← Frameworks

Correlated Execution Risk

Two operating investments that share a team, a stack, a review queue, or a failure mode are not statistically independent. When the shared dependency fails, both fail together. Sharpe ratios cannot be summed; NPVs cannot be added without an adjustment. Correlation is the input that makes Stage 3 different from Stage 2 - it is why portfolio construction beats project-by-project selection.

Two projects, equal E[R] and σ. Portfolio risk depends on ρ.ρ = -1perfect hedgeσ_p = 0(theoretical)ρ = 0independentdiversificationσ_p = σ/√2ρ = +1shared team,shared stack,shared reviewσ_p = σ(most ops default)Lower bars mean lower portfolio risk. Higher correlation = higher portfolio risk = tighter capacity.

Where Correlation Comes From

Shared team. Two projects staffed from the same team share the same risk of that team getting reorganized, losing a key member, or being pulled to a fire. A 50% shared roster implies meaningful positive correlation.

Shared stack. Two automations built on the same model version, the same data pipeline, or the same verifier share the same risk of that stack failing. When the stack breaks, both projects break.

Shared review queue. Two projects whose throughput depends on the same human reviewer are correlated in time - the reviewer’s attention is the shared resource. Slippage in one compresses capacity for the other.

Shared narrative. If two projects are both pitched as evidence for the same strategic thesis, the failure of one undermines the other even if they are technically independent. The correlation is in the audience’s mind, and that is still real.

Why the covariance matrix is the hard part

For N instruments, the covariance matrix has N(N−1)/2 off-diagonal entries. For 20 candidate instruments, that is 190 pairwise correlations. In public equities, these are estimated from long historical return series. In operating investments, they must be estimated from execution-dependency structure - which teams, which stacks, which review queues, which narratives.

A common failure mode: treat projects as independent because correlation is hard to estimate. The portfolio looks well-diversified on paper; the actual portfolio has a handful of shared failure modes that make it concentrated. The calibration of the covariance matrix - not the solution of the optimization - is where the operational alpha of Stage 3 actually comes from.

Rosetta Stone

Four circles, four readings of the same object. Each role reads the artifact through its own lens.

Related