Consulting Assessment Practice 2026: What Free Tests Buy

What each publisher gives away

Everything in this table was read on the publisher's own site on 30 September 2026, except the HireVue practice interview, read the day before.

The four SHL tests that state they return a report are the interactive versions of numerical reasoning, deductive reasoning and inductive reasoning, plus mechanical comprehension. Two state the opposite in as many words: the interactive general ability test says "This test does not currently provide a report", and the learning potential questionnaire says "This test does not provide a report". The rest say nothing, so they promise nothing. The catalog is at shl.com/shldirect.

One condition is attached and it is worth respecting. SHL writes that practice results "are only to be used by you to familiarize yourself with our products", and that sharing them with a third party is treated as misuse of its material.

The ceiling the publisher admits to on assessment practice

Read the sentence SHL uses to explain why practice helps: familiarity with the question types and the timing "will help you feel more relaxed in the real assessment situation, giving you a better opportunity to show your true ability."

That is the company selling the test, describing the benefit as access to your existing ability. It is a smaller promise than the market makes, and it is the one that comes with a source. The same page tells candidates not to guess at what observers want, because guessing wrong is a real possibility.

So the honest model of practice on these formats has two parts. A real floor: sitting an unfamiliar interface under a clock for the first time costs you items, and that cost is removable in a couple of sessions. And a low ceiling: once the format is familiar, further repetitions mostly measure you.

Nobody publishes the size of the gain. We looked for a first sitting against second sitting comparison from any of the four publishers and found none. That page is at shl.com/shldirect, checked 30 September 2026.

Why there is no question bank to memorize

This is the part the paid market has no reason to tell you. TestGorilla documents how its items are served, and the design makes a memorized bank worthless.

Its own guide says each multiple choice test holds "a bank of a minimum of 100 questions", and that only 12 to 20 of them are used per sitting. Questions are "randomly delivered on a per-assessment basis", so a candidate taking the same test in two applications faces fresh items. And each question has a limited lifetime: "Once it has been presented a certain number of times we retire it. This reduces the risk of questions leaking online."

The publisher also refuses to show the items to its own paying clients: "It isn't possible to view the questions and answers to the tests." That page is at support.testgorilla.com, read on 30 September 2026 through a text proxy after a direct request returned 403.

Two practical conclusions follow. A site that claims to hold the live items is selling a guess or a leak, and a leak carries a retirement date set by the publisher. The only asset that transfers is the method for a family of items, which is why the work below is organized by family rather than by question.

The same design explains something candidates report as unfair and which the publisher documents as deliberate. Two people sitting the same named test on the same day answer different items, so comparing scores with a friend tells you nothing. What is comparable is the percentile, and that is computed by the platform against candidates who took the test under the standard clock.

A worked checking item

Checking is the family nobody practices and several consulting screens include. SHL describes its checking test as measuring "how well you can review written text for accuracy, at a detailed level". Sova sells the same family as Checking and Accuracy. The item below is ours.

Compare the entered records against the source records. How many entered records contain an error?

Two of the five. Row two reads 1196 at source and 1169 as entered. Row four reads 8241 at source and 8214 as entered. Both errors are transpositions of adjacent digits, the type that leaves the shape of the string unchanged.

The method that works is mechanical. Compare in fixed chunks rather than reading each record as a word: letters first, then the first two digits, then the last two. Transpositions survive a whole word comparison because the shape of the string is unchanged, and they do not survive a chunked one. Speed comes from the routine and not from reading faster.

Two habits raise the score on this family more than practice volume does. Fix the chunk size before the test starts and keep it for every row, because switching pattern midway is what produces a missed row. And resist confirming a match twice: once a row is cleared, move on. Time spent reconfirming a row you already cleared buys nothing, and the clock is the binding constraint on this family.

A worked calculation item

SHL sells a separate calculation test, described as assessing "your understanding of mathematical equations and your ability to solve problems". Sova says plainly what you need at the desk: "you will also need a pen, paper and calculator for the ability questions."

What is 18% of 2,450?

441. Split it. Ten percent of 2,450 is 245. One percent is 24.5, so eight percent is 24.5 times 8, which is 196. Adding 245 and 196 gives 441.

Now the check that costs three seconds and catches a mistyped digit. Twenty percent of 2,450 is 490, so the answer has to sit a little under 490. A calculator that returns 44.1 or 4,410 fails that test immediately, and a misplaced decimal is the most frequent error on a timed numerical screen.

Build the habit in that order: estimate the band, compute, compare against the band. It costs almost nothing per item and it is the one defense against an input error you cannot see. It also transfers directly into a case interview, where nobody hands you a multiple choice list to check yourself against.

Percentages, fractions and ratios carry most of these items, so the useful drill is small. Learn the one percent value of the number in front of you and build everything from it, as in the answer above. That single move replaces long multiplication on the majority of items and it leaves a trail you can restate out loud, which is what an interviewer asks for later.

The assessment practice questions inside the real test

Two publishers put unscored practice inside the live assessment, and candidates skip past them to save time. That is a bad trade.

