Learn
Allocation sleeves vs. asset classes
The two-layer way to describe a portfolio - strategic sleeves on top, granular asset classes underneath - and why keeping both layers makes targets and drift meaningful.
Most allocation advice gives you a single layer: "60% stocks, 40% bonds." That is fine until you actually have to manage the portfolio, at which point one layer is too coarse to act on and too vague to hold yourself to. A more workable model uses two layers: strategic sleeves on top, granular asset classes underneath.
The two layers
- A sleeve is the strategic bucket - the role a slice of the portfolio plays. A common set is Income, Equity, Alternatives, and Cash. Sleeves answer "how much of the portfolio is doing each job?"
- An asset class lives inside a sleeve and is the analytical unit you can actually map a holding to - US large-cap, international developed, TIPS, investment-grade corporates, REITs, and so on.
The Equity sleeve might hold US