Hiring · August 3, 2026 · 10 min read

How to hire a financial analyst: skills and tests

How to hire a financial analyst: the judgement that separates real analysts from template-drivers, the work samples to run, and the AI-era skills to test.

← Part of The five pillars of hiring: what assessments measure

This guide is for the finance leaders, founders, and hiring managers hiring a financial analyst who cannot afford to get it wrong. The stakes are quiet but large: a financial analyst's numbers feed the board pack, the fundraise, the hiring plan, and the decision to open or close a market. A weak one produces a beautiful model built on an assumption nobody stress-tested, and the error surfaces months later as a missed forecast. The failure mode is invisible on a CV — everyone lists the same tools and the same modelling experience. What separates a real financial analyst from a plausible-sounding one is judgement, and judgement is exactly what a résumé and a friendly chat cannot show you. It is harder in 2026, because AI drafts the first-pass model and writes the commentary, so the job has shifted from producing numbers to catching the wrong ones.

What does a financial analyst actually do?

A financial analyst turns raw finance data into the story leadership uses to decide. Day to day that means building and maintaining forecasts and models, running variance analysis when actuals diverge from plan, and writing the commentary that explains why — in language a non-finance executive can act on. Before you assess anyone, be honest about the job at your company, because it flexes wildly.

• Build and maintain forecasting models — revenue, headcount, cash — and keep them honest as the business changes underneath them.
• Run variance analysis: actuals came in off plan, and the analyst finds why, distinguishing a data glitch from a real trend the business needs to act on.
• Write the commentary in the board pack or monthly report — the paragraph that turns a table of numbers into a decision.
• Partner with the business: sit with the sales or operations lead, understand what drives their numbers, and translate that into the model's assumptions.
• Defend the assumptions when a director pushes back — knowing which input actually moves the output and which is noise.
• Pull, clean, and reconcile data from the ledger, the CRM, and a dozen spreadsheets that never quite agree.

Notice how little of that is 'operate the spreadsheet'. The mechanics are table stakes. The role lives in the judgement layer — which is why hiring on tool proficiency is the classic trap.

Do you actually need a financial analyst yet?

Not every company that thinks it needs a financial analyst does, and hiring one too early wastes a good person on work that does not exist yet. The honest test is recurrence. If your finance questions are one-off — a single fundraise model, one annual budget — a fractional analyst, a contractor, or your accountant handling reporting is usually the right call. You need a full-time financial analyst when the questions repeat and someone has to own the answers.

• You need one when there is a monthly forecast the leadership team genuinely relies on, and variance analysis that someone must own week after week.
• You need one when decisions are stacking up — pricing, hiring, market entry — that all need a numbers person in the room, not a report emailed later.
• You probably do not need one yet if your data lives in three unreconciled tools and nobody owns the source of truth; fix the plumbing first, or you will hire an analyst to babble at noise.
• You may not need one if what you actually want is bookkeeping and the monthly close — that is an accountant. See how to hire an accountant for where the line sits, and hire an operations analyst instead if the recurring questions are operational rather than financial.

What separates a strong financial analyst from a plausible-sounding one?

This is the crux of the hire, because the impostor here is specific and common: the model-template driver. They can populate a three-statement model, they know the shortcuts, and they interview beautifully — but they have never defended an assumption in their life. Their models are inherited templates with new numbers dropped in. When a figure looks wrong, they trust the spreadsheet over their own eyes. The strong financial analyst is the opposite: they treat every number as guilty until traced. The skills that actually predict success are quieter than the ones on the CV.

• Assumption judgement — knowing which input is load-bearing and which is decoration. Ask any analyst to name the three assumptions their forecast is most sensitive to; the impostor lists inputs, the real one lists the ones that move the answer.
• Investigative instinct — when actuals miss plan, they chase the variance to its root cause rather than restating it. The interview question that surfaces this is simple: walk me through a variance you investigated and what you changed as a result.
• Numeric scepticism — a strange number triggers a hunt, not a shrug. This is the single trait most correlated with analysts who never let a wrong figure reach the board.
• Business translation — they can explain a trade-off to a non-finance executive without jargon and without dumbing it down.
• Communication under scrutiny — they can defend a number to a sceptical director and change their view when the director is right, without either caving or digging in.
• The AI-era skill: reviewing machine output. AI now builds the first draft of the model and writes the commentary; the analyst's real job is catching the confidently-wrong number before it ships.

That last point is the shift nobody writes about for this role. When AI drafts the model and the narrative, producing a plausible-looking output is no longer the skill — it is the commodity. The scarce, hireable skill is discernment: reading the AI's confident forecast and knowing, from a feel for the business, that the churn assumption is a fantasy. A financial analyst who cannot review AI output is now the risk, not the safeguard. This is the same judgement layer that separates real signal from polish across every role — cognitive rigour and situational judgement over a tidy-looking output.

