The Two Sigma Factor Lens is a return-based factor model powering Venn, Two Sigma’s cloud-based portfolio analytics platform. Unlike holdings-based models like MAC3 or Barra, it requires only a return stream — no position-level transparency — making it applicable to hedge funds, private assets, and alternatives where holdings are unavailable.
The 18 factors
The lens expanded from 8 factors (2018) to 18 (2023), organized hierarchically:
Core Macro (4)
| Factor | Proxy |
|---|---|
| Equity | MSCI ACWI |
| Interest Rates | Global sovereign bonds, 7-10yr maturity |
| Credit | Residualized blend of US/EU IG and HY corporate bonds |
| Commodities | S&P GSCI, residualized against rates and equity |
These four core factors alone explain 80-98% of variance for typical institutional portfolios.
Secondary Macro (4)
| Factor | Proxy |
|---|---|
| Emerging Markets | Equal-risk blend of relative EM credit and equity vs. developed |
| Foreign Currency | GDP-weighted basket of G10 currencies vs. local |
| Local Inflation | Inflation-linked vs. nominal rates (USD/GBP) |
| Local Equity | Domestic vs. foreign equity (home bias) |
Macro Styles (4)
| Factor | Proxy |
|---|---|
| Equity Short Volatility | Rolling short S&P 500 put options |
| Fixed Income Carry | Long high-yielding, short low-yielding 10yr bond futures |
| Foreign Exchange Carry | Long high-yielding, short low-yielding G10 currencies |
| Trend Following | Multi-asset futures based on 6-12 month trailing returns |
Equity Styles (6)
Market-neutral, region-neutral long-short portfolios: Value, Momentum, Quality, Low Risk (related to residual-volatility), Small Cap, and Crowding (short widely-shorted stocks).
Methodology
Hierarchical residualization
Factors are constructed to be orthogonal via a top-down residualization process:
- Equity and Interest Rates are the base factors (directly observable)
- Each subsequent factor is the residual from regressing its proxy on all higher-order factors
- This ensures low cross-correlations even during market stress
Regressions use rolling exponentially-weighted windows (3 years of daily returns) with Newey-West adjustments for lead-lag effects across asset classes.
Portfolio analysis
To analyze a fund, Venn runs return-based regression of the investment’s return stream against the factor set. No holdings data is needed.
Four design principles
- Holistic: high R-squared with few factors
- Parsimonious: avoid spurious exposures from too many regressors
- Orthogonal: low cross-correlations by construction
- Actionable: stable factor-mimicking portfolios that translate to asset allocation decisions
Nonlinear risks are addressed by explicitly including skewed factors (Credit, Equity Short Volatility). Residual return is not automatically “alpha” — it can be uncompensated idiosyncratic risk.
Use cases
- Manager evaluation: decompose returns, detect style drift, compare factor profiles across managers
- Portfolio analytics: identify overlapping factor exposures across sleeves, measure true diversification at the factor level
- Risk decomposition: attribute risk to systematic factors vs. residual; scenario and stress analysis
- Trend analysis: rolling factor exposures over time to track strategy evolution
Manager evaluation with Venn
Two Sigma’s Venn Manager Evaluation Guide describes a structured workflow for using the Factor Lens in manager due diligence.
Qualitative vs. quantitative due diligence
Qualitative evaluation (meetings, surveys, team culture, strategy history) provides context. Quantitative factor analysis connects what a manager says with what they actually do. Factor analysis decomposes risk to show what drove performance, going beyond surface-level “what happened” to the “how.”
Style drift detection
A common use case is checking whether a manager’s factor exposures match their stated mandate. For example, a manager claiming to invest based on value characteristics but showing zero or negative value factor exposure signals style drift. Rolling factor analysis over shorter time frames reveals how exposures changed over time, including around manager transitions or market stress.
Residual return as alpha
The Factor Lens is designed to explain as much risk as possible. The unexplained portion (residual return) may represent genuine manager skill (alpha). Analyzing residual contribution to risk and return — and its trend over time — helps distinguish persistent skill from temporary favorable conditions. Fading residual over time may indicate the investment style has become less effective. Sudden changes in residual may mark manager changes or strategy shifts.
Manager comparison
Factor analysis enables deeper manager comparisons beyond raw risk-return metrics:
- One manager may have achieved returns through favorable factor exposure (e.g., riding a momentum tailwind) while another delivered genuine residual alpha
- Similar-looking returns may mask very different risk profiles or factor bets
- Peer group analysis (e.g., within US Large Blend) provides relative context — 20% returns in a year where peers earned 30% is actually underperformance
Portfolio-level evaluation
Individual manager evaluation is a starting point; the real value comes from understanding how a manager fits into the overall portfolio. Managers with different stated objectives may share underlying factor exposures, reducing true diversification. Venn allows users to analyze correlations among individual holdings or sleeves and to source contribution to portfolio-level risk at the manager or asset level. The manager whose residual return is least correlated with the existing portfolio provides the most diversification benefit.
Connection to smart beta disruption
Venn’s factor decomposition directly relates to the Kahn & Lemmon (2016) framework: the portion of a manager’s active return explained by systematic factors is smart beta (available more cheaply via rules-based products), while the residual is the pure alpha that justifies active fees.
Comparison to other models
| Two Sigma Factor Lens | MAC3 | Liquid Factor Models | |
|---|---|---|---|
| Type | Return-based | Holdings-based | Return-based |
| Input | Return stream | Full holdings | Return stream |
| Factors | 18 (macro + styles) | 14 styles + industry/country | 10 (equity + rates + credit) |
| Asset classes | Multi-asset | Equity only | Multi-asset |
| Primary use | Manager evaluation, allocation | Risk management, optimization | Hedging, overlay |
| Provider | Two Sigma | Bloomberg | Academic (Rosenthal 2024) |