Front Office Investing interview prep.

Thinks in hypotheses, information coefficients, and out-of-sample evidence, allergic to a good backtest with a bad story.

What interviewers look for

  • Does the candidate distinguish in-sample from out-of-sample? A great backtest is the start of the investigation, not the conclusion, the question is whether it survives OOS.
  • Do they have a hypothesis BEFORE the data? An economic or behavioural reason a signal should work guards against data-mining; signals with no story are usually overfit.
  • Are they rigorous about bias, look-ahead, survivorship, data snooping, multiple testing, and can they name how each inflates a backtest?
  • Do they reason in net-of-cost, real-world terms, turnover, transaction costs, market impact, capacity, alpha decay, not just gross Sharpe?
  • Can they reason about signal combination and portfolio construction. IC, breadth, correlation between signals, risk model, optimization, not just one factor?
  • Intellectual honesty: are they more interested in why a signal might be WRONG than in selling it? Quant teams screen hard for researchers who try to kill their own ideas.

Behavioural questions to expect

  1. Walk me through your CV.

    What it tests: Story coherence + genuine fit for systematic research. Funds want evidence of quantitative rigor and a builder's instinct (research + code), not a narrative stock-picker drawn to 'markets'.

  2. Tell me about your most rigorous piece of research.

    What it tests: Research maturity + intellectual honesty. Tests whether the candidate validated a result (out-of-sample, robustness) rather than reporting an in-sample finding.

  3. Tell me about a weakness, a failure, or feedback you've received and worked on.

    What it tests: Self-awareness + research humility. Cross-role canonical. Fake weaknesses downgrade immediately. Quant research is mostly failed ideas, teams want people who kill their own signals.

  4. Why systematic / quant investing? Why not discretionary?

    What it tests: Authentic interest in the systematic approach vs cycling buy-side recruiting. Tests whether the candidate is drawn to evidence, breadth, and process, not a gut-feel narrative.

  5. Why this firm?

    What it tests: Whether the candidate has done the homework. Bar: firm-specific evidence from style, data / tooling edge, and research culture, not generic 'great track record'.

  6. Why a {multi-manager pod / single-manager} platform, or why {mid-frequency / high-frequency}, over the alternative?

    What it tests: Whether the candidate understands the structural trade-offs and has made an informed choice, not just chasing the first offer.

  7. How would you describe this firm's research process and edge in your own words?

    What it tests: Whether the candidate has internalized HOW the firm makes money, data, research, execution, not just WHAT it trades. Tests whether they understand the research-to-production pipeline, not a buzzword summary.

  8. How do you think this firm manages risk on its signals and at the book level?

    What it tests: Whether the candidate understands systematic risk is about factor exposure, crowding, and capacity, and that a decaying or crowded signal is a risk even if the backtest looks fine. Probed hardest at pod platforms.

Technical concepts to master

Designing a signal, the workflow

Hypothesis first
An economic or behavioural reason the signal should predict returns, stated before touching the data.
Data + point-in-time
The dataset and the discipline of using only what was knowable at decision time (point-in-time, properly lagged).
Construction
Turning raw data into a signal, cross-sectional ranking or z-scoring, winsorizing outliers, neutralizing obvious factors.
Evaluation
Measure predictive power. IC, quantile spread returns, Sharpe, t-stat, in-sample then out-of-sample.

Backtesting hygiene

In-sample vs out-of-sample
In-sample is where you built the signal; out-of-sample is untouched data used only to confirm it.
Walk-forward + cross-validation
Re-fit and test on rolling forward windows; for time series, purge and embargo around test sets to avoid leakage.
Multiple testing + deflated Sharpe
Trying many signals on one dataset guarantees lucky winners; the deflated Sharpe penalizes for the number of trials.
Survivorship + look-ahead
Survivorship drops dead names; look-ahead uses future information, both systematically inflate backtests.

The economics of a signal

Turnover
How much of the book trades per period; high-frequency, fast-decaying signals turn over more.
Transaction costs + market impact
Commissions, spread, and the price you move by trading; impact rises with size and falls with liquidity.
Capacity
How much capital the signal carries before market impact erodes the alpha toward zero.
Alpha decay / half-life
How fast a signal's predictive power fades after the data point, its half-life.

The fundamental law + combining signals

Grinold's fundamental law
Information ratio is approximately IC times the square root of breadth: IR = IC x sqrt(breadth).
Breadth + independence
Breadth only counts if bets are independent; correlated bets reduce effective breadth.
Signal combination
Weight signals by IC and, crucially, by their correlation; uncorrelated signals diversify the alpha.
Risk model + neutralization
Strip unwanted exposures (sector, size, beta) so the bet is on the signal, not on a factor in disguise.

Why signals fail live

Overfitting (the big one)
The signal fit noise, not structure; it shines in-sample and dies out-of-sample.
Costs + capacity reality
A gross edge that doesn't survive transaction costs, market impact, or that has trivial capacity.
Regime change
The relationship the signal exploited shifts when the macro or market regime changes.
Crowding + decay
Others find the same signal; the edge compresses, decays, and can unwind sharply in a deleveraging.

Practical drills

  • Design an alpha signal for the firm's asset class and horizon. 5 min prep, 5-7 min delivery. Be ready to be probed for 10-15 min on bias controls, net-of-cost economics, and what would kill it.
  • Your signal has an information coefficient of 0.05 and you make ~200 independent bets per year. (a) What's the implied information ratio? (b) What if you raise breadth to 800 independent bets? (c) You add a second, uncorrelated signal with the same IC, roughly what happens to the combined IR?
  • A signal shows 8% gross annual alpha. It turns the book over 20x per year (round-trip), and round-trip transaction cost is 20bp. (a) What's the net alpha? (b) The firm wants to 5x the capital and market impact doubles the cost to 40bp, now what? (c) What does this tell you about capacity?

Smart-question anchors

  • Research-to-production pipeline, how a signal goes from idea to live capital, and who owns data, execution, and risk
  • Data + tooling edge, proprietary / alternative data, the backtesting framework, and how research is validated before deployment
  • Signal ownership + collaboration, whether researchers own signals end-to-end or feed a central book; how credit and P&L are attributed
  • Capacity + cost discipline, how the firm models transaction costs / market impact and thinks about capacity vs assets
  • Risk + monitoring, factor neutralization, crowding monitoring, live-IC-vs-backtest tracking, and kill criteria for decaying signals

Sourced from

Ready to Generate Your Own Prep?

Drop your CV and a job description on the home page. A couple of minutes later you get a report with everything you need to land the job.

Generate my first prep