The model is 
the interview. 

Aptence runs realistic Excel assessments in the browser, grades every formula against hidden scenarios and records what a proctor would. Built for investment banks, private equity funds and finance programs.

DCF valuationProctored42:17
HomeInsertFormulasSubmit
B17fx9.0%
ABCD
1USD m2025A2026E2027E
2Period012
3Revenue growth 8.0%7.5%
4Revenue1,240.01,339.21,439.6
5EBITDA margin21.5%22.0%22.5%
6EBITDA266.6294.6323.9
7D&A(49.6)(53.6)(57.6)
8EBIT217.0241.1266.3
9Taxes on EBIT(54.3)(60.3)(66.6)
10Capex(62.0)(67.0)(72.0)
11Change in NWC (9.9)(10.0)
12Unlevered FCF   
13Discount factor   
14PV of FCF   
15   
16Sum of PV   
17WACC9.0%  
  • DCF valuation
  • LBO returns
  • Three-statement model
  • LBO with a debt stack
  • Income statement forecast
  • Balance sheet roll-forward
  • Indirect cash flow statement
  • Debt repayment and liquidity
  • PP&E and asset disposals
  • FCFE equity valuation
  • WACC and financing weights
  • DCF sensitivity to WACC and growth
  • Interest accrual and cash payments
  • Three statements with a revolver
  • Regional tax-loss carryforwards
  • Insurance equity multiples
  • Working capital and cash conversion
  • Definite-life amortization
  • Shareholders equity roll-forward
  • Share count and EPS dilution
  • Adjusted present value
  • Forecast reasonableness review
  • Operating forecast drivers
  • Price and volume sensitivity
  • Fixed assets and book-tax timing
  • Current and deferred income taxes
  • Acquisition funding and goodwill
  • Intangibles and impairment schedule
  • Inventory reserve and write-offs
  • Analytical R&D capitalization
  • Credit category and borrowing cost
  • Bond repricing and duration
  • Portfolio risk and performance
  • DCF with an equity bridge
  • LBO with an IRR grid
  • Merger model (accretion / dilution)
  • Trading comparables (comps)
  • Precedent transactions
  • Dividend discount model (DDM)
  • Black-Scholes option pricing
  • Sum-of-the-parts and football field

Platform

Everything a technical screen needs, in one browser tab.

Five parts that work as one: the spreadsheet the candidate builds in, the grader that checks it, the integrity layer that watches the session, the workspace where you decide, and the library of models you test on.

fx=C4*(1+$C$3)
C
D
E
F
G
137.0
190.0
243.0
296.0
349.0
174.0
227.0
280.0
333.0
386.0
211.0
264.0
317.0
370.0
423.0
F4CtrlRAlt=Ctrl⇧↓Ctrl[Ctrl1

01

An Excel your candidates already know

Formulas, fill, absolute references, number formats, trace arrows and the shortcuts analysts use without thinking. Given cells are locked; the candidate builds everything else.

0

score

02

Graded on every formula

Each workbook is recalculated on hidden scenarios. A hard-coded number fails the moment the assumptions move.

  • 00:00Full screen entered

03

Integrity signals, with evidence

Full screen, focus, pastes, second screens and webcam snapshots, summarised for a person to judge.

CandidateScoreRank
  • RR. Haddad64#1
  • SS. Okafor78#2
  • AA. Martin92#3
  • LL. Chen71#4

04

A workspace built for recruiters

Create a test, invite by link or e-mail, follow attempts and compare candidates on the same scale.

75 min

DCF valuation

90 min

Leveraged buyout

45 min

Merger model

90 min

Three-statement

05

Real finance, not puzzles

DCF, LBO, M&A, three-statement, comps and options pricing, with figures drawn per candidate.

How it works

From job post to shortlist in four steps.

Step 1

Choose a model

Pick an assessment from the library, set the time limit and the level of supervision, and add the written questions you want answered.

