What an OA actually is
A timed, auto-graded coding test, usually sent within days of applying, typically 60 to 90 minutes for two to four problems. You write code in a browser, it runs against hidden test cases, and a score goes into a system. In most pipelines no human reads your code unless the score clears a threshold.
- No partial credit for approach. An elegant solution that fails three edge cases scores below an ugly one that passes everything.
- Hidden tests. You cannot see what you are failing, only that you are.
- Hard time limit. Running out of time on problem three is extremely common and is the main thing to design your strategy around.
- Sometimes behavioral or aptitude sections bolted on, which are usually screening for consistency rather than scored heavily.
How to actually take one
- Read every problem first. Two minutes spent reading all of them tells you which is cheapest, and the order they are presented in is not the order you should solve them.
- Take the easy one first, completely. A finished easy problem outscores two half-finished mediums.
- Write the brute force if the clock is against you. A correct slow solution that passes most tests beats an unfinished optimal one.
- Handle the obvious edge cases explicitly - empty input, single element, duplicates, negative numbers. Hidden test suites are full of them.
- Watch the clock per problem, not overall. Set a budget and abandon to the next problem when you hit it. Returning later is cheap; running out with one problem untouched is not.
What to practise, and how
The format is the thing you are practising, not just the problems. Working untimed problems one at a time is a different skill from clearing three problems in seventy minutes with no feedback.
- Run full timed sets: pick three problems, set 70 minutes, no notes, no searching, no stopping.
- Do it in a plain browser editor with no autocomplete, since that is the environment you will be in.
- Afterwards, grade yourself only on tests passed. Not on whether your approach was clever.
- Build fluency in the handful of things you will type under pressure - a hash map, a sort with a custom key, a breadth-first traversal - so none of them costs you thinking time.
For the underlying problem material, work the Blind 75 by pattern and the 15 patterns guide for recognition. Amazon's is the most documented variant, covered in the Amazon online assessment.
What a score actually means
Thresholds vary by company and by season, and nobody outside publishes them reliably. What is consistent is the shape: passing most or all test cases on most problems clears the bar at most companies, and the marginal value of an elegant solution is close to zero. Optimise for test cases passed and nothing else.
One honest caveat: some OAs are proctored, and some are not. Where a company states conditions, follow them. Using assistance against a stated rule is not a grey area, and the downside - a rescinded offer, a blacklist - is considerably worse than a weak score.
Where CoPilot Interview fits
Disclosure: we build an interview assistant, so treat this as a disclosed pitch. CoPilot Interview is built for live rounds - the coding, system design and behavioral conversations where a human is talking to you - with a permanent free tier that needs no credit card. For the online assessment itself, prepare so that you do not need help, and follow whatever conditions the company has set.
Frequently asked questions
How long is a new grad online assessment?
Typically 60 to 90 minutes for two to four problems, sent within days of applying. Some companies add a behavioral or aptitude section, which is generally screening for consistency rather than being heavily scored.
What score do you need to pass an online assessment?
Thresholds vary by company and hiring season and are not published. The reliable guidance is that passing most or all hidden test cases on most problems clears the bar in most pipelines, and solution elegance adds essentially nothing to the score.
Should I write the optimal solution in an OA?
Only if you can finish it. Online assessments give no partial credit for approach, so a correct brute force that passes the tests scores above an unfinished optimal solution. Write something that works, then optimise if time remains.
Why do so many candidates fail the online assessment?
Because it is a different exam from an interview and most people prepare only for interviews. There is no interviewer to give hints, no credit for explaining your reasoning, and a hard time limit that punishes spending too long on one problem.
Walk into the round with a second pair of eyes
CoPilot Interview is a native Windows and macOS app that puts suggested answers on your own screen in about four seconds, across coding, system design and behavioral rounds. The free tier is permanent: 15 minutes a day, no credit card.
Try it free