Skills test before or after an interview? How to choose
Decide whether a skills test should come before an interview. Compare staff time and candidate effort, avoid repeated assessments and adapt an invitation template.
· 15 min read

A candidate finishes your writing test, then learns that the role requires evenings they cannot work. You have an assessment to mark, but neither of you is closer to a hire.
Confirm the working arrangements first, then choose the earliest assessment that can answer a real hiring question without asking for unnecessary candidate work.
A short customer-reply task may be useful before an interview for a writing-heavy role. A substantial take-home deserves a fuller conversation about the opportunity first. Compare the staff and candidate effort in each route.
What should happen before a skills test?
Explain the responsibilities, pay, workplace and hours before requesting an assessment. Read the application for answers already supplied. An unresolved availability question may need only an email:
Before I send the assessment, could you confirm whether you can work Tuesdays and Thursdays, 2–10pm, at [workplace]? We need both shifts covered for this support role.
Let me know whether both days work or which hours need discussing. You don’t need to explain your personal circumstances.
Skip the question if you already have the answer. This wording asks one person to work both shifts; change it if applicants can choose either day.
Then match the assessment to what you need to learn:
| What do you need to learn? | What to ask for |
|---|---|
| What did the applicant do in a team project? | A short conversation about their own actions and decisions. |
| Can they write a clear customer reply? | A reply to a customer message, using a policy you provide. |
| Can they interpret a report? | A small report to check, with an explanation of what they found. |
| Why did they approach the task that way? | A discussion of the answer already submitted, not the whole task again. |
A work-sample test asks applicants to do something the job involves, such as writing a customer reply or checking a report. This advice concerns practical work-sample tasks rather than personality questionnaires.
When should the test come first?
If your support team answers customers by email, ask applicants to write one customer reply. Supply the message and the relevant policy. You can then assess the work itself rather than ask whether they are “good at customer service”.
Keep an early task short. Discuss the opportunity before requesting a substantial take-home project so the candidate can decide whether to invest that time.
Give applicants the information and tools they need. Assess what the role requires them to know already, not company procedures you intend to teach. The US Office of Personnel Management’s work-sample guidance makes this distinction between skills needed on entry and skills learned after selection.
A customer-reply exercise you can adapt
Give the applicant the message and the facts they are allowed to use:
Customer message: “My order arrived with one of the two mugs I paid for. I need the missing one by Friday. Can you send it today?”
What you know: The order shows two mugs. You have logged the missing item and asked the warehouse to check stock. The warehouse has not replied.
Policy for this exercise: Update the customer by noon on the next working day, even if the check is unfinished. Do not promise a dispatch or delivery date before the warehouse confirms it. Ask your supervisor for help if a replacement is unavailable.
Your task: Write the next email to the customer. Use only the information above. We are looking for an accurate explanation and a clear next step, not a particular phrase or a long answer.
For reviewer practice, one acceptable reply is:
I'm sorry one of your mugs was missing. We've asked the warehouse to check stock, but we can't yet confirm dispatch today or delivery by Friday. I'll update you by noon on the next working day, even if we're still waiting for the stock check.
The reply acknowledges the missing item, makes no unsupported delivery promise and gives an update time. It does not promise a replacement before stock is confirmed. Keep this sample out of the candidate's task and accept different wording that achieves the same things.
A friendly promise of Friday delivery fails the accuracy check. A reply that merely says “we are investigating” leaves the customer unsure when to expect news.
Agree your exercise policy, permitted tools and expected effort before sending the task. Use the answer in the interview to explore a decision that needs explaining. Assess the writing itself separately: explaining an email well is a different skill from writing it clearly.
When should you interview first?
Talk first when there is a real question about whether the job suits either side. An applicant who expects to redesign your support process may not want a job mainly answering tickets. Explain the day-to-day duties before asking them to complete a task.
A conversation can also clarify their own contribution to a team project. Ask which decisions and work were theirs, then decide which skills still need testing.
Keep the main assessment consistent for the vacancy. Different follow-ups can clarify different answers without giving every applicant a different test.
Could you discuss the task during the interview instead?
Use a short live exercise when you need to discuss decisions as the candidate makes them or introduce new information. Explain the format and allowed tools beforehand, including any agreed adjustments.
Choose it for that purpose. Writing while someone watches and asks questions differs from drafting independently, and it uses interviewer time too.
How much time could testing first save?
Compare the interview work avoided with the extra marking and administration. Then check that the earlier test still answers a relevant question. A cheaper stage that cannot distinguish the skills you need is not a useful saving.
Interview or test 60 applicants first?
Use a 20-minute interview plus five minutes for notes, with one interviewer. The writing task takes the candidate 20 minutes and one reviewer five minutes to mark.