How do you test those skills?

You test judgement the way you test any skill that hides on a CV: give the candidate the real work and watch how they do it. The most predictive thing you can do is a job-shaped work sample, not a trivia round on formula syntax. The case for this is covered in work-sample tests, and it applies with force here because the whole job is what a candidate does with a spreadsheet, not what they can recite about one.

Run a modelling work sample with a planted problem

Give candidates an anonymised dataset and a short brief — build a simple forecast, or investigate why a month came in off plan. Plant one deliberately odd input: a mis-keyed figure, an assumption that does not survive contact with the business, a total that does not foot. The template-driver builds a clean model around the bad number and moves on. The real financial analyst stops, flags it, and asks the question you were hoping they would ask. Then add the second half: have them write two sentences of commentary a non-finance executive could act on, and defend one assumption when you push back. That single exercise separates the two candidate types faster than a full interview loop.

Test the underlying reasoning, then the tool

Modelling sits on top of numeric reasoning, so confirm that foundation is solid — a numerical reasoning test checks whether a candidate reasons with numbers under pressure, not just whether they memorised a template. Tool fluency matters too, and an Excel test verifies the mechanics are genuinely there rather than merely claimed on the CV. Treat these as fast confirmations, not the whole assessment, so your work sample can focus on judgement. This is the skills-first sequence set out in the skills-based hiring guide: screen on demonstrated ability, not pedigree.

Because a modern financial analyst works with AI at hand, the assessment should let them. Banning AI tests a version of the job that no longer exists. Run the work sample in an environment where AI tools are genuinely available — an AI Sandbox work-sample session — so you can watch not just the model they produce but how they treat the machine's first draft. Do they accept the AI's forecast, or interrogate it? That behaviour is the whole modern skill, and you can only see it by observing the work, not by asking about it.

What does the interview loop look like?

Keep the loop tight and structured: one strong work sample plus one structured interview beats five rounds of gut-feel conversation, and it protects candidate experience for people who have other offers. Structured means the same questions, same order, and the same scoring rubric for every candidate, so you are comparing evidence rather than reacting to who reminded you of yourself. The structured interviews guide covers why this matters for fairness and accuracy, and an interview scorecard template gives you the anchors to score each answer independently instead of forming one blurry overall impression.

Who interviews, and for what

• The hiring manager or finance lead reviews the work sample with the candidate live — walk me through this model, why this assumption, what would you check first — probing the reasoning behind the output.
• A business partner from sales or operations tests translation: can the analyst explain a number to someone non-finance and take a challenge to it without either caving or getting defensive?
• A structured behavioural round surfaces the investigative instinct: the variance they chased, the forecast they got wrong and how they found out, the time they told a director the plan was unrealistic.
• Every interviewer scores against the same anchors before the debrief, and nobody shares scores until all are in — so the loudest voice does not set the room.

The questions that work ask for specifics, not philosophy. 'Walk me through a variance you investigated and what you changed as a result' is the single best question for this role — it is almost impossible to fake, because a real answer has a messy middle and a concrete outcome. Follow up hard on what the candidate personally did, not what the team did. 'Tell me about a forecast you got wrong' surfaces whether they own error or deflect it. 'Which assumption in your last model were you least sure about, and why did you keep it?' separates the analyst who reasons about uncertainty from the one who inherited a template and never looked underneath it.

Compensation, seniority, and the first 90 days

On compensation, hire on demonstrated judgement rather than anchoring on a title, because 'financial analyst' spans an enormous range — from a first analyst supporting a monthly close to a senior partner who models a fundraise and sits in strategy meetings. Because the role blends numeric rigour with business partnering and, increasingly, the discernment to govern AI output, a strong financial analyst sits toward the higher end of a finance band for their seniority. Benchmark against your own market and level the role to the scope of decisions it will influence. Getting the level wrong is its own kind of mis-hire, and it compounds quietly through every decision the numbers touch.

What good looks like at 90 days

• They own the source of truth — the forecast and the key models are theirs, and people trust the numbers because the analyst can always explain where they came from.
• They have already caught something: a variance nobody had explained, an assumption that was quietly wrong, a report that had been misleading leadership for months.
• They partner rather than serve — the business leads pull them into decisions early instead of asking for a report after the fact.
• Their commentary is read. The paragraph they write in the board pack is the one people quote, because it says what the numbers mean, not just what they are.
• They use AI to move faster and visibly check it — you can see the drafts they rejected as clearly as the ones they shipped.

The best financial analysts are not the ones with the cleanest models or the fastest formulas. They are the ones who look at a plausible number the whole room believes, feel that something is off, and chase it until they know why — with AI drafting in one hand and their own scepticism in the other. Hire for that, assess for that, or you will keep mistaking a tidy spreadsheet for a trustworthy one.