Q72 is a portfolio optimization platform for independent wealth managers, family offices, and institutional investors. It runs the same asset universe and constraints through four portfolio construction methodologies — Q72 Confidence Alpha, Markowitz, Risk Parity, and Black-Litterman. The twelve classical portfolio outcomes are evaluated with walk-forward out-of-sample validation. An optional quantum-assisted refinement applies only to Q72 Confidence Alpha and is reported separately.

How is Q72 different from Sharpe-based tools like Bloomberg PORT or BlackRock Aladdin?

Q72 Confidence Alpha combines confidence-weighted forecasts, CDaR risk control, and correlation-aware selection. Markowitz, Risk Parity, and Black-Litterman remain independent reference methodologies under the same mandate and constraints. Q72 evaluates the twelve classical portfolio outcomes with walk-forward out-of-sample validation instead of relying only on in-sample metrics.

Q72 is designed for independent wealth managers, family offices, RIAs (Registered Investment Advisers), and institutional investors who need validated, side-by-side comparison of multiple optimization methodologies rather than a single black-box recommendation. It is a B2B SaaS platform — not available to retail or consumer investors.

What optimization methods does Q72 use?

Q72 compares four portfolio construction methodologies under the same mandate and constraints: Q72 Confidence Alpha, Markowitz mean-variance, Risk Parity, and Black-Litterman. The three reference methodologies retain their own independent logic and do not receive Q72 Confidence Alpha pre-weighting. Quantum is not a fifth methodology; it is an optional two-stage refinement of Q72 Confidence Alpha on real quantum hardware.

How is Q72's quantum refinement evaluated?

The twelve classical portfolio outcomes are evaluated with genuine walk-forward out-of-sample validation. The optional quantum refinement is assessed separately: a quantum candidate is retained only when it improves the same Q72 objective without increasing volatility, otherwise Q72 Classic remains. Any historical quantum performance figure is labelled as a replay of the final live weights, not as a walk-forward out-of-sample track record.

Does Q72 store client data?

Portfolio calculations require ticker symbols, capital amounts and allocation inputs, not end-client names or account numbers. Q72 also processes professional account, billing and support data, and consented analytics as described in its Privacy Policy. Results are scoped to the professional account.

The platform runs asset universes and constraints through multiple optimization engines simultaneously instead of relying on a single methodology.

Handling Large Asset Universes with Spectral Clustering

Key points in this article

Quantum hardware has qubit limits, but portfolio universes routinely contain 30–100+ assets. Q72 uses spectral clustering to partition the universe into correlation-coherent sub-groups, runs the quantum stage in parallel across clusters, then recombines. This preserves the correlation structure that makes the portfolio problem meaningful.

One of the most common questions we receive from technically sophisticated users is straightforward: quantum hardware has qubit limits. Portfolio universes routinely contain 30, 50, or 100 assets. How does Q72 apply quantum optimization to portfolios larger than the hardware can represent directly?

The answer involves a pre-processing stage that runs before any quantum circuit executes: spectral clustering. The technique partitions the asset universe into structurally coherent sub-groups, reduces the problem to a tractable size for quantum processing, and does so in a way that preserves the correlation structure that makes the portfolio problem meaningful in the first place.

01Why naive decomposition doesn't work

The obvious approach to a large universe is to divide it arbitrarily — take the first 20 assets, optimize them, take the next 20, optimize those, combine the results. This approach is computationally straightforward and produces reliably bad results. The reason is that portfolio optimization is fundamentally about exploiting the correlation structure between assets. An arbitrary partition destroys this structure.

Quantum hardware has qubit limits, but portfolio universes routinely contain 30–100+ assets.

02Spectral clustering: the intuition

Spectral clustering is a technique from graph theory and machine learning that partitions a set of objects based on their pairwise relationships — in this case, the correlation matrix of asset returns. The clustering algorithm finds the partition of this graph that minimizes the connections between groups relative to the connections within groups.

Intuitively, this produces clusters of assets that are 'close' to each other in correlation space — assets that tend to move together — while ensuring that the connections between clusters are as weak as possible. Each cluster is internally coherent, and the clusters together span the full correlation structure of the universe.

Portfolio construction & investment research

03How this integrates with Q72's quantum stage

In Q72, spectral clustering runs as the first step in the optimization pipeline for any universe above the direct quantum processing threshold. The quantum stage then runs in parallel across all clusters simultaneously, producing an optimized sub-portfolio for each cluster. The final step recombines the cluster results into a single portfolio allocation.

04What this means for portfolio quality

The practical effect is that the quality of the quantum optimization does not degrade significantly as the universe grows. This is not a compromise. Spectral clustering is itself a well-validated technique with a strong theoretical foundation and decades of empirical support in financial applications.