In each route, 30 applicants reach the second step and everyone finishes the work requested. This lets you compare the workload. It does not establish that the same 30 people would progress, or that half your applicants should pass.
| Order | Work for the team | Staff time | Candidate time across the group |
|---|---|---|---|
| Interview first | Interview 60; mark 30 tasks. | 27 hours 30 minutes | 30 hours |
| Test first | Mark 60 tasks; interview 30. | 17 hours 30 minutes | 30 hours |
Staff timeCandidate time
Testing first saves ten staff hours under these assumptions.
Applicants spend 30 hours in total either way. Thirty people complete one 20-minute step; the other 30 complete both. The experience still differs: test-first asks everyone to do the task before speaking to your team.
See the calculation
Interview first: 60 × 25 minutes for interviews and notes, plus 30 × 5 minutes for marking = 1,650 staff minutes.
Test first: 60 × 5 minutes for marking, plus 30 × 25 minutes for interviews and notes = 1,050 staff minutes.
The difference is 600 minutes, or ten staff hours. Applicant time in either route is 30 × 20 minutes plus 30 × 40 minutes = 1,800 minutes, or 30 hours.
The result depends on how many applicants reach the second step. If all 60 pass the test, marking and interviewing take 30 staff hours. That is more than either route in the table because more people receive both assessments, not proof that testing first is inherently worse.
Do not make the pass mark stricter to preserve the ten-hour saving. Set the standard from the job and plan enough review time for the people who meet it.
Count every person involved. Two interviewers attending a 20-minute interview use 40 staff minutes before notes. Add preparation, scheduling, support and unfinished attempts. Reserve review time in the diary, too: a shorter task will not produce a quicker decision if nobody can mark it this week.
Should you review CVs before sending the test?
There is another decision earlier in the process: should every applicant receive the task, or should you review applications first? The next comparison starts with 300 applications. It is separate from the 60-person example, so do not add the two sets of totals.
Test all 300 applicants or review CVs first?
Compare inviting all 300 applicants with reviewing their CVs and inviting 100. Use an 80% completion rate for each route, a 20-minute task and five minutes to mark it. Allow two minutes per CV review, including the decision note.
| Work | Test all 300 | Review CVs, then test 100 |
|---|---|---|
| Completed tasks at the assumed 80% rate | 240 | 80 |
| Staff time reviewing CVs before the task | None at this stage | 300 × 2 minutes = 10 hours |
| Staff time marking completed tasks | 240 × 5 minutes = 20 hours | 80 × 5 minutes = 6 hours 40 minutes |
| Staff time for these steps | 20 hours | 16 hours 40 minutes |
| Applicants' time on completed tasks | 240 × 20 minutes = 80 hours | 80 × 20 minutes = 26 hours 40 minutes |
CV-first uses three hours 20 minutes less staff time with these inputs. Applicants spend 53 hours 20 minutes less on completed tasks in total. Those are separate savings for different people.
The tradeoff is that you receive 80 completed tasks instead of 240. Two hundred applicants are screened out before the task. Check why they were excluded: did their applications miss an agreed requirement, or did unfamiliar job titles and software names hide relevant experience?
A saving only helps if the process still gives you a useful shortlist.
The result changes if reading each CV takes four minutes. CV-first would then use six hours 40 minutes more staff time than testing everyone, while still asking fewer applicants to do a task. Use the time a proper review takes; do not rush the review to make this route look cheaper.
Check the calculation with a longer CV review
At four minutes per CV, 300 reviews take 20 hours. Marking 80 completed tasks takes another six hours 40 minutes: 26 hours 40 minutes altogether, versus 20 hours to mark 240 tasks without the initial CV review.
The routes use equal staff time when the average CV review takes two minutes 40 seconds: 300 reviews then take 13 hours 20 minutes, plus six hours 40 minutes of marking. This is a comparison of these inputs, not a target review speed.
These totals stop before interviews, scheduling and support. Count that work before comparing the full process. Include any CV reviews done after the test as well; moving them later does not remove their cost.
How to count time spent on unfinished tasks
In the CV-first calculation, 20 of the 100 invitees do not finish. If ten never start and ten spend six minutes before stopping, add an hour of candidate effort. The total becomes 27 hours 40 minutes, including completed and unfinished tasks. Keep staff time separate.
You need the same information for the other route before comparing total candidate effort. A missing submission does not mean the person spent no time on the task.
Did the faster route still produce a useful shortlist?
Ask the hiring manager what they can now judge from the applications and what remains unanswered. Keep the time spent clarifying, correcting and repeating work in the comparison. A stage that saves one recruiter time may simply move the work to the next person.
Check exclusions against the job, not just past decisions
Reproducing past decisions can look reassuring without showing that a process selects well. In a set of 100 applications with five hires, predicting “not hired” for everyone agrees with 95 records, and misses every hire.
The other 95 records are not all failed assessments either. Some people may have withdrawn, never been interviewed or applied for a role that was cancelled. Check what the old outcome actually means before using it as a standard.
Research on hiring and exploration addresses the underlying problem: employers usually lack comparable outcome information for people they never interviewed. Agreement with old decisions cannot fill that gap.
