Why reading all four comes first
In an online assessment the problems are not always sorted by difficulty, and how hard one looks is a poor guide to how long it takes. About a minute and a half per problem tells you three things: what each one asks, what its input constraints are, and which ones you can already see a solution for.
The constraints matter more than the story around them. A limit of n ≤ 1,000 says an O(n²) loop is probably fine. A limit of n ≤ 200,000 says it is not, and that you need something closer to O(n log n). Knowing this before you start stops you writing the wrong solution twice.
Reading everything also lets part of your mind work in the background. A problem you read at minute four and parked often looks easier at minute forty.
Then rank them. The useful measure is expected points per minute: how confident you are, times what the problem is worth, divided by how long you think it will take. If no point values are shown, assume equal weight and rank by confidence.
A seventy-minute plan
Time management in an online assessment is a plan you make at minute six and check at fixed moments, not a feeling you consult when you get nervous. Once the four problems are ranked, a workable split looks like this.
| Block | Minutes | Running total | What you do |
|---|
| Read and rank all four | 6 | 6 | Note constraints and a first idea for each |
| Easiest problem | 10 | 16 | Solve, check edge cases, submit |
| Second easiest | 15 | 31 | Solve, submit a working version |
| Third | 20 | 51 | Brute force first, improve if time allows |
| Hardest | 14 | 65 | Brute force, or the cases you can handle |
| Return to best partial | 5 | 70 | Improve whichever is closest to full marks |
The numbers are not sacred, but every block has an end, and you look at the clock when it arrives. If the second problem is still broken at minute 31, either you can name the bug or you move on and come back. Overrunning a block without deciding to is how candidates finish with one perfect answer and three blanks.
Bank points early with partial credit
Many coding assessments score each problem by the share of hidden test cases it passes. Others score a problem as all or nothing. The instructions usually say which, and you should know before the timer starts. When partial scoring applies, a correct but slow brute force will often pass the small hidden tests and time out only on the large ones.
Do the arithmetic once so the trade is obvious. The figures below are an illustration, not any platform's scheme.
Assume: 4 problems, 100 points each, scored by share of hidden tests passed
Problem 3 has 15 hidden tests
Brute force, 20 minutes: passes the 6 small tests -> 6 / 15 x 100 = 40 points
Optimal attempt, 20 minutes: does not compile in time -> 0 points
Session A: 100 + 100 + 40 + 20 = 260 points (everything attempted)
Session B: 100 + 0 + 0 + 0 = 100 points (problem 1 polished, rest untouched)
So write the simple version, submit it, then optimise. Whether a later submission replaces an earlier one, and whether repeated submissions cost anything, differs between platforms and employers. Check the invitation and the instructions page before you begin, not during the session.
When to stop polishing one problem
A problem is done when it passes the samples, survives the edge cases its constraints imply, and has a complexity that fits the largest input. The usual edge cases are empty input where allowed, a single element, the maximum size, duplicates and negative values. Once those pass, renaming variables or shaving a constant factor buys almost nothing, because automated hidden tests check output and running time, not style.
If the instructions say a person will read your code later, readable names start to matter. Even then, a working answer to the next problem is worth more than a tidier version of this one.
The harder call is a problem that feels nearly finished. If you are one bug away and can say where the bug is, finish it. If you are "one bug away" but cannot say where, that is sunk cost talking. Submit what passes, jot a one-line note about the suspected bug, and move to the next block. Coming back twenty minutes later often shows you the problem in seconds.
Why page order and perfectionism lose points
The weak plan is to open problem one, solve it beautifully, and carry on in order. If problem two happens to be the hardest of the four, that plan spends the freshest thirty minutes of the session on the problem with the lowest points per minute. What the employer usually sees first is a score, not the care you took, so the order you attempt problems in is part of what is being tested.
Good time management here also means accepting an untidy final state. Four problems at 100, 100, 40 and 20 is a stronger online assessment result than one immaculate answer and three empty editors. If you want to practise this, run past problem sets under the same seventy-minute timer and review the order you chose, not only the answers.
Read everything, rank by what you can finish, time-box each problem, and submit the working version before you try to make it fast.
© 2026 Preptima. Originally published at preptima.com (opens in a new tab).