For companies · Coding test for hiring

Coding test for hiring
what a good one measures, and how ours works.

A coding test should tell you how a candidate works in code like yours, not how many puzzles they have memorised. This guide covers what to measure, which format to pick, and how the Revvox coding round does it: a timed task in a small real codebase, tests run in a sandbox, every run on record.

By the Revvox team · Updated 19 October 2026

The short answer

A coding test for hiring is a short programming task that candidates complete before a live interview, so you only spend interview time on people who can write working code. A good one looks like the job: reading existing code, fixing a bug or extending a feature, and running tests. The Revvox coding round does this in the browser: candidates open a link, work on a small multi-file TypeScript service, run the tests in a sandbox, and submit. Your team sees every run and the result of hidden tests on the final code. It is in early access.

What a coding test should measure

Most engineering work is changing code that already exists. A test built from algorithm puzzles mostly measures practice with puzzles, which favours people with time to grind and can screen out experienced engineers who do not. A test built from a small, realistic codebase measures what the job needs: reading unfamiliar code, finding the right place to change, making the change safely and checking it with tests.

  • Can they read code they did not write and find their way around it?
  • Do they run the tests, read the failures and use them?
  • Is the change correct on cases they were not shown?
  • Is the code they hand in something a teammate could review and merge?

Example coding round tasks

The kind of task a realistic coding round uses, written by us. Two of these shapes are in our library today; the rest show how to write your own.

  1. A refund function in a billing service pays back more than was charged. Find and fix it.Reading the existing code, writing or running a test that shows the bug, and a fix that does not break other cases.
  2. An orders API returns every order at once. Add paging without breaking existing clients.A backwards-compatible change, sensible defaults and limits, and tests for the first and last page.
  3. A date-range filter drops records on the last day of the month. Fix it.Boundary handling and time zones, checked with a test on the edge case.
  4. Add input validation to an endpoint that accepts user profiles.Clear errors, no crash on bad input, and the happy path still working.
  5. A function is correct but too slow on large inputs. Make it fast enough.Finding the costly part, a better approach, and tests still passing.

Timed, take-home or live: which format

Many teams combine them: a short timed task on realistic code to narrow the list, then a live round with an engineer for the few who pass. The timed task should be short enough that good candidates do not drop out, which usually means under 90 minutes in total.

Three common coding test formats
FormatGood forWatch out for
Timed online taskScreening many candidates fairly, with the same task and time for eachPuzzle-style tasks that test recall more than work
Take-home projectSeeing design and code quality on a larger taskDays of unpaid work; strong candidates often decline
Live coding interviewHearing how someone thinks and takes hintsCosts an engineer's hour per candidate; nerves skew results

How to design a fair coding round

  1. Start from your own work. Pick a bug or a small feature like the ones your team handles every week, in a codebase small enough to read in ten minutes.
  2. Write tests the candidate can run. Visible tests show them what done looks like. Keep a second set hidden to check the final code on cases they were not shown.
  3. Set a time that fits the task. Time your own engineers doing it, then allow more. A rushed task measures speed, not skill.
  4. Give everyone the same conditions. Same task shape, same time, same instructions. Vary the details per candidate so answers cannot be passed around.
  5. Decide what passing means before you look. Agree the bar first, so the result does not depend on who reviews it.

How the Revvox coding round works

The coding round is part of Revvox Dial, and sits between Vira's phone screen and the final interview. It is in early access, set up with each company by a founder.

  • Candidates open a private link: no account and nothing to install. They see the company, the role, the tasks and the time allowed, and give consent before anything is recorded.
  • They work in a browser editor with a file tree, tabs and TypeScript highlighting, on a small multi-file service on Node.js.
  • They run the tests in an isolated sandbox as often as they like. The timer is kept on our server, so a reload costs nothing.
  • When they submit, or at the time limit, hidden tests they never saw run on the final code.
  • Your team sees each run: when, how many tests passed, how long it took and the failures, next to the phone-screen report.
  • Each candidate gets their own variant of the task, with a different planted bug and different numbers.
  • Today's library: a billing service whose refunds pay back too much (45 minutes, fix a bug) and an orders API that needs paging (60 minutes, extend a feature). Pick one or two tasks per role, up to 90 minutes in all.

What is not built yet

  • Invites by WhatsApp and email: today you share the candidate's link yourself.
  • Other languages: tasks run in TypeScript on Node.js today.
  • Vira in the room, talking the candidate through the task by voice.
  • A marked report scoring skills against a written rubric. Today your team sees the test results and every run, and decides.

Pricing

Early access, priced with you.
No public price yet.

Founding companies set it up with us a few at a time, and agree a price per round before the first candidate. Phone screening is live with public plans.

Mock interview prices

Questions

Questions
people ask about this.

Is the coding round available now?
In early access. We are setting it up with a small group of companies, one at a time. Ask for early access and a founder will get in touch.
Which languages can candidates use?
TypeScript on Node.js today. Other stacks will be added to the library, and we will say when they are ready.
Does AI decide who passes?
No. The round shows your team what the candidate did and which tests pass. Your team decides who moves on.
Does the candidate see their result?
No. They see their own test runs while they work. The hidden tests and the results go only to your team.
Do you detect cheating?
We do not claim to. Each candidate gets their own variant of the task, so an answer passed around does not fit the next person's, and you see every run they made on the way.
How is this different from a puzzle platform?
Tasks are small, realistic codebases with their own tests, closer to everyday engineering work than algorithm puzzles.

Try the coding round on your next tech hire.

Early access opens to a few companies at a time. A founder reads every request.

Get early access to Revvox

Takes a minute. A founder reads every one.

  1. 1 You
  2. 2 Your needs
Which product do you want?