SD-02.4 — Quantitative & Systematic Research

Business Domain: BD-02 Securities Research & Selection (Front office) · Applies: PUB

Purpose

Researches factors and signals and builds the models behind systematic strategies. Where fundamental research (SD-02.2, SD-02.3) studies one issuer at a time in depth, quantitative research studies the cross-section: it hypothesises a factor, tests whether it predicts return, and combines validated factors into an alpha model. For a systematic manager the “selection decision” is an output of that model, not a discrete human judgement — so the discipline of this Service Domain is as much about defending against overfitting as about finding signal.

Service Operations

Research factors and signals — hypothesise an economically plausible factor and construct it from clean, point-in-time data.
Evaluate signals — test predictive strength: the information coefficient and its stability, factor returns, factor decay.
Develop the alpha model — combine validated signals into a multi-factor model.
Backtest — simulate the model on historical data with realistic transaction-cost assumptions.
Validate the model — out-of-sample and walk-forward testing, and the overfitting defences (purged cross-validation, the deflated Sharpe ratio) that gate a signal into production.
Monitor alpha decay — track live performance against the backtest and retire signals as they decay.

Inputs and outputs

Inputs: historical price, fundamental and alternative data; point-in-time databases; compute infrastructure; the investable universe (SD-02.1).
Outputs: validated factors, an alpha model and its validation evidence — consumed by SD-02.5 (the systematic selection signal) and by portfolio construction in BD-05.

Entities

Consumes: Instrument / Asset (E-02), Listed Equity (PB-01), Debt Instrument (PB-02), Price & Market Data (E-08), Asset Class (E-09); the SD-02.1 investable universe.
Owns: none — the factor library and alpha model are analytical artefacts.

Standards

No external standard governs quantitative research. The named methodologies are the discipline: the information coefficient, walk-forward analysis, purged cross-validation, the probability of backtest overfitting, the deflated Sharpe ratio. Model validation is also subject to the firm’s model-governance regime (SD-14.4).

Open extensions

The boundary with SD-09.5 Investment Analytics & Insight — SD-02.4 builds the models that select securities; SD-09.5 produces analytics on what is held.
The relationship to SD-14.4 Model Governance for the independent validation of production alpha models.