The rise of skills tests in hiring and how to prepare

A decade ago, most hiring outside software and a few professions ran on the same three steps: application, interview, references. Somewhere in the last few years a fourth step has become normal almost everywhere. You are asked to write a memo, analyze a spreadsheet, record a video answering scenario questions, sit a timed reasoning test, or spend a weekend building something the company will evaluate. Marketing coordinators get writing tests. Operations managers get case studies. Customer service candidates get situational judgement tests before a human has read their application.

This is a real shift and it is still moving. This article looks at why employers are doing it, what the common formats are actually measuring, how to prepare for each, and the increasingly important question of when a “test” has crossed the line into unpaid work you should decline.

Why employers are adding tests

Several forces are pushing in the same direction, and they reinforce each other.

Applications have stopped being informative. AI writing tools mean a polished cover letter and a keyword-matched resume can be produced in minutes by anyone. Employers report that the top of the funnel has become both larger and less useful. A work sample is a way to find out what a person can do when the words on the application no longer tell them. We covered the application side of this at /news/how-ai-screening-is-changing-the-way-you-should-apply-for-jobs/.

Degree requirements are being dropped. When a company stops using a degree as a filter, it needs something else to stand in for it. Skills-based hiring, in practice, often means skills-tested hiring.

The research has been on the side of work samples for a long time. Industrial psychology has been telling employers for decades that structured tests of the actual work, along with structured interviews and general ability tests, predict job performance far better than unstructured interviews or years of experience. Schmidt and Hunter’s well-known meta-analyses made this case in the 1990s; what changed is that the tooling to run such tests cheaply at scale became available.

Mis-hires are expensive and remote work made them more likely. A bad hire discovered at month four costs many months of salary once you count recruiting, lost output and the re-hire. Managers who are hiring people they will rarely meet in person want more evidence before committing.

The result is that a candidate applying for a mid-level office role in 2026 should expect at least one assessment step and should not be surprised by two or three.

The main formats and what they are really testing

Knowing what the employer is looking for changes how you prepare. Most tests fall into one of six types.

Take-home assignments

A brief, a deadline of a few days to a week, and a deliverable: a marketing plan, a data analysis, a piece of code, a design, a written report. This is the format closest to real work, which makes it the most informative for the employer and the most burdensome for you.

What it tests: whether you can scope a problem, make sensible assumptions, produce something finished, and explain your choices. It also, unavoidably, tests how much time you are willing to give away.

How to prepare: read the brief twice and write down the two or three things it is really asking for. Decide on a time budget before you start and stick to it. Produce something complete and modest rather than ambitious and unfinished. Include a short note (half a page at most) on what you assumed, what you would do with more time, and what you chose not to do and why. That note shows judgement rather than just effort.

Live exercises and working sessions

You are given a problem in the room (or on the call) and work through it with an interviewer, sometimes alone with them watching. Pair programming, a whiteboard case, a “here is a customer complaint, draft the response now” exercise.

What it tests: how you think under mild pressure, whether you ask good questions, and how you take feedback. The interviewer is usually less interested in whether you reach the right answer than in watching how you get there.

How to prepare: practice thinking aloud. Most people go silent when they concentrate, which leaves the interviewer with nothing to assess. Narrate your reasoning, state assumptions, and when you are stuck say what you are stuck on and what you are considering. Ask a clarifying question early; it is almost always rewarded. Do not pretend to know something you do not.

Situational judgement tests

Multiple-choice scenarios: “A colleague misses a deadline that affects your work. What do you do?” with four or five options to rank or choose from. Common in customer service, retail, public sector, graduate schemes and healthcare recruitment.

What it tests: whether your instincts match the organization’s stated values and the way it wants people to handle conflict, priorities and customers.

How to prepare: read everything the company publishes about its values and how it describes its culture, because the scoring key was written by people who believe those statements. In most SJTs the preferred answers involve taking ownership, communicating early, escalating appropriately rather than either ignoring a problem or going over someone’s head, and treating customers or colleagues with respect while still following process. The extreme options (do nothing; confront angrily; go straight to senior management) are almost always scored low. Free practice tests exist for the major formats and half an hour on one will show you the pattern.

Coding and technical tests

Timed problems on platforms like HackerRank, Codility or a company’s own system, followed sometimes by a live technical interview. Also used for data roles (SQL and analysis tasks) and some finance jobs.