For CV-first screening, ask a second reviewer to examine some excluded applications against the requirements before seeing the original verdict. Include borderline cases and unfamiliar descriptions of relevant work. Record overlooked evidence and questions worth asking, not an invented performance score for someone who never took the assessment.
For test-first screening, distinguish an unsuitable answer from a missing submission. Check for inaccessible links, help requests and withdrawals. Feedback from people who reply can explain their cases; it cannot tell you why everyone else stopped.
Alongside the time figures, record what you learned, any relevant experience the first review missed and where candidates are waiting. Those findings tell you what to change next.
Email template: invite a candidate to a skills test
Ask a colleague to try the task using only the instructions the candidate will receive. Find out where they needed help and whether the time estimate was realistic. The invitation describes an untimed exercise expected to take about 20 minutes.
Subject: Next step for the support role: a short writing exercise
Hi Sam,
We’d like you to complete a short customer-reply exercise for the support role. It includes the situation and policy you need. It should take about 20 minutes and isn’t timed.
We’ll assess how accurately and clearly you explain the next step to the customer, not whether you know our product.
Read the task: [Task link or named attachment]
Send your answer: [Submission link or monitored email address]
Submission deadline: [Date, time and time zone]
Tools and help you may use: [The agreed rule, including AI use]
The support manager will review your answer. I’ll email you by [update date] to let you know where your application stands.
Please reply if you need help, more time or a different format, including if a problem comes up after you start.
Alex
Test the links without recruiter permissions, fill in the tool-use rule and book the reviewer’s time. If reading and submitting happen on one page, one clearly labelled link is enough. Leave time between the submission deadline and your promised update.
If the task is timed, replace the untimed sentence with the actual limit and what starts the clock. Do not leave “about 20 minutes” as the only instruction for a test that closes automatically.
Say whether applicants may use reference material, AI or another person’s help, and what they must disclose. Writing without assistance and improving an AI draft demonstrate different things. Decide which you want to assess before writing the instructions.
Use an exercise designed for assessment, not free work on an unfinished company project. For hiring in Great Britain, Acas’s interview guidance covers adjustments and payment considerations for work trials; calling work a “test” does not settle those obligations.
If the candidate reports a broken task link
Reply to the problem rather than send the invitation again. After confirming that the deadline and related reminders have been paused, you could write:
Hi Sam,
Thanks for letting me know the task link won’t open. I’m checking it with our team and have paused your assessment deadline. You don’t need to keep trying the link.
I’ll email you by [date and time], even if we’re still fixing it. Once the task is accessible, we’ll agree a new submission deadline.
Alex
Check that the task, reminders and reviewer all use the revised deadline. If the system cannot pause it, arrange a workable alternative before promising one. Once the problem is fixed, send the working link and agreed date rather than expect the candidate to keep checking.
Reserve time to review the work you request
With 40 results arriving each working day and eight minutes needed per review, the team needs five hours and 20 minutes to keep up. Two hours clears only 15 results, leaving 25 more waiting each day. These are planning inputs, not a recommended review speed.
Sending a task earlier will not produce an earlier decision when marking remains unassigned. Name the reviewer and reserve the time before increasing invitations.
Test the actual invitation, task, review and follow-up you intend to use. A study comparing interview formats does not establish the best order for your writing task and interview.
Compare when each answer became ready, when someone reviewed it and when the candidate heard the decision. You will see which wait the new process actually reduced.
Follow the whole invited group, including people who do not finish. A reduced shortlist caused by broken links or an unreasonable workload is not evidence that the earlier test selected better candidates.
Screened’s interview-process guide explains the transcripts, topic summaries and assessments available to your team. Read the existing answers before adding a step, and keep a written task where writing remains unassessed.
How do you choose the order for your vacancy?
For the support role, three questions determine the sequence: do the hours work, can the applicant write an accurate customer reply, and can they explain their choices?
Confirm the first without an unnecessary meeting. Use the supplied writing task for the second, then discuss the answer rather than repeat the exercise. Agree the assessment standard before marking; the scorecard template can hold those criteria.
A proposal to the manager could say:
Let’s confirm the hours before sending the writing task.
That should avoid asking candidates to complete a task before discovering they cannot work the shifts.
We’ll keep the task and marking standard unchanged.
I’ll track how often the hours need clarification, how many tasks are completed and the review time involved. Can we agree who will handle those replies before we start?
Check the revised process with test applications before using it in an approved live trial. Note other changes, such as a new job board or reviewer, when you compare results.
Give the next interviewer the submitted work and the question still unanswered. You should be able to explain the sequence to the candidate just as clearly: “First we confirm the hours, then you try a short writing task, then we discuss your approach.”
Keep each stage only when it resolves a remaining question. Confirm practical arrangements before asking for work, use the submitted answer in the next conversation and assign the review time before expanding invitations.


