How to Choose a Pre-Employment Assessment Tool: A UK Buyer's Guide
Most pre-employment assessment tools look alike from the outside. A candidate completes something online, a score comes back, and the vendor's site says the score predicts performance. The differences that decide whether the tool improves your hiring or just adds a step are underneath that surface, and they are not hard to find if you know which questions to ask. This guide is the set of questions, in the order that saves the most time, for an HR team, a hiring manager or an agency owner choosing a tool for the first time or replacing one that has not earned its place.
It assumes you already know what an assessment is for. If not, what is an employee suitability assessment? covers the ground and compares the main formats. This piece is about choosing between tools once you have decided you need one.
Start with the decision, not the demo
Before looking at any product, write down which hiring decision the tool has to support and how often. Screening two hundred applicants for a customer service team is one job. Choosing between three shortlisted finance managers is another. Confirming a senior appointment that will take two years to unwind if it goes wrong is a third. A tool that is excellent at one of those can be useless at the others, and vendors are rarely candid about which one they are built for. Volume, stakes and the point in the process where the tool sits are the three facts that filter the market before a single demo.
The eight questions
1. What does it measure, and where does that come from?
Ask to see exactly what a candidate is assessed on for one of your real roles, and ask where that list came from. There are two honest answers. Either the tool builds the assessment from your role, so the scenarios and criteria are specific to the job, or it draws from a library built for a job family, so you are getting the generic version. Both can be legitimate, but the second is much more common and is often described as if it were the first. A test that cannot tell a marketing manager from a finance manager is measuring something, but it is not measuring suitability for your vacancy.
2. What is the evidence that it predicts performance?
Every vendor will say their tool is validated. Ask what the method is and what the research behind that method says, rather than accepting a figure from the vendor's own case study. The independent research base for selection methods is decades deep and it is not secret: work samples and structured methods rank near the top, unstructured interviews and personality inventories well below. The science behind evidence-based hiring sets out the primary sources. A vendor who can place their method in that literature is on solid ground. A vendor who can only cite their own numbers is asking you to trust them.
3. Can you see why a score is what it is?
Take a sample report and pick any score in it. Can you trace it back to something the candidate actually did or wrote? If the report offers a number and a paragraph of generated prose with nothing attached to the candidate's own words, you cannot check it, you cannot explain it to the hiring manager, and you cannot defend it if the unsuccessful candidate asks why. This is the single question that separates tools you can act on from tools you can only quote. It also matters more from January 2027, when the ordinary unfair dismissal qualifying period in Great Britain falls from two years to six months: the record of why you appointed someone becomes the record of a decision you may have to stand behind sooner.
4. How does it handle fairness?
Three sub-questions. What does the scoring actually look at, and is any of it a proxy for a protected characteristic? Personality-based scoring and anything described as culture fit deserve a hard look here. What adjustments are available for candidates who need them, and can you make them without the vendor's help? And what does the tool give you to check for adverse impact after the fact: can you see progression rates by group, or is that left to you to reconstruct from a spreadsheet? The Equality Act 2010 makes the outcome your responsibility regardless of whose tool produced it, and adverse impact checking is easier when the tool was built expecting to be checked. Tools that score observable behaviour against the requirements of the role start from a better position than tools that score traits.
5. What is it like to be the candidate?
Complete the assessment yourself, on a phone, at the point in the process where a real candidate would meet it. Note how long it took, whether the instructions made sense, whether it felt like a fair chance to show something or like a hurdle. Candidates you want are usually candidates other employers want too, and a forty-minute personality battery at first contact loses some of them before anyone has seen them. The right length depends on the stakes of the decision: a short scenario is proportionate for a first screen, a longer assessment is proportionate for a final decision, and the tool should let you match one to the other rather than impose one length on every role.
