Can You Cheat HireVue Games? What the Ladder Actually Measures

HireVue publishes exactly what it measures, and it is not your answers

Most arguments about cheating a test assume the test is made of questions with answers. That assumption is what makes leaked answer lists feel valuable. It does not hold here, and we know it does not hold because HireVue's own scientists published the feature list in a peer-reviewed paper.

Those features go into a regression trained to predict scores on a conventional cognitive ability test, with a penalty term that trades a little accuracy for smaller subgroup differences. There is no per-question mark anywhere in the pipeline. A shortcut needs something to copy, and this model does not read anything a candidate could be handed in advance.

Levels are generated to a difficulty, not pulled from a list

The second reason a leaked list is useless is that there is no list. A HireVue cognitive game does not draw item 14 from a bank of 200. It builds the level you are on out of the difficulty you have climbed to, and HireVue describes the escalation in its own words: level complexity rises through the addition of shapes, colours or distraction effects.

The ladder is bidirectional and it is what makes two sittings incomparable at the item level. Win a level and you move up. Fail one and you move down. So the sequence you see is a function of your own performance minute by minute, and somebody else's sequence describes their run rather than yours.

The advice that is everywhere, and that lowers your score

This is the most useful thing on this page and it is the opposite of what the preparation industry teaches. Search for HireVue game strategy and you will be told to maximise throughput: accuracy per minute, get through as many levels as you can.

HireVue's own data says that is backwards. In the scoring sample reported in its peer-reviewed paper, total levels played correlated negatively with measured cognitive ability on Shapedance, at r = -0.19 against the questionnaire measure and r = -0.63 against the game-based score itself. On Numerosity the correlation was r = 0.08, which is nothing at all.

The mechanism is not mysterious. Losing a level is fast, because you commit to a wrong answer and move on, and it knocks you down a position. Churning through levels is therefore what failing looks like from the outside. The candidates who score well reach a high level and win most of what they play: the published sample averaged about 28 levels in three minutes on Numerosity with 79% won.

So the single highest-value change most people can make is to slow down by a beat and win the level in front of them. That is not a trick. It is just the opposite of the trick everyone is being sold.

The other three products each fail a shortcut differently

Portrait, PortraitXT and Teamchat

These have no right answers at all, so there is nothing to look up. What there is to get wrong is the temptation to answer as the job description rather than as yourself. Two things make that expensive. PortraitXT reverses roughly half its statements, so a constructed persona has to stay consistent across items that are worded in opposite directions, on about six seconds an item. And a profile built to match a job you have guessed at aims you at a role you may not want: HireVue reports these games correlate .52 to .72 with the IPIP Big Five pool, which means they are measuring something stable enough that the mismatch shows up later.

The Virtual Job Tryout

The tryout's own structure defeats the main shortcut, which is skimming. One section gives you customer details and a later section asks you to recall them, with no warning that it is coming. There is no way to prepare for that except to actually read what you are given. It is also a job-specific product, with more than forty role versions, so material written about somebody else's tryout describes a different set of tasks.

CodeVue

The only product in the family where copying is a realistic route, and the only one with a countermeasure aimed at it: HireVue advertises that it automatically detects potentially plagiarised submissions. The language is the employer's choice, not yours, and a technical conversation usually follows. A pasted solution you cannot walk somebody through is a worse position than a partial solution you wrote.

CodeVue is also the one product with an official practice mode, offered straight after the Begin page. Most candidates skip it. Using it is free and tells you how the editor and the test runner behave before anything counts.

What actually moves the result

Four things, one per product family, in the order they pay off.

On the games, practise the ladder rather than the puzzle. The task in front of you is easy at the bottom and hard at the top, and what you are training is the pace at which you can stay accurate while it gets harder. Our HireVue practice suiteruns all eleven games on the same bidirectional ladder, and reports back the two features HireVue's model actually weights: your ceiling and your win ratio.

Watch the instruction video properly. The games have no practice round. The video before each one is the only preparation the real product gives you, and it is replayable from the pause screen. Most early collapses are a misunderstood rule rather than a weak candidate, and HireVue deliberately makes the opening levels easy so you can settle in.

On the personality games, answer quickly and honestly. About six seconds an item is not enough time to maintain a character across fifty statements, and it is not meant to be.

Know which product you have been sent. This is where most preparation is wasted with this vendor. HireVue also runs the recorded video interview, and one invitation frequently contains both: its own example graduate assessment is six interview questions plus three games. If yours includes a recorded interview, practise that separately at our interview practice, and read which HireVue product you have been sent if the email is ambiguous.