What it tests: core competence, and increasingly, how you use AI tools. Some employers now permit or expect assistant use and assess your ability to direct and verify it; others ban it and monitor for it. Find out which before you start.

How to prepare: practice the platform’s format, not just the skills, because unfamiliarity with the interface costs more time than most people expect. For data roles, be ready to explain your query choices as well as write them. In live sessions, the same thinking-aloud rule applies.

Writing and communication tests

A short written task under time limits: summarize a document, draft an email to an unhappy client, write a briefing for a manager. Increasingly common for any role where writing matters, partly because AI has made unsupervised writing samples worthless.

What it tests: clarity, structure, tone and the ability to work out what the reader needs. Not vocabulary, and not length.

How to prepare: write short. Lead with the point. Use headings if the task is more than a few paragraphs. Read it once for anything you would be embarrassed to have a manager see. If the brief specifies an audience, write for that audience and nobody else.

Aptitude and psychometric tests

Numerical, verbal and abstract reasoning tests, and personality inventories. Widely used in graduate recruitment and at large employers.

What it tests: general reasoning speed and, for personality tools, fit with a profile the employer has decided it wants. Personality tests generally have no right answers, though they often have consistency checks that catch people trying to game them.

How to prepare: reasoning tests reward practice more than any other format, because much of the difficulty is time pressure and question style rather than underlying ability. Two or three practice sessions on the same type of test typically improve scores noticeably. For personality inventories, answer honestly and consistently.

One-way video interviews

Not exactly a test but often bundled with one: recorded answers to preset questions, sometimes scored by software before a person watches. The practical setup and delivery advice at /job-search/how-to-prepare-for-a-video-interview/ covers these in detail.

A preparation checklist that works for any format

• Ask what the assessment is, how long it should take, and how it will be evaluated. Employers who cannot answer the third question have not thought it through, which is useful to know.
• Ask whether AI tools are permitted. Get the answer in writing if the test is significant.
• Find out who will see your work. A take-home reviewed by the hiring manager is worth investing in; one screened by a junior recruiter against a checklist is not.
• Set a time budget and tell the employer what it was in your submission. It shows you can manage scope, and it protects you from the assumption that you spent a weekend on it.
• Do the test in the conditions it will be assessed in. Practice a timed test timed.

When to decline unpaid work

Here the trend has a genuine problem attached. Some “assessments” are not tests; they are free consulting. The signs are consistent enough that you can spot them.

The brief uses the company’s real, current problem rather than a constructed one. “Write a content strategy for our Q4 launch” is not an assessment.

The deliverable would be usable by the company as-is. A full campaign plan, a working feature, finished designs, a client-ready report.

The expected time is large. There is no fixed line, but once an unpaid task exceeds about a working day, it has stopped being a sample and become a job.

There is no rubric, no defined scope and no named reviewer, and the deadline is tight for the amount asked.

The task comes early in the process, before anyone has spoken to you, and to a large pool of candidates. That is the pattern of a company crowdsourcing ideas.

You do not have to refuse outright. The better move is to renegotiate. A script that works:

“I am glad to do a work sample and I want to give you something you can assess properly. The brief as written looks like several days of work, and I want to be sure I understand the scope. Would it work to limit it to [a specific narrower piece], with a two to three hour time budget? Alternatively, I would be happy to walk through my approach live with the team, or to complete a paid trial project if you want to see finished output.”

Most employers acting in good faith will accept one of those options or explain what they actually need. An employer who refuses all three and insists on the full unpaid deliverable has told you something about how it treats people who work for it.

Two situations deserve a pause before declining: a role that is a real step up with an employer you specifically want, and creative fields where speculative work has long been the norm. In both cases, never hand over editable source files or full working code until there is a contract, and put a note on anything you submit stating that it is a sample for assessment and remains your property.

What this means for how you plan a search

Budget for it. If each serious application now carries two to eight hours of assessment work, the volume strategy of firing off fifty applications a week does not survive. Fewer, better-targeted applications where you are willing to do the test properly beat many where you rush it. And keep a copy of everything you submit: a good take-home, with the company’s details removed, becomes a sample you can show the next employer.

And keep some perspective. Skills tests are, on balance, good for candidates who can do the work but do not interview well, do not have the right degree, or do not have a network. The fairness problems are real, mostly around unpaid time, but the underlying shift toward judging people on what they can do rather than who they know or where they studied is one that most readers of this site should welcome. The task is to make sure you are being tested, not used.