6. Can it be gamed?
Any assessment with a right answer can be prepared for, and any assessment with a fixed list of options can be reasoned through without the ability being tested. The question has sharpened since candidates gained access to tools that write fluent, plausible responses on demand, which is the problem set out in AI-written CVs have broken the CV. Ask the vendor two things: whether the format has a memorisable right answer, and what the tool does to detect responses that are generic or appear not to be the candidate's own work. A tool with no answer to the second question has not thought about the world its candidates live in.
7. Where does the data go, and who decides?
Candidate responses are personal data, and in many tools they are processed automatically to produce a recommendation. Ask what is stored, for how long, where, and who can see it. Ask whether the vendor can support a data protection impact assessment, because for automated assessment of applicants you will usually need one. And ask who decides. Since February 2026, UK data protection law allows a recruitment decision to be made solely by automated means for most data, provided the candidate is told, can make representations, can obtain meaningful human intervention and can contest the outcome. A person with real discretion reviewing the result keeps the decision outside those rules altogether, which is the simpler position and the one to want: the tool as an input, a person making the call, stated plainly by the vendor and reflected in how the product is designed. A tool that produces a hiring verdict rather than evidence for a human to weigh is a harder thing to defend and, for most employers, a harder thing to live with.
8. How does it fit the process you already run?
Who sends the invitation, who reads the report, and how long does the report take to arrive? A tool that needs a trained assessor and returns results in a week sits differently in a process from one that returns a report within minutes of the candidate finishing. Ask whether it connects to the systems you already use or lives alongside them. And ask how it is priced: per assessment, on a monthly plan, or on an annual contract with a minimum. Per-assessment pricing lets you pilot on one role without committing; a subscription makes sense once you know the volume. Be wary of any contract that asks for a year before you have seen the tool on a live vacancy.
Red flags that end the conversation
A single overall score with no evidence attached. Scoring that includes culture fit or personality type. An accuracy claim with no method behind it. One assessment used for every role regardless of level. A recommendation to reject that the tool makes on your behalf. Any reluctance to let you complete the assessment yourself before buying. None of these is necessarily disqualifying on its own, but two together usually mean the product was designed to look rigorous rather than to be checked.
How to pilot one
Pick one live vacancy you are hiring for anyway, ideally one where a poor hire would be costly enough that people will pay attention. Run the assessment alongside the process you already use rather than instead of it, so every candidate meets both. Do not let the interviewers see the report until after they have scored the interview; then compare what the assessment surfaced against what the interview found, and note where they disagreed and who turned out to be right. At the end, look at progression rates across the pilot cohort for any group difference, and look at how many invited candidates did not complete. A pilot that runs for one hiring round on one role will tell you most of what a year's contract would, at a fraction of the cost, and it turns the vendor's claims into your own evidence.
One caution. A pilot cannot tell you whether the people you hired went on to succeed, because that takes months to know. What it can tell you is whether the tool produced evidence you could act on, whether the evidence was fair, and whether the candidates tolerated it. Those are the things to decide on. The performance data comes later, and it is worth setting up now the record that will let you look at it when it does: the real cost of a bad hire is the figure that pilot will eventually be measured against.
Where PRODICTA sits against these questions
PRODICTA builds each assessment from the role details you provide, so the scenarios and scoring criteria are specific to the vacancy rather than drawn from a library. Candidates respond to realistic work situations with no single right answer in their own words; the format and the research behind it are set out on the science page. Every strength and watch-out in the report is anchored to a direct quote from the candidate, and the report includes a confidence level, interview questions and a plan for the first ninety days. Scoring is behaviour-only. Three assessment depths let you match length to stakes: fifteen minutes and two scenarios for a standard hire, twenty-five minutes and three scenarios for a difficult final decision, and an extended assessment for a senior appointment. Sift, a short screening scenario for whole applicant pools, is available on monthly plans and gives an early signal rather than a verdict. Reports are one input to your decision, not the decision itself. Pricing is per assessment or on a monthly plan, set out on the pricing page, and the demo lets you walk a real role through to a full report before you speak to anyone.