Cross-asset portfolio diversification is often reduced to a search for more asset classes. Yet the experience of 2022 showed why that can be misleading: equities and government bonds may carry different labels, but at times they may share exposure to the same macroeconomic forces. When inflation and discount rates rise together, a traditional 60/40 bond/equity portfolio can behave like one single concentrated regime trade.

For multi-asset investors, a more practical question is therefore not how many assets or markets a portfolio owns, but how many genuinely distinct return engines it contains, be it beta or alpha return streams. That shifts the focus from asset selection to regime analysis and signal construction.

Some concepts travel well across equities, bonds and FX. For example trend, carry, valuation and macro-surprise signals can be applied across assets and markets, provided they are translated into the economics of each universe and normalised for differences in risk, liquidity and trading costs. What does not travel is the raw indicator itself. A positive growth surprise, for example, may support equities, lift bond yields and strengthen a currency - but the magnitude and even the direction of those responses may depend on inflation, central-bank reaction functions and starting valuations among others.

When designing a signal and strategy, a strong full-sample Sharpe ratio is not enough. A robust cross-asset signal should have firstly a clear economic or behavioural rationale, survive walk-forward and out-of-sample testing, and remain useful after realistic turnover, financing and execution costs. Most importantly, its conditional behaviour matters: a strategy with a low average correlation to a 60/40 portfolio is not a reliable diversifier if its correlation rises sharply during the portfolio's worst months. Smoothing is not diversification, and neither is a return stream whose apparent independence disappears in the tails.

Process-level diversification can also help manage alpha decay. Instead of relying on a single formulation, investors can diversify the "what, how and when" of a strategy: instruments traded, signals modelling, investment horizon and rebalancing rules. Combining structurally different signals with dynamic risk scaling can reduce dependence on any one model or market regime. With regards to implementation, liquid derivatives can be more capital-efficient, while leverage may be used to scale a diversified portfolio — rather than concentrating risk. All these considerations obviously require a robust investment platform, ensuring high quality data, detailed transparent monitoring and robust implementation.

Ultimately, the goal is not to collect the largest possible number of assets or signals, but to build a portfolio of positive return streams failing at times for different reasons. In a market regime where traditional correlations are less dependable, investors must understand the difference between owning many positions and being genuinely diversified.