Financial Modeling for Beginners: The Skill That Opens Doors Across Industries

Financial modelling is the most portable quantitative skill in business, and the one most consistently faked. A candidate who can build a three-statement model that ties, defend a discount rate, and state the two assumptions that decide the answer is employable in M&A, private equity, corporate finance, infrastructure, energy trading and strategy. This is a European build guide: IFRS-anchored, euro-denominated, and structured as a ten-week programme with an honest account of what the skill is worth in France.

1. What financial modelling actually is

A financial model is an auditable argument about the future of a business, expressed as arithmetic. It has three obligations, in order: it must tie (the balance sheet balances, cash flows reconcile), it must be traceable (every input has one source cell and a documented origin), and it must be decision-shaped (it answers a specific question: pay this price, build this plant, refinance now).

Four model families cover almost all professional use:

• Three-statement operating model — P&L, balance sheet, cash flow driven by revenue and cost drivers. The foundation; everything else is built on it.
• DCF valuation — unlevered free cash flow discounted at WACC, with an explicit terminal value.
• LBO — debt schedule, cash sweep, covenant headroom, IRR and MOIC to the sponsor.
• Project / infrastructure finance — the fastest-growing family in Europe, driven by grid, renewables, nuclear and industrial-capacity mandates: long-dated cash flows, debt sizing on DSCR, and contracted revenue.

2. The European accounting reality

Most training material online is US GAAP. In Europe you will work in two frames simultaneously and confusing them is a visible error.

• IFRS applies to consolidated accounts of EU-listed groups. Consequences for your model: IFRS 16 puts operating leases on the balance sheet as a right-of-use asset and a lease liability — which inflates EBITDA and changes every leverage multiple you quote. IFRS 15 governs revenue recognition timing. Development costs can be capitalised under IAS 38 conditions, unlike US GAAP practice.
• French statutory accounts (Plan comptable général) are what an unlisted French mid-cap actually files, retrievable from the registry. Layout, provisions and the treatment of crédit-bail differ from IFRS. If your comparable set mixes statutory and IFRS figures without adjustment, your multiples are wrong.
• Tax and social charges. The French corporate income tax rate of 25% (Legifrance, CGI art. 219, 2024), the CVAE trajectory, CIR/CII research credits, and employer social contributions of roughly 25-42% of gross payroll (URSSAF contribution schedules, 2024) all belong in the model explicitly. Employer charges are the single most common omission in student models of French businesses.

Building a euro discount rate

Do not import a US WACC. Build it: risk-free rate from the 10-year French OAT or German Bund, an equity risk premium for the relevant European market, a beta from listed comparables re-levered to your target capital structure, and a cost of debt taken from the company's actual facilities or comparable euro-market spreads. Then say out loud what a 100bp move in the discount rate does to your valuation. If you cannot, you do not yet own the model.

3. Structural discipline — the part that separates professionals

Errors in professional models are almost never mathematical; they are structural. Six rules eliminate most of them.

1. Separate inputs, calculations and outputs into distinct sheets, with one consistent colour convention for hardcoded inputs. No hardcode anywhere in a calculation block.
2. One row, one formula. A formula that changes mid-row is the most common source of silent error in audited models.
3. Build checks that fail loudly. Balance-sheet check, cash-flow reconciliation, sum-of-segments to total, and a single master flag at the top of every sheet.
4. Time as columns, never as sheets. Include a hardcoded model-date row and drive every period from it.
5. Sensitivity is not decoration. Two-way data tables on the two variables that actually move the answer, plus a downside case with the covenant consequence stated.
6. Document the sources. One assumptions log listing each input, its value, its source and its date. This is what makes a model defensible in a diligence room.

4. The toolchain in 2026

Excel remains the deliverable format; it is no longer the whole workflow. What is now expected at entry level in European finance teams:

• Excel to a professional standard — INDEX/XLOOKUP, SUMIFS, iterative calculation for circular interest, data tables, Power Query for ingesting messy registry and ERP extracts. Keyboard-only navigation, because speed is assessed in modelling tests.
• Python or SQL for data preparation — pulling and cleaning multi-year filings, market data, or operational extracts before they reach the model.
• Generative AI as an accelerator, never as an author. It is genuinely useful for boilerplate structure, formula debugging and documentation drafts. It hallucinates figures. Any number that enters your model from a model, not a source, is a liability with your name on it.

5. A ten-week build programme

Sequenced so each week produces an artefact you can show, not a tutorial you watched.

• Weeks 1-2 — Accounting fluency. Read one full IFRS annual report of a CAC 40 industrial group and one French statutory filing of an unlisted mid-cap. Reconcile net income to operating cash flow by hand.
• Weeks 3-4 — Three-statement model from filings. Rebuild three historical years of the group you read, then forecast five, driver by driver. It must tie without a plug.
• Week 5 — DCF. Build the euro WACC from OAT, re-levered beta and real facility pricing. Add a sensitivity grid on WACC and terminal growth.
• Week 6 — Comparables. Six European peers, adjusted to a consistent accounting basis, with IFRS 16 treatment made explicit.
• Weeks 7-8 — LBO. Debt schedule with cash sweep, covenant tests, IRR and MOIC decomposition into deleveraging, multiple expansion and operating improvement.
• Week 9 — Project finance model. A 20-year renewable or grid asset: DSCR-sized debt, contracted and merchant revenue split, and the equity IRR under two price cases.
• Week 10 — Ship the portfolio. Publish three models with a one-page investment memo each. The memo is what gets read; the model is what gets audited.

