Do Partial Test Cases Count in an OA? Hidden Tests, Weighted Scores, and Cutoffs

Partial test cases often count in an online assessment, but the scoring unit matters. On many coding platforms, each individual test case is pass or fail: you receive all of that case's assigned points or none. The coding question can still earn a partial score because its total is the sum of the cases you passed.

Passing 7 of 10 cases does not always equal 70%. Hidden cases may carry different weights, samples may be worth zero, and some platforms add completion bonuses. Employers can then apply a separate cutoff or review.

The safest interpretation is: preserve every working case you can, but do not infer your hiring result from the pass count alone. Practice with PracHub interview questions and written solutions, record which edge cases you missed, and treat the exact invitation and in-test instructions as the source of truth.

Partial test case scoring in online assessments with hidden tests weights and cutoffs

Quick answer

Question	Short answer
Do passed test cases earn partial credit?	Often yes. Many coding questions add the points assigned to each passed case.
Is one test case partially correct?	Usually no. A case normally passes only when the output and execution requirements are satisfied.
Are all cases worth the same amount?	Not necessarily. Employers or question authors may assign different weights.
Do visible sample cases count?	Sometimes, but they may be assigned zero points.
Do hidden cases count?	Yes when they are scored; they commonly test boundaries, performance, and unseen paths.
Does a partial platform score clear the company cutoff?	It can, but the cutoff and later review process are employer-specific.

A solution that passes meaningful cases is usually better than a blank submission, but the dashboard may not reveal the employer's decision rule.

What "partial credit" can mean

Candidates use one phrase for several different mechanisms. Separating them prevents most scoring confusion.

Partial score for a coding question. Your code passes some test cases and fails others. The platform adds the points from the passed cases, so the question receives something between zero and full credit.

Partial credit inside one test case. This is less common for ordinary exact-output coding tests. A case generally succeeds only when the program returns the required output within the applicable limits. A nearly correct output is still a failed case.

Partial progress in a module. A platform can award base points for progress and reserve extra points for completing the whole module. CodeSignal's current Assessment Score uses this kind of two-tier model.

Partial credit on another question type. Multi-select questions, manually reviewed repository tasks, approximate-solution problems, and project assessments can use different rubrics. HackerRank, for example, documents partial credit for selecting some correct options in a multiple-answer MCQ, subject to its configured rules.

When someone says there is "no partial credit," determine whether they mean each case is binary, each whole question is all-or-nothing, or the employer uses a fixed threshold. Those are different rules.

HackerRank: why two official statements sound contradictory

HackerRank provides the clearest illustration. Its employer documentation says each test case has a predefined score. Passing the case earns that full score; failing it earns zero. It then says partial scoring does not apply. In the same section, it explains that the total question score is the sum of all passed cases.

Its candidate documentation describes the result from the other level: pass all cases for a full score, pass some cases for a partial score, and pass none for zero.

Both statements can be true:

One test case: binary, with no fractional credit inside that case.
One coding question: partially scored by adding the points from passed cases.

HackerRank's official example uses two sample cases worth zero, three easy hidden cases worth 5 points each, four medium cases worth 10 each, and two difficult cases worth 20 each. A candidate who passes all three easy cases, three medium cases, and one difficult case earns 65 out of 95.

That candidate passed seven of nine hidden cases, or about 78% by count, but earned about 68% of the available points. The difference comes from weighting. The missed difficult case costs more than a missed easy case.

Hidden tests can be worth more than visible tests

Visible samples explain input, output, and basic behavior; they are not necessarily a miniature score. HackerRank allows sample tests to carry zero points.

Hidden tests are unseen inputs used to evaluate paths the samples do not expose. Common categories include:

empty, minimal, or maximum-size inputs;
duplicates, ties, negative values, or overflow boundaries;
parsing details such as whitespace and output formatting;
disconnected graphs, cycles, or degenerate trees;
large inputs that distinguish an efficient algorithm from one that times out.

A solution can pass every sample and still score poorly because the hidden suite measures completeness and robustness. Fix visible failures first because they reveal a known mismatch, but do not stop testing after the samples turn green.

How major platforms handle partial results

Scoring changes over time, and employers can configure assessments. This comparison summarizes current official documentation rather than promising what your invitation uses.

Platform	How partial work can affect the result	Important qualification
HackerRank	Passed coding test cases contribute their predefined points; multi-answer MCQs can award partial credit	One coding case itself is pass/fail, and case weights can differ
CodeSignal	Base points reflect progress or completion within modules, while full module completion unlocks bonus points	A partly solved module and a fully solved module do not necessarily scale linearly
Codility	Task scores combine into the overall score; coding tasks can measure correctness and, where applicable, performance	Employers can weight tasks differently, and "performance" means algorithmic efficiency
CoderPad Screen	Working code can contribute points, reports can show point and comparative scores, and timed-out code is submitted for validation	Candidate-visible validators may differ from the final validators; employers can review the detailed work
Coderbyte	Coding test cases contribute points; custom unit-test challenges can score by the fraction passed	Questions can be weighted and employers can set a qualifying score

CodeSignal's completion bonus means finishing one nearly solved module may be worth more than scattering a little progress across several. Codility adds a vocabulary trap: its performance score can measure runtime behavior on large data sets, not how quickly the candidate clicked Submit.

