Picking good factors is only the start. Turning factor scores into executable positions requires a portfolio optimization layer — this is where weights get allocated, risk gets controlled, and industry exposure gets managed. This post covers the core constraints and attribution discipline (methodology only; no parameters).
If your factor is especially effective in renewables, and you don't constrain industry exposure, the model will "overweight renewables." Surface explanation: factor working. Reality: industry beta driving everything. When the industry style rotates, the whole portfolio drawdowns together.
Industry-neutral constraints force the portfolio's exposure to each industry to match the benchmark (e.g., CSI 300 industry weights). This strips out industry beta, leaving relatively pure factor alpha. The cost is some upper-bound return, but stability improves dramatically — which matters more for institutional capital and licensed-advisor environments.
Risk attribution decomposes total return/volatility into risk sources:
| Source | Meaning |
|---|---|
| Industry | Industry exposure contribution |
| Style (Barra) | Size, Value, Momentum, Quality, Volatility, ... |
| Country / Market | Aggregate market beta |
| Specific | Residual = "pure alpha" |
The ideal factor strategy: specific (residual) alpha significant and positive; industry/style contributions near zero. If attribution shows 80% of return comes from industry beta, the "alpha factor" is actually industry timing — not sustainable.