked like a short problem.
Uber	2 coding questions	90 min	1.3 q/h	Useful for pacing, but the report is a single candidate account.
DRW	3 coding questions	150 min	1.2 q/h	Longer window; report notes that hidden-case feedback may be limited.

The rows are a transparent sample, not a claim that these are the only employers with timed assessments.

What this index measures

The index measures how quickly a candidate must move between explicitly counted tasks within a stated timed window. A higher number usually means less time for reading, clarification, implementation, testing, and recovery after a wrong turn. It can help you decide whether to practice one long implementation or several short problems in one sitting.

It does not measure the number of hidden tests, the score cut-off, the employer’s selectivity, the amount of proctoring, or the probability of receiving an interview. It also does not estimate algorithmic difficulty. A single 90-minute project may require more design, refactoring, and compatibility work than five short questions.

Dataset and deduplication

The six-page fetch contained 538 records when retrieved on September 6, 2026; the API records themselves had updates through September 2, 2026. A public record is a candidate account or question entry, not a company policy document.

The same remembered OA can appear as several question entries. To avoid counting every extracted question as another assessment attempt, the analysis groups records by employer and normalized report text, then keeps a single format when the report contains an explicit count and duration. Reports with only “four levels,” “several sections,” or “about an hour” are kept as qualitative evidence but are not inserted into the ranked calculation unless the count and timer can be stated without inference.

The formula and its limits

Use this calculation when the unit is comparable:

questions per hour = explicitly counted questions ÷ timed minutes × 60

For example, three questions in 90 minutes equals 3 ÷ 90 × 60 = 2.0 questions per hour. Keep the original count and minutes beside the normalized value. A rounded number without its denominator can create false precision.

For a progressive project, retain the unit in the label: Airbnb’s report is shown as 2.7 levels per hour, not 2.7 questions per hour. For a mixed Salesforce report, the table says tasks per hour because one task is a project and the other is a coding problem. These rows are useful for planning but should not decide a strict employer ranking.

How to interpret the density bands

Above 3 questions per hour: switching cost dominates

Upstart’s five-in-one-hour report gives roughly 12 minutes per question before any reading or submission overhead. The right practice is not to rush every problem blindly. Train a fast first pass: identify the input/output contract, write one example, choose a basic data structure, and decide quickly whether to move on. Reserve a final pass for test cases and syntax errors.

A rapid screen can still contain long prompts. One candidate report describes six questions with individual eight-minute limits and lengthy statements. The raw questions-per-hour number would hide the fact that reading time, not typing speed, is the bottleneck. Practice with unfamiliar wording and a visible timer rather than memorizing a narrow set of patterns.

Around 2 questions per hour: balanced but still timed

J.P. Morgan, Meta, Morgan Stanley, Booking.com, and Rubrik land around two counted questions per hour in the selected reports. This is enough time to explain an approach and test a solution, but not enough to spend most of the session on an elegant optimization that has not been validated.

Use a two-pass plan: spend the first quarter of the window understanding both prompts and sketching solutions, then implement the more certain problem before returning to the harder one. If the assessment allows movement between questions, leaving a partial but compiling solution can be safer than polishing one answer while ignoring the second.

Below 2 questions per hour: depth may replace breadth

Citadel’s two questions in 75 minutes, DRW’s three in 150 minutes, and Uber’s two in 90 minutes look less dense. That does not make them easy. More time per task can indicate harder constraints, multiple follow-ups, C++ requirements, limited hidden-test feedback, or a need to explain complexity. The preparation target becomes correctness under depth: write tests, check boundary cases, and keep the code easy to revise.

Salesforce’s mixed coding-plus-project report is a reminder to budget by format. A code-review or project section may consume time through repository navigation, test interpretation, and debugging. Treating it as “one question” makes the index look artificially relaxed.

Why questions per hour is not true difficulty

A density index cannot see prompt complexity, implementation surface, testing conditions, candidate background, or scoring weights. A short-looking question may hide several rules; a progressive project can create rework when an early design shortcut breaks later levels. Proctoring, hidden-test feedback, and language constraints change risk without changing the arithmetic.

For those reasons, the index should sit beside a format checklist. Record the platform, whether questions are independent, whether levels build on the same codebase, whether a project or review is included, and whether the timer applies to the whole test or to each section.

A practical preparation plan

Start with the invitation or assessment setup screen. Write down the actual timer, the unit of work, whether you can revisit questions, the allowed language, and any proctoring rules. If the employer has not published the format, use the index only to choose a rehearsal pace; do not assume the row is a guarantee.

Then run three rehearsals:

• Density rehearsal: six short problems in 60 minutes, with five minutes reserved for testing and submission.
• Balanced rehearsal: two medium problems in 60–75 minutes, switching after a fixed checkpoint instead of getting stuck.
• Depth rehearsal: one 90-minute project or multi-level implementation, with a clean interface and regression tests after each feature.

After each run, record reading time, first compiling solution time, test time, and the point where you changed tasks. That breakdown tells you whether your issue is algorithms, prompt parsing, debugging, or transitions. A higher questions-per-hour target only helps if you know which component is consuming the clock.

Question	Practice use
Find the Earliest Pair with a Target Sum	Fast implementation, tie-breaking, and edge-case checks.
Compute a Depth-Weighted Sum of a Nested List	A medium problem where parsing and depth tracking must be decided early.
Design a Concurrent Image Processing Service	Long-form design and concurrency reasoning.
Debug Distributed System Performance Problems	Investigation under incomplete information rather than pattern recall.
Implement a Set with Readable Snapshots	Progressive implementation and API compatibility practice.

Common mistakes when reading an OA index

Do not divide by an invitation window. A test that stays open for 24 hours may still contain a 90-minute timed section. Do not count every subpart as a new question when the candidate described one project. Do not average reports from different job levels or platforms without retaining the format label. Do not treat an absent duration as unlimited time. Finally, do not turn a candidate memory into a company promise; the invitation and assessment rules control your actual session.

Frequently asked questions

What is a good questions-per-hour target?

There is no universal target. Use the employer’s stated format first. For a short-question screen, practice around the reported density while preserving a testing buffer. For a project or multi-level assessment, measure completed requirements and regression-test time instead of forcing a question count.

Does a higher index mean an easier OA?

No. It means more explicitly counted tasks per timed hour in the selected reports. Short questions may be easier, but a high-density set can also use long prompts or strict section timers. Difficulty requires separate evidence.

Why are some companies missing?

The report may state a count without a duration, a duration without a count, or a format that cannot be compared with independent coding questions. Excluding it is more honest than assigning a zero or guessing a denominator.

Are these official 2027 company rules?

Platform documentation can verify that certain assessment structures are available, but only the employer’s current invitation and assessment screen define your test.

Sources and Further Reading

• CodeSignal: General Coding Assessment structure — four questions and a 70-minute assessment window.
• CodeSignal: Industry Coding Assessment rules — one project-based question with four progressive levels and a 90-minute maximum.
• HackerRank: Section-based testing — employer controls for question counts, random selection, and section time limits.
• HackerRank: Best practices for screening — explains question shuffling and time constraints as configurable test controls.

Recheck the assessment invitation and platform rules before relying on any company-specific timing.*