Platform score and company cutoff are separate

The assessment platform calculates evidence. The employer decides how to use it.

A cut score is a threshold selected for a role or hiring process. CodeSignal's employer guidance explains that organizations calibrate it to the job, candidate pool, and hiring goals. The same platform score can therefore lead to different outcomes across employers or roles.

Companies may use:

an absolute score threshold;
a benchmark or comparative score;
different thresholds by role or location;
resume, eligibility, integrity, or work-style review;
manual review for borderline or project-based submissions.

An anonymous claim that "the cutoff is 70%" is therefore weak evidence without the exact employer, role, location, and assessment version. Partial credit can matter without being sufficient: the same score may advance at one employer, miss another threshold, or enter manual review.

Decision flow for protecting partial credit and handling hidden tests in an online assessment

What to do when only some tests pass

Protect working behavior before chasing a full score:

Save the strongest working version so a risky change does not erase earned cases.
Read the failure category. Wrong answer, runtime error, timeout, memory limit, and formatting problems need different repairs.
Map failures to the contract. Recheck empty inputs, boundaries, duplicates, tie rules, indexing, output order, and numeric limits.
Check complexity. If small cases pass but large cases fail, inspect time and space complexity before adding conditionals.
Change one thing at a time, retest, and submit the stable version before time expires.

Avoid hardcoding visible examples or guessing hidden inputs. The useful goal is a general solution you could defend in a technical follow-up.

How to allocate the final minutes

For multi-question tests, use the scoring model you can actually observe.

Situation	Best next move
A nearly complete question has one identifiable boundary bug	Fix and retest it
A blank question has a straightforward baseline solution	Implement the baseline and bank valid cases
A working solution times out only on large inputs	Improve complexity if the change is localized and understood
Several questions have scattered progress, and one is close to completion	Finish the closest one when the platform rewards completion bonuses
The remaining fix requires a full redesign with little time	Preserve the stable submission and document the idea for later practice

Choose the action with the clearest path to additional verified behavior, not cosmetic polishing or an emotional reaction to one hidden failure.

Practice hidden-test discipline with PracHub

These PracHub question-bank records train edge-case coverage and partial-progress judgment. They are practice material, not predictions of your exact assessment. Each complete title in the first column opens the question and written solution.

PracHub question	Practice focus	Why it helps
Solve Five OA Coding Tasks	Independent tasks, parsing, and hidden suites	Trains banking progress across several separately graded implementations
Solve Two OA Coding Problems	In-place updates, integer operations, and boundaries	Shows how two complete subproblems can demand different test strategies
Group strings that are anagrams	Hashing, canonical keys, and duplicates	Exposes hidden failures caused by key design and repeated values
Merge Overlapping Intervals	Sorting, endpoint rules, and invariants	Forces a precise decision about touching and overlapping boundaries
Design comprehensive OA test cases	Equivalence classes, oracles, and stress cases	Turns hidden-test preparation into a reusable checklist

Frequently asked questions

If I pass 7 of 10 test cases, is my score 70%?

Not necessarily. It is 70% only if all ten cases carry equal points and there are no separate bonuses, penalties, or task weights. Some platforms let question authors assign different case values, and visible samples can be worth zero. Treat 7/10 as coverage feedback unless the scoring rubric confirms equal weights.

Do hidden test cases count more than sample tests?

They can. Hidden cases may receive the same weight, a higher weight, or simply be the only scored cases. HackerRank's official example assigns zero to sample cases and increasing values to easy, medium, and difficult hidden cases. Your own assessment may use another configuration.

Can partial credit still get me an interview?

Yes, when the partial score clears the employer's threshold or the submission receives favorable manual review. It is not guaranteed. Employers can combine the assessment with resume fit, eligibility, integrity signals, hiring capacity, and later screens.

Should I attempt every question or finish one question completely?

Use the visible scoring structure. If tasks are independently scored, a simple working baseline on each can protect points. If full completion unlocks a bonus, finishing a nearly solved module may be more valuable. Always preserve working submissions before switching.

Can recruiters see which hidden tests I failed?

Authorized reviewers can often see question results, code, execution information, and activity history while candidates see only a count or generic failure. Keep the final code readable and defensible.

Final takeaway

Partial test cases often count, but they do not all have to count equally. A coding case is commonly binary, while the whole question receives the sum of points from every passed case. Hidden cases can be weighted, samples can be unscored, and platforms such as CodeSignal can reserve additional value for complete modules.

Your best response is not to guess the hidden cutoff. Preserve working code, test the contract systematically, fix the highest-confidence failure, and submit a stable version. Then use the outcome to improve the next attempt: solve a PracHub coding question, build your own boundary tests before opening the explanation, and record which hidden-case category your first solution missed.

Sources and Further Reading

Research note: This guide was checked on August 29, 2026. Assessment providers and employers can change scoring, weighting, cutoff, and reporting settings; your invitation and in-test instructions control your specific attempt.