New assessment

Model

DCF valuation ⌄

Time limit

75 min

Closes

Oct 31

Supervision

Full screen and focusWebcam snapshotsScreen share

Written question

Which assumption moves enterprise value the most, and why?

Step 2

Invite candidates

Send one link or personal invitations by e-mail. Every candidate receives a case of their own, with figures nobody else has.

Candidates

Invite
Invite link · /t/7f3a-summer-analystCopy
  • a.martin@example.comSubmitted
  • s.okafor@example.comIn progress
  • l.chen@example.comOpened
  • r.haddad@example.comSent
  • m.rossi@example.comSent

Step 3

They build in the browser

No download, no add-in. The workbook opens in a locked, supervised tab and the clock is kept by the server, not the laptop.

Supervised · full screen38:04
Revenue234.0265.0296.0327.0
EBITDA331.0362.0393.0424.0
Taxes on EBIT428.059.090.0121.0
Capex125.0156.0187.0218.0
Unlevered FCF222.0253.0284.0315.0
PV of FCF319.0350.0381.0412.0

The server keeps the clock and every edit. Leaving the tab is recorded, never punished automatically.

Step 4

Read the report

A score by layer, the cells that failed and why, a replay of the build and the integrity timeline. Shortlist with the evidence in front of you.

A. Martin

DCF valuation · 41 min

92

  • Accuracy92
  • Integrity95
  • Structure100
  • Reasoning80
  • Speed85

Findings

  • !Terminal value discounted from year 4 in 2 of 10 scenarios
  • ✓Every output depends on the assumptions
  • ✓One formula per line across all years

Grading

The browser is never trusted. The model is recalculated.

When a candidate submits, our servers reopen the workbook with the same formula engine they used and run it next to the reference solution on scenarios the candidate never saw. A model that reaches the right number by accident fails as soon as growth turns negative.

Enterprise value on hidden scenarios

10/10 match the reference

7001,0001,300S1S2S3S4S5S6S7S8S9S10
Reference solutionCandidate matchesCandidate is off

Every output follows the assumptions. It matches the reference on all ten scenarios.

score weightsrelative
  • Accuracy45

    Outputs against the reference on every hidden scenario. A line that only inherits an earlier mistake keeps 75% of its credit.

  • Integrity20

    No errors, no typed outputs, no constants buried in formulas, no circular reference unless iterative calculation is on, outputs that depend on the inputs.

  • Structure15

    One consistent formula along each line, the way the reference is built.

  • Speed10

    Time used, measured by the server and weighted by accuracy: a fast wrong model earns nothing.

  • Knowledge20

    When the test has written questions: each answer graded by AI against a reference answer and a fixed rubric, and any grade can be changed by your team. AI never decides whether a number is right.

  • Reasoning10

    Soon: a written justification of the model, read the same way.

Integrity

Signals with evidence. Decisions stay human.

Every attempt comes with a timeline of what happened during the test: when the candidate left the tab, what was pasted, what the webcam saw, and how the model grew minute by minute.

▶Replay · S. Okafor · DCF valuation00:00 / 45:00
0 min153045
  • 00:12Full screen entered
  • 00:30Webcam on · one face in frame
  • 07:42Window left for 4 s
  • 12:06Paste blocked · 212 characters
  • 18:30Webcam snapshot
  • 23:12Developer tools shortcut blocked
  • 31:48Second screen connected
  • 44:00Submitted · 312 edits

A person decides

Signals are evidence for a recruiter. No candidate is ever rejected automatically.

Nothing hidden from candidates

Before the clock starts, candidates see exactly what will be recorded and why.

Supervision you choose

Full screen, webcam, screen share, microphone, identity check: switched on or off for each assessment.

No facial recognition

Photos are kept as evidence for a person to look at. No emotion analysis, no automatic face matching: your team compares the identity photos.