Sova says you get "the opportunity to complete a practice question at the beginning of each part of your assessment", and that those answers "do not count towards your score". TestGorilla says the same about the items that open most of its tests, adding that the employer never sees them. SHL tells candidates to read the instructions carefully and to complete any practice and example questions they are given.

So the first live item of each section is free information about the interface: how the answer is entered, whether the question can be revisited, where the timer sits. Use it, then start.

One rule differs by publisher and it decides what you do when stuck. TestGorilla tells candidates: "Skipped questions count as wrong answers. If you don't know the answer, guess." A blank and a wrong answer are the same result there, so never leave one empty on that platform.

Three sessions, and what the fourth is for

1. Session one, no clock. One set per family the assessment names, with the correction read on everything you got wrong. Write the failure mode next to each miss rather than the score.
2. Session two, under the real clock. Compute the pace from the product fact sheet, which is allowed time divided by maximum items, and hold it. This is the session that removes the interface cost.
3. Session three, failing family only. The list from session one will have one recurring mode. Work that one until the correction is automatic.
4. A fourth session only if a family is still losing you items. Past that point the number moves less than the anxiety does, and the time is better spent on case work, which is decided by judgment rather than by pace.

The diagnosis is the whole plan. A candidate who does eight sets across four families and never labels the misses has practiced volume and learned nothing transferable.

What assessment practice cannot move

The comparison group. SHL reports a percentile against a named population printed on the report itself, so your result is relative to a group you did not choose.

The adaptive path. Most SHL Verify products carry the word "Adaptive" on their fact sheet, which means the next item depends on your last answer. Preparation changes your answers, and it does not change the mechanism.

The employer's threshold. No publisher and no firm publishes it. The publisher hands over a percentile and a band, and the decision on what continues belongs to the hiring firm.

What a video section rewards. Sova states that "Time is not recorded on the video recording sections, so this will not affect your results", so rehearsing speed there is wasted effort. HireVue's public practice interview is free, repeatable and run against what it calls a fictitious company, and it says your recording "will not be reviewed by an employer" and is deleted automatically after 30 days.

The firm-built screens no publisher sells

One class of consulting screen sits outside everything above, and a reader looking for practice on it needs to know that before spending an evening. BCG's own US Associate page names its first screen and times it: "The Consulting Career Assessment (CCA) will help us learn more about you, above and beyond what's on your resume and application", read on 6 October 2026. Three clocks sit around it. "CCA invitations will be sent 30-90 minutes after your application is submitted", then "You will have 48 hours to complete the CCA", and the assessment itself "takes approximately 30-35 minutes". A smaller group then sits an online case of about 35 minutes, invited roughly two weeks after the application deadline, and the page adds that "candidates that have previously taken the online case will not be prompted to take it again".

What that page does not carry is a practice version, a sample item, or the name of an outside publisher. Asked how to prepare, it points to its own case interview resources and nothing else. So a search for CCA practice has no official object: the firm publishes none, and the four publishers above do not sell it, because this screen was built by the firm rather than bought in.

Two things still transfer to it. The interface cost described above is removable on any timed screen, whoever wrote the items. And the order of magnitude check in the calculation item is the same defense against a mistyped digit on any platform. What does not transfer is a bank of questions. The published stages, the timings and what the firm says the assessment adds to a decision are collected on the BCG case interview guide.

Where our tools help, and where they do not

We have no publisher style aptitude drill, and the last version of this page implied we had six simulators running. That was wrong, and it is corrected below.

What exists today: free drills that score case work and return a dimension by dimension breakdown, and how an answer is assessed, which publishes the weights they use, read from the code that does the scoring. On the math rubric, Answer 40 carries the result and Working 25 carries the trail, which is our own scoring and not any firm's. Our assessment simulators cover firm built formats rather than publisher tests, and that page states that none of the six is live yet.

So for publisher tests, the publishers are the better starting point, and they are free. For the formats they do not cover, aptitude tests in consulting has the product level timing table, verbal reasoning works the cannot say rule, numerical reasoning works exhibit questions, and the assessment center written case covers the day that follows the screen. For the publisher whose only rehearsal is an unlimited practice question inside its own video interview, the Outmatch assessment collects what it documents and what it withholds.

What we corrected here, and when

30 September 2026. The previous version of this page claimed we had rebuilt six simulators carrying the timing and constraints of the formats they reproduce. No simulator is live today, which our own assessments page states in its own header, so the claim was false and it has been removed. It was the kind of error that matters more than a wrong date, because it was a commercial claim about our own product, and a reader could check it in one click.

Two other things changed with it. The page had never cited a publisher, and it now rests on the candidate material of all four, each with the date it was read and the method used where a direct request was blocked. And the old version advised rehearsing a simulator without saying what the publishers give away for free, which meant the most useful answer to the query was missing. The inventory table at the top exists because of that gap.

6 October 2026. The section on firm-built screens was added. Search Console showed this page being served on bcg cca practice test and bcg cca practice between 19 September and 5 October 2026, and the page named no firm-built screen anywhere, so a reader arriving on either query found nothing here about the test they had typed. The page still covers publisher tests, which is its subject, and it now says plainly that the firm-built screens are not among them and why.

If you find another claim on this site that you can falsify in one click, the fastest correction route is to tell us which page and which sentence.