Methodology

How this is put together

Everything here is either generated by us or sourced from something a firm published about its own process. This page says which is which, so you can weigh it yourself.

Questions are generated, not stored

There is no question bank to work through. Each game and each assessment section is a generator: it defines the shape of a problem - a two-step Markov chain, a fraction to reduce, a three-outcome ranking - and builds a fresh instance every time you start. Sitting the same paper twice gives you two different papers.

That is a deliberate trade against realism. A real screen draws from a fixed bank, so its questions are more polished than ours will ever be. What generation buys is that practice cannot turn into memorisation, which is the failure mode of every PDF of past problems.

Generators are checked two ways. Automated tests run each one thousands of times and assert the properties that have to hold - that no two items in a sitting repeat, that every multiple-choice question has exactly one correct option, that no ranking question has two answers tied for the same place. Beyond that, every generator and every scoring anchor is read by a mathematics PhD before it ships.

Where the formats come from

Each of the 34 mock screens reproduces a format, not content. Item counts, section timings, marking rules and question types are taken from what the firm publishes about its own hiring process, supplemented by what candidates consistently report in public.

The two are labelled differently on purpose. A published figure is stated plainly. A candidate-reported one - a pass mark, a cutoff, a duration a firm has never confirmed - is always marked as candidate-reported, because firms move those and do not announce it. If you see a bare number on this site that reads like fact and is not, that is a bug worth reporting.

Where a firm runs no online assessment at all, we say so rather than inventing one. Several roles screen by interview only, and those sittings model the work instead of anyone’s test, with the firm name off the paper.

How scoring is set

Marking follows the format being modelled. Where the real section penalises a wrong answer, ours does, at the same rate - which is what makes skipping a real decision rather than a free option. Where it does not, ours does not either.

Your rating out of 200 is not a percentage of questions answered. It is built from accuracy by skill, weighted toward the skills the screens actually gate on, so a strong score in one area does not paper over a weak one in another. The breakdown at the end of every sitting shows the components rather than a single number.

When a format changes

Firms rewrite their screens. When we learn a format has moved, the sitting is updated and the rules on its page change with it - the section list, the timings and the marking notes on each assessment page are generated from the same definition the test runs on, so a page cannot describe a format the test no longer uses.

If you have sat a screen recently and it did not match what we describe, that is the most useful thing you can send us. Contact is the fastest route, and corrections go in ahead of new features.

What we do not claim

Outcry Arcade is not affiliated with any firm named here and has no access to any firm’s assessment content. Nothing on this site is a leaked question, a real paper, or a guarantee about any hiring outcome. More on who we are on the about page.