The Modelling Test in Transaction Services Interviews: What to Expect

Twenty minutes into the case, a candidate is building a beautiful waterfall chart. The bridge behind it doesn't foot. He never noticed, because he never built a check row and never glanced at the total. He'll fail — not for lack of Excel ability, but for the one thing the test is really measuring: can you produce a number a reviewer can trust? The modelling test is where TS interviews get real. The behavioural rounds tell the firm whether they want to work with you; the Excel case tells them whether you can actually do the work.

For many candidates it's the most intimidating stage — but it's also the most learnable, because the formats are predictable and the things being assessed are narrower than people fear. You're not being asked to build a leveraged buyout model from scratch in twenty minutes. You're being asked to show you can handle financial data cleanly, structure an analysis, and reason commercially about what the numbers mean. This article sets out what the test looks like, what graders actually reward, the traps that sink candidates, and a worked walk-through of how to approach a typical case.

The common formats

TS modelling tests come in a few recognisable shapes. You may get one, or a combination, depending on the firm and the level.

• The Excel case study. You receive a trial balance, a P&L, or a set of monthly management accounts and a short brief. You build an analysis — often a QoE-style earnings normalisation — and answer questions.
• The build-from-scratch. You construct a specific schedule from underlying data: an EBITDA bridge, a net debt schedule, or a normalised working capital calculation.
• The error-spotting exercise. You're handed a pre-built model with planted mistakes — broken links, wrong signs, hard-codes — and asked to find and fix them. This tests review skills, which are central to FDD, where you spend as much time reviewing management's numbers as building your own.

Most tests run 60–90 minutes, sometimes with a follow-up discussion where you explain your output. That discussion matters as much as the spreadsheet — graders want to hear your reasoning, and a candidate who can narrate why they made each choice outscores one who silently hands over a correct file.

The format also varies by firm type. A Big Four assessment may be a structured, standardised case; a boutique may hand you messier real-deal data and watch how you cope. Either way, the underlying competencies are the same.

What is actually being assessed

This is the part candidates misread. They try to dazzle with complexity. Graders are looking for something far more grounded — the hard skills a TS team relies on every day.

Notice what's not on the list: flashy formatting, exotic functions, or a dashboard.

No one is impressed by a nested array formula they cannot audit. A clean SUMIFS the reviewer can follow beats a clever one they cannot.

The TS world runs on models that get reviewed, handed over and built upon. Legibility isn't a nicety — it's a professional skill. The partner who signs the report has to trust your workings, and a manager under deadline has to be able to update them at 11pm without a phone call to you.

Common traps

Candidates lose marks in predictable ways. Watch for these.

• Diving in before reading the brief. You build the wrong thing beautifully. Spend the first five minutes understanding exactly what's asked.
• Hard-coding numbers into formulas. A typed constant inside =A1*1.15 is invisible and unauditable. Every input belongs in its own labelled cell.
• Ignoring sign conventions. Mixing positive and negative for costs, or for debt, produces a bridge that doesn't foot and a net debt figure that's simply wrong.
• No sense-check. Handing in a model with a 200% margin because you never glanced at the output is an instant fail.
• Treating every "one-off" as genuine. Cases routinely plant an add-back that recurs, testing whether you probe EBITDA adjustments or accept them at face value.
• Double-counting between bridges. An item pulled into the net debt schedule and the working capital normalisation flatters the equity bridge. Reviewers look for this specifically.
• Running out of time on cosmetics. Formatting last, numbers first.

Modelling hygiene graders reward

Good hygiene is the cheapest way to score well, because it's entirely within your control — it needs no cleverness, only discipline.

• Colour-code inputs and formulas. Blue for inputs, black for calculations is the standard convention. It tells the reviewer instantly what's hard data and what's derived.
• One calculation per cell, building left to right, top to bottom. Avoid mega-formulas that fuse five steps into one unauditable cell.
• Never hard-code inside a formula. If you need an assumption, put it in a labelled cell and reference it.
• Label everything — row headers, units (£'000?), and time periods.
• Build a check row. A balance check, or a bridge foot that returns TRUE or zero, catches your own mistakes before the grader does.
• Keep a clear flow: inputs → workings → outputs.

These habits are what separate someone who has modelled professionally from someone who has only done coursework. They also happen to be exactly the standards you'll be held to on your first live engagement, so building them now is not just interview theatre.

A worked walk-through: an EBITDA-to-net-debt case

Suppose you're handed three years of P&L and balance sheet data and asked to (1) calculate adjusted EBITDA, (2) build a net debt schedule, and (3) comment. Here's the approach — and the approach is what you're graded on as much as the answer.

Step 1 — Read and plan (5 mins). Skim the brief. Identify what's asked and what data you have. Sketch the three deliverables mentally before touching a cell. Resist the urge to start typing.