Credentials help at the margin: the CFA programme for investment roles, the AMF certification for regulated market functions in France, and a recognised modelling certificate for signalling. None substitutes for three defensible models and a memo you can argue.

6. What the skill is worth in France

Indicative gross annual first-job packages, Paris, 2025-2026: M&A or transaction services analyst €50k-€70k plus bonus; private equity analyst €60k-€85k plus carry exposure; infrastructure and energy project finance €45k-€60k; corporate FP&A or corporate finance analyst €38k-€48k; industrial controlling €36k-€45k. The differential is driven less by the modelling itself than by the deal exposure attached to it — which is why the transferable move is to pair modelling with a sector you can speak to credibly: energy, defence, healthcare, semiconductors.

7. Worked mini-case: sizing debt on a €120m grid asset

Abstract instruction produces abstract skill. Work the sequence below on paper before you build it, because it is the exact logic a project-finance interviewer will test.

1. Establish contracted revenue. A €120m capex asset earning an availability-based tariff produces, say, €14m of contracted revenue a year, indexed to inflation. Separate contracted from merchant revenue in different rows — lenders will only size debt on the contracted part.
2. Derive operating cash flow. Deduct operations and maintenance, insurance, land or connection fees, and local taxes. Assume €3.5m, leaving €10.5m of cash available for debt service.
3. Apply the coverage constraint. Lenders require a minimum debt service coverage ratio — typically 1.30x to 1.45x for contracted European infrastructure. At 1.35x, annual debt service cannot exceed €7.8m.
4. Convert to a debt quantum. Annuity-amortise €7.8m over an 18-year tenor at an illustrative 5% all-in euro cost: roughly €91m of senior debt, implying about €29m of equity, or 24% of capex.
5. Test the downside. Cut availability by 5% and add 50bp to the cost of debt. If DSCR breaches 1.10x, the structure is not financeable and the sponsor must inject more equity or renegotiate the tariff.

The output of that exercise is not a number but a sentence: in this worked example the deal works at 1.35x coverage with 24% equity, and breaks if availability falls below 92%. That sentence is what a credit committee buys.

8. Five errors that end candidacies in modelling tests

1. Hardcoded numbers inside formulas. The single fastest way to fail a technical screen, because it proves the model cannot be updated or audited.
2. Circular references resolved by deleting the interest link. Interest on average debt is genuinely circular. Handle it with iterative calculation or an explicit circuit-breaker switch, and say which you used.
3. EBITDA quoted without stating the lease treatment. Post-IFRS 16 EBITDA and pre-IFRS 16 EBITDA are different numbers. Comparing a multiple across the two is a credibility-ending error in a European interview.
4. Terminal value carrying, for example, 85% of the valuation, unremarked. If it does, say so and defend the terminal growth rate against long-run euro-area inflation and GDP expectations. Silence reads as not having checked.
5. No sensitivity, or sensitivity on variables that do not matter. Run the two drivers that actually move the answer. Sensitising, for example, a 0.3%-of-revenue cost line signals you have not identified what the model turns on.

9. How the skill is actually assessed

European recruiters test modelling in three formats, and each rewards different preparation. A timed build (60-180 minutes, from a blank sheet or a template) tests structural discipline and keyboard speed. A model review hands you a broken file and asks what is wrong — the fastest way for a firm to detect real experience, because only someone who has inherited bad models finds the errors quickly. A case defence asks you to justify your own model: which assumption is weakest, what would change the recommendation, and what you would verify with one more day of data.

Prepare for all three by building models you intend to defend rather than complete. Before publishing any model in your portfolio, write the three sentences an interviewer will ask for: what the model concludes, which two assumptions drive it, and what evidence would overturn it. Candidates who can say those sentences without reopening the file are the ones who get offers.

10. Where to get real European data

Models fail on inputs, not formulas. These are the primary European sources worth knowing before you pay for anything:

• Company accounts. Listed groups publish IFRS annual and half-year reports with segment detail. Unlisted French companies file statutory accounts to the commercial court registry, accessible through the national open-data company registry.
• Regulated market disclosure. The AMF publishes prospectuses, transaction filings and shareholding declarations — the fastest route to real deal terms and capital structures.
• Macro and sector series. Eurostat and INSEE for prices, wages, industrial production and construction cost indices; the finance acts for tax parameters; the OECD for cross-country comparisons.
• Energy and infrastructure operations. RTE for grid and load data, and national regulator publications for tariffs and network charges — indispensable for any industrial or project model.
• Labour cost. Sector collective agreements set minimum classification salaries; APEC data anchors cadre pay. Model payroll from the agreement upward, not from a guess downward.

Log every input with its source and retrieval date in the assumptions sheet. In a diligence room, an input without a date is treated as an input without a source.

Build order matters more than tool choice: accounting fluency first, structural discipline second, valuation technique third. Candidates who invert that order produce models that look professional and cannot be defended for ten minutes.

Method and limits

Accounting treatments summarise IFRS and French PCG principles at the review date and are not a substitute for the standards themselves or professional advice; IFRS 16, IFRS 15 and IAS 38 applications vary by fact pattern. The corporate income tax rate, CVAE trajectory and employer contribution ranges reflect the French framework at the review date and change with each finance act; employer charge ranges vary widely by salary level and applicable exemptions. Pay bands are indicative gross annual Paris packages triangulated from published salary surveys, APEC cadre data and recruiter ranges; they are ranges, not offers. Discount-rate construction describes standard practice, not a single correct method — equity risk premia are estimates that differ materially across published sources.