What the integrity layer records

  • Full-screen exits
  • Time away from the tab
  • Pastes from outside, blocked
  • Copying out, blocked
  • ID document and face, before the test
  • Webcam snapshots
  • Faces counted in each photo
  • Virtual cameras
  • Voices nearby, no sound kept
  • Screen capture
  • Room photo
  • Second screen
  • Test open twice
  • Developer tools, print, save
  • Browser automation
  • Network changes
  • Page reloads
  • Silent periods
  • Model built in one burst
  • Answers typed faster than a person

Library

41 models, from DCF to Black-Scholes.

Every model is checked against its reference solution before a candidate sees it, and draws its own figures for each attempt, so an answer key passed around a class is worthless.

  • valuation60 min

    DCF valuation

    Five-year unlevered free cash flows, terminal value, enterprise value, equity value and value per share.

    31 cells to build→
  • valuation75 min

    DCF with an equity bridge

    Normalised terminal year, full EV-to-equity bridge, treasury stock method and a centred WACC × growth grid.

    119 cells to build→
  • deals60 min

    LBO returns

    Entry price and financing, five years of cash flow sweeping the debt, exit value, money multiple and IRR.

    31 cells to build→
  • deals90 min

    LBO with a debt stack

    Sources and uses with a term loan B, senior notes and PIK mezzanine; five-year forecast to cash taxes; debt schedule with mandatory amortisation, revolver and cash sweep; exit, management option pool, sponsor MOIC and IRR.

    84 cells to build→
  • deals90 min

    LBO with an IRR grid

    Term loan B with cash sweep plus bullet notes, tax after D&A and interest, returns attribution and an entry × exit IRR grid.

    175 cells to build→
  • deals45 min

    Merger model (accretion / dilution)

    Purchase price allocation, amortisation in EPS from day one and the highest price that keeps EPS flat.

    61 cells to build→
  • modeling90 min

    Three-statement model

    Five years of integrated income statement, balance sheet and cash flow statement from an opening balance sheet: working capital, PP&E, term debt, dividends and a revolver.

    150 cells to build→
  • modeling90 min

    Three statements with a revolver

    Link the three statements with a revolver that draws below minimum cash, with no circularity.

    305 cells to build→
  • valuation45 min

    Trading comparables (comps)

    Compute peer multiples, apply them and correct the revenue multiple for the target's margin.

    59 cells to build→

Recruiter workspace

Every candidate on the same scale, with the evidence one click away.

Follow an assessment while it runs, rank candidates on one score, and open any attempt to see what they built, what failed and what happened during the test.

Summer Analyst 2027 · DCF Live · closes Oct 31

Invited

0

Completed

0

Median score

0

Flagged for review

0

Score distribution · 31 attempts10 → 100
CandidateScoreTimeSignals
AMAmélie Martin9241 min0
SOSamuel Okafor8152 min2
LCLina Chen7738 min0
RHRami Haddad6470 min1
MRMarco Rossi3822 min5

Amélie Martin

DCF valuation · 41 min

92

  • Accuracy92
  • Integrity95
  • Structure100
  • Reasoning80
  • Speed85

What failed

  • ×Terminal value discounted from year 4 in 2 of 10 scenarios

During the test

  • •Full screen kept for the whole test
  • •8 webcam snapshots, one face in each
Open the workbookReplay

Illustrative data. Click a candidate to open their attempt.

Built for

Teams that hire on technical skill.

01

Investment banking

Screen analyst and associate classes on the DCF, merger and three-statement models they will build in their first month.

  • DCF
  • Merger model
  • Comps

02

Private equity

Replace the take-home LBO graded by a tired associate with one recalculated on hidden scenarios.

  • LBO
  • Debt schedule
  • Returns

03

Corporate finance and FP&A

Test forecasting, working capital and the discipline of a clean model before an offer, not after.

  • Forecast drivers
  • Working capital
  • Three-statement

04

Universities and programs

Run graded modeling exams for a whole cohort, with figures drawn per student and a report per class.

  • Exams
  • Cohorts
  • Practice

Next: consulting, accounting and data analysis, wherever a spreadsheet is the job.

Hire on the model, not the résumé.

Aptence is in early access for banks, funds and finance programs. Take the DCF test yourself, then put it in front of your next class.