Step 2 — Set up the structure. A tab or block for inputs, one for workings, one for outputs. Lay the three years across columns with consistent period labels. Two minutes here saves twenty later.

Step 3 — Build reported EBITDA. Start from operating profit, add back depreciation and amortisation. Tie it to the source data. Build a check that it agrees to the statutory numbers.

Step 4 — Layer the adjustments. Walk through the notes. A "one-off" restructuring cost that appears in all three years is not one-off — challenge it. A genuinely non-recurring legal settlement is a valid add-back. This is the heart of the EBITDA bridge: each adjustment is a labelled, sourced line, and the bridge must foot from reported to adjusted EBITDA in every year.

Step 5 — Build the net debt schedule. List gross debt items, deduct cash and equivalents, then consider debt-like items — overdue payables, pension deficits, deferred consideration. Mind the sign convention throughout. If the case touches working capital, separate the normalised position cleanly using NWC analysis so you don't double-count items between the bridges.

Step 6 — Sense-check and comment (last 10 mins). Does adjusted EBITDA move sensibly across the years? Does the margin trend have a story? Write two or three crisp observations:

"Adjusted EBITDA grew from £4.1m to £5.3m, but around £0.6m of the improvement reflects the recurring 'restructuring' add-back we've challenged; on an underlying basis, margin was broadly flat at ~14%."

That comment is what turns a spreadsheet into FDD. Notice the discipline across all six steps: plan, structure, build with checks, then interpret. Candidates who reach Step 6 with time to spare consistently outscore those still wrestling with broken links at the buzzer — because the interpretation is the value TS actually sells.

How to practise

You can't cram a modelling test the night before, but a few focused weeks make an enormous difference. The skill is muscle memory, and muscle memory is built by repetition.

1. Drill the core schedules — EBITDA bridge, net debt, normalised working capital — until you can build each from raw data without thinking about mechanics.
2. Practise against the clock. Speed comes from repetition, not panic. Time yourself and shave minutes off each rebuild.
3. Build a review eye. Open a model, plant errors, close it, reopen it cold, and learn to spot broken links and sign flips fast.
4. Always finish with a comment. Train yourself to interpret, not just compute — write the "so what" every single time.

Slot this into your broader interview preparation plan so the modelling drills run alongside your technical and behavioural prep. If you're coming from an audit background, you already have half the raw skill — the shift is from ticking to the accounts to asking what a buyer would pay for them.

In the interview

The modelling test is often followed by a live discussion of your file. Treat it as a chance to show reasoning, not to apologise for what you didn't finish. If asked "walk me through how you approached this," a strong answer sounds like:

"I spent the first five minutes on the brief so I built the right three outputs rather than diving in. I set up separate input, working and output blocks with a check row on each bridge, so the reviewer can audit every number. On EBITDA, I noticed the restructuring cost recurred across all three years, so I flagged it rather than accepting it as a one-off — that's the difference between reported and genuinely underlying earnings. I ran short on the debt-like items, so I've listed the pension deficit as a flag to size with more time rather than guessing at it. My headline comment is that the earnings growth is real but partly cosmetic, driven by that add-back."

That answer shows planning, hygiene, scepticism about add-backs, honesty about what's unfinished, and a commercial conclusion. It would beat a fully finished file delivered in silence.

Common mistakes

• Building before reading. The single most expensive error. Five minutes of planning beats fifty minutes of rework.
• Chasing perfection on one section. A model with all three deliverables at 90% beats one section at 100% and two blank.
• Silent hand-over. If there's a discussion, narrate your logic. Graders reward reasoning they can hear.
• Accepting management's labels. "One-off," "non-recurring" and "exceptional" are claims to test, not facts to accept.
• Skipping the check row. It costs thirty seconds and catches the errors that fail you.
• Cosmetics before substance. A footing bridge in plain black beats a broken one in corporate colours.
• No closing comment. A spreadsheet without an interpretation is data entry, not diligence.

Clean beats clever

The candidates who pass modelling tests are almost never the ones with the fanciest formulas. They're the ones whose files are clean, foot correctly, and arrive with a sentence that tells you what the numbers mean. The test isn't a puzzle to outsmart; it's a rehearsal of the job — producing a number a partner can defend to a client and a buyer can act on. Master the boring disciplines — plan, structure, check, interpret — and the modelling test stops being the scary stage and becomes the one where you pull ahead. Build the habits now, and on the day you'll be the candidate calmly writing a comment while the rest are still hunting for a broken link.

The Transaction Services Interview Programme (€119.99, one-time) includes timed Excel case studies — EBITDA bridges, net debt and working-capital builds — with model answers, a hygiene checklist and an error-spotting drill set. Enrol today.