What makes a fund quantitative
A quantitative fund uses explicit data and models to drive a substantial portion of security selection, forecasting, portfolio construction, execution, or risk. Humans still decide which questions to study, how data is interpreted, which model is acceptable, and when a relationship may have stopped working.
Research architecture
The process begins with a hypothesis that can be translated into observable variables. Data is acquired with rights and provenance, cleaned, timestamped, and joined. The researcher defines a target, features, benchmark, training period, test period, and evaluation criteria. Results are examined across time, markets, regimes, and realistic costs.
Signal versus strategy
A signal is information that may predict return, risk, flow, liquidity, or another variable. A strategy converts signals into positions using portfolio constraints, risk models, turnover, financing, and execution. The portfolio can fail even when individual signals have predictive power if they overlap, concentrate, or cost too much to trade.
Portfolio construction
Portfolio construction determines weights and constraints. It can target volatility, neutralize factors, limit sectors, control turnover, cap individual positions, and incorporate expected cost. Optimization can formalize tradeoffs but remains dependent on forecast and covariance estimates that are uncertain.
Execution
The model’s expected return must survive bid–ask spread, market impact, fees, borrow, delay, and adverse selection. Execution can be a separate strategy and research area. A backtest that uses closing prices for a signal not known until after the close contains a timestamp error.
Production systems
Live trading requires data validation, order controls, model versioning, code review, monitoring, alerts, recovery, and a clear process for disabling the strategy. Research notebooks don't provide those controls automatically.
Monitoring
Track performance, exposure, turnover, cost, data quality, feature behavior, model drift, and operational incidents. Compare live behavior with research assumptions. A performance decline can come from random variation, market change, crowding, data changes, execution, or a software error.
Governance and model risk
Document model purpose, assumptions, limitations, ownership, validation, approval, and change history. Independent challenge can reduce the risk that the research team becomes attached to a result. The degree of formal governance varies by organization and regulated activity.
Capacity
A strategy has capacity limits because larger orders move prices, consume liquidity, and reveal information. The scalable return is more important than the return of a tiny simulated portfolio.