Acumen Logic®How we work
We would rather show our working.
Most prep sites ask you to take their accuracy on trust. This page is the actual process behind every replica assessment, every drill and every percentile on the platform: where the material comes from, how it is checked, and the rules we hold ourselves to. If anything here does not hold up, email me at nicholas@acumenlogic.co.uk. There is no ticket queue in between.
How a replica is built
Four steps, in the same order every time.
Start from the source standard.
Every replica is built from the publisher's own published material wherever it exists: fact sheets, candidate guides and official practice pages. Publisher-published material outranks everything else. Prep-industry sites make many claims about these tests and some of them are wrong, so a prep-industry claim is never, on its own, the basis for anything in a replica.
Where a publisher publishes nothing on a detail, we make a conservative choice rather than presenting a guess as fact. SHL and Aon publish detailed specifications. Cappfinity does not publish item counts, so our Cappfinity set is built to the format and scoring model Cappfinity does describe, and we say so.
Match the mechanics exactly.
Each replica matches the real test's timings (the overall limit and the pace per question it implies), question formats (item types, answer mechanics, on-screen layout), scoring rules (including negative marking where the publisher uses it) and navigation behaviour (whether you can go back). Where our replica is deliberately stricter than the real test, the assessment page says so.
Prove every question before it ships.
Every question must have exactly one defensible correct answer. If a second answer is arguable, someone gets marked wrong for being right. So where a machine can check a question, one does: a script rebuilds each logic grid and tries every legal filling to prove only one completion works; another recomputes each rotation item to prove only one option is the turn described. If a check fails, the question does not ship, and neither does the site.
Verbal reasoning and situational judgement cannot be settled that way. Those are read and checked by hand. We would rather tell you that than imply every question is machine-proven. The wrong options are never filler: each one is a specific mistake a real candidate makes under time pressure. Answers are scored on our server against a fixed key, so a score cannot be inflated or edited in the browser.
Machine-check every explanation.
Before any deploy, an automated guard recomputes every arithmetic claim in every drill explanation. If a single number is wrong, the release is blocked. The check is a gate, not a report. Wrong teaching is worse than no teaching.
The Benchmark
One attempt, taken cold. That is the point.
20 questions, three sections, 17 minutes. Free, with free results: score, estimated percentile, per-section strengths and weaknesses, and the correct answer for every question. Most tools give unlimited attempts, which is useful for practising but means the score stops being information. The Benchmark works the other way round. Practice belongs after the diagnosis, not inside it.
The estimated percentile
A model is not a measurement, so we label it.
Your percentile is an estimate from a statistical model calibrated to typical performance on the aptitude tests UK employers set — not a live count of other people's scores. We will refine the model with real candidate results once enough have accumulated. It's meaningful from your very first attempt, and we label it as estimated because a model is not a measurement. We will never present a modelled figure as if it were a direct count of people you beat. If a prep tool shows you a percentile without telling you where it comes from, that is worth asking about.
Corrections
When a publisher changes, the replica changes.
Formats are retired, timings revised, new versions rolled out. When a publisher's published specification changes, the affected replica is corrected. If we find a replica deviates from what the publisher has published, correcting it takes priority over building new content.
Your results
Individual results are yours.
Results are kept for as long as your account is open. In aggregate, de-identified results are used to calibrate and refine the percentile model. Full details are in the privacy policy.
Where we stand
Where we stand on test integrity
Acumen Logic exists for learning and preparation: understanding assessment formats and building genuine reasoning skill before the day. It is never for gaining an advantage inside a live assessment. To be specific:
- Never use Acumen Logic during a live assessment.
- Never share live test content with us. We do not accept it, and we do not build from it.
- Always follow your employer's own assessment rules. Where an employer publishes official preparation, use it. It is the best guide to their process, and our employer guides link to it.
Standards
What we don't do
Standards we hold ourselves to:
No published success rates or score-lift claims.
We publish nothing we could not substantiate to you line by line.
No auto-renewing payments.
Pro is one payment for a fixed period. No subscription, nothing renews, no card kept on file.
No selling of data.
Your results exist to teach you, not to be a product. Individual results are yours; in aggregate, de-identified results are used only to calibrate and refine the percentile model. The privacy policy sets that out in binding terms.
Know where you stand
before they decide.
The Benchmark is free. Seventeen minutes, an instant estimated percentile, and the correct answer for every question.
Start Free Benchmark