Applied Skill Tests 2026: Numerical Practice

Applied, found at beapplied.com, is a UK hiring platform built to reduce bias: employers replace CV screening with anonymised, job-related questions that reviewers score blind, and they can add timed skill tests. One of those is a numerical reasoning test. The version observed in autumn 2026 had 18 questions in 30 minutes with a single countdown, interactive charts with hover tooltips, figures you have to scroll to, data sufficiency and error-checking items, plus a separate untimed practice test. This guide explains how Applied works for applicants and how to prepare for its numerical test, with original practice questions.

Key facts

How beapplied (Applied) hiring works

Applied grew out of the Behavioural Insights Team, the behavioural-science unit originally set up by the UK government, and became a separate company in December 2016. Its design draws on research about bias in CV screening. According to Applied and its backers:

• Anonymisation. Reviewers do not see your name, university or other socio-demographic details when they score your answers.
• Work-sample questions instead of CVs. You answer short questions based on real tasks of the job ("How would you approach…?"). These are scored against criteria the employer sets in advance.
• Chunking and multiple reviewers. Reviewers score one question at a time across all applicants rather than one applicant at a time, the order is randomised, and several reviewers score each answer independently, so one person's impression of you does not colour everything.
• Skill tests. Employers can add timed tests, such as the numerical test described below, to measure specific skills.

For you as an applicant this means two things. Your written answers carry much more weight than on a CV-based platform, so treat each one as a mini work sample: answer the question asked, show your reasoning, and stay within the word limit. And the timed test is scored on its own, with no CV to compensate.

What the Applied numerical test looks like

The test ran on app.beapplied.com in a "timed test" area. The header showed a test ID and a line confirming that you had agreed to and signed the test rules before starting. The layout had a fixed-height panel on the left for the question and figures, and options below.

Practice test. A separate practice run offered sample questions with no timer. Its buttons read "Exit practice test" and "Review my answers". It is worth doing the full practice run: the real test uses the same layout and the same item types.

Timer and progress. The real test had one 30-minute countdown, shown as an hourglass between the Previous and Next buttons (on a narrow window it became a floating clock). A progress bar labelled "PROGRESS (TOTAL ANSWERS GIVEN)" counted answered questions. Because Previous and Next work, you can skip a hard item and return.

Hover and scroll. Most chart items carried a hint: "Hover the mouse over the points for precise values". Tooltip values used a decimal comma (for example "27,35"). Items with several figures said: "Please use the 4 figures below to answer the question. You may need to scroll down to see them all." Tables could also scroll sideways, hiding their last column. The grey hover hint appeared even on text-only items, where there was nothing to hover.

Pace. 18 items in 30 minutes is 100 seconds each on average, but the spread is wide. Word problems took under a minute; multi-figure case items (a cost table plus a revenue chart, or two bar charts that must be combined) took two to four minutes each. In the observed sitting almost the full 30 minutes were used: the last item was reached with about a minute and a half left.

Question types with original examples

The observed test mixed seven types: table and chart arithmetic, word problems, multi-step growth projections, multi-figure "case" items, data sufficiency, "which information do you need", and error checking. Some options were written out in words ("three million, four hundred…").

1. Word problem (two unknowns). A monitor and a docking station cost £418 together if bought separately. The monitor costs £146 more than the dock. Bought as a bundle, the dock is 25% off. What is the bundle price? £313.50 / £384.00 / £350.00 / £418.00 / £136.00

Answer: £384.00. Dock = (418 − 146) ÷ 2 = 136; monitor = 282. Bundle = 282 + 136 × 0.75 = 282 + 102 = 384. £313.50 applies the 25% to the whole price.

2. Proportional scaling. Twelve printers print 3,600 pages in 2 hours. How many pages do 30 printers print in 5 hours? 9,000 / 22,500 / 15,000 / 4,500 / 90,000

Answer: 22,500. Rate = 3,600 ÷ (12 × 2) = 150 pages per printer-hour. 150 × 30 × 5 = 22,500. Scaling only the printers or only the hours gives 9,000.

3. Data sufficiency. Four points P, Q, R and S lie on a straight line in that order. How far apart are Q and R? Statement 1: PS = 30 m and PQ = RS. Statement 2: PR = 18 m. Options: Statement 1 alone is sufficient, but statement 2 is not / Statement 2 alone is sufficient, but statement 1 is not / Each statement alone is sufficient / Both statements together are sufficient, but neither alone is / Statements 1 and 2 together are not sufficient.

Answer: Both together are sufficient, but neither alone is. Let PQ = RS = x. Statement 1 gives 2x + QR = 30; Statement 2 gives x + QR = 18. Either alone has two unknowns. Together: subtracting gives x = 12, so QR = 6 m.

4. Information needed. A charity wants to know whether moving its newsletter from print to e-mail will save money over three years. Which information is needed? A) The yearly print and postage cost. B) The yearly cost of the e-mail software. C) The one-off cost of setting up the e-mail system. D) The number of staff in the communications team. E) The printing supplier's founding date. A & B / A, B & C / A, B, C & D / B & C / A & E

Answer: A, B & C. Savings over three years = 3 × (print cost − software cost) − set-up cost. Team size (D) is not part of the comparison unless the stem says staff time changes, and the founding date (E) is irrelevant. Read the time horizon in the stem: a one-off cost matters for "over three years".

5. Error checking (origin of the error). A table shows Budget, Actual, Difference and % Difference (all £000s) for four products and a total:

Product   Budget  Actual  Difference  % Difference
Kettles     420     441        21        5.00%
Toasters    360     342       -18       -5.00%
Blenders    250     265        15        6.00%
Mixers      180     189         9        5.00%
Total     1,210   1,273        63        5.21%

Which cell contains the original incorrect calculation? Budget Total / Actual Total / Difference for Total / % Difference for Total / % Difference for Blenders

Answer: Actual Total. 441 + 342 + 265 + 189 = 1,237, not 1,273 (two digits transposed). The total difference (63) and percentage (5.21%) were calculated from the wrong total, so they are follow-on errors. All row values check out: 21 ÷ 420 = 5%, −18 ÷ 360 = −5%, 15 ÷ 250 = 6%, 9 ÷ 180 = 5%.

6. Projection with a red herring. A line chart shows a museum's ticket revenue in £000s; hovering over 2025 shows "640,0". A note under the chart says the average ticket price is £16. Revenue is expected to rise 5% a year. What is the projected revenue for 2028, to the nearest thousand pounds? £740,880 / £741,000 / £736,000 / £672,000 / £800,000

Answer: £741,000. Three years of growth: 640 × 1.05³ = 640 × 1.157625 = 740.88 (£000s), which rounds to £741,000. The ticket price is not needed. £736,000 adds 3 × 5% = 15% instead of compounding; £740,880 does not round as asked.

7. Time puzzle. Your train leaves at 09:52. Your phone shows 09:41, but you know it runs 6 minutes slow. The walk to the station takes 14 minutes. If you leave now, how do you arrive? 3 minutes early / 9 minutes late / On time / 3 minutes late / 9 minutes early

Answer: 9 minutes late. Real time = 09:41 + 6 = 09:47. Arrival = 09:47 + 14 = 10:01, which is 9 minutes after 09:52. Using the phone's time gives 09:55, 3 minutes late, the trap.

Traps and how to avoid them

• Figures below the fold. Before you calculate anything, scroll the left panel to the bottom and sideways. An item that looks unanswerable usually has another figure further down.
• Tooltips with a decimal comma. "27,35" is 27.35. Convert before you compare with options written with a point.
• Red herrings. Exchange rates when every series is in one currency, a revenue chart in a cost question, a price per ticket when revenue is already given. Ask "which numbers does this question actually need?"
• Compounding periods. Count the years: from 2025 to 2028 is three growth steps, not four.
• Answers in words. Convert "three million and forty thousand" to 3,040,000 before comparing, and watch place value.
• Digit transpositions in error checks. Errors were single swaps (two digits exchanged, a letter changed). Scan systematically, one row at a time.
• Ambiguous items. At least one observed item's options fitted only a slightly different reading of the question (a different year). If nothing fits after a second read, choose the closest option, flag it mentally and move on; the Previous button lets you return if time remains.

How to prepare

One week, 30–45 minutes a day.

1. Day 1: Do the Applied practice test if you have access to it, untimed, to learn the layout, hover and scroll behaviour.
2. Days 2–3: Word problems and proportional scaling: two unknowns, discounts on one item, worker-hour and printer-hour rates. Target under 60 seconds each.
3. Day 4: Data sufficiency. Learn the five fixed options and test each statement alone before combining. Practise "information needed" items by writing the formula first.
4. Day 5: Error checking: column comparisons and "origin of the error" tables. Recalculate totals first; most errors live in a total.
5. Day 6: Multi-figure items: combine a cost table with a chart, and percentage-change charts where you back-calculate earlier months.
6. Day 7: Full mock: 18 mixed items in 30 minutes. Answer quick items first, then return to the long ones with Previous / Next.

Setup. A laptop or desktop with a full-size browser window (narrow windows rearrange the layout and push the question below the charts), a mouse for hovering, a calculator, pen and paper.

Written questions. If the application also has written work-sample questions, draft answers offline, keep to the word limit and answer exactly what is asked: reviewers score each answer on its own, without seeing your name or CV.

Sources

• Applied job board
• Nesta, "Goodbye John and Jennifer, hello Applied" (name, university and socio-demographic blinding)
• Nesta press release, investment in Applied (spin-out from the Behavioural Insights Team, December 2016)
• Blackbird Ventures, Applied portfolio note (anonymised, randomised, reviewed by several team members)
• The Psychologist (BPS), "Five minutes with David Halpern" (Applied prototype vs CV sift, chunking)
• Association for Psychological Science, "Harnessing the Wisdom of Crowds to Improve Hiring"
• City AM, anonymous skills-based tests and hiring outcomes (Applied-commissioned research)

Geschrieben von

Adrian, Spezialist

MSc Psychologie, 7 Jahre Testablegen.

Zuletzt aktualisiert: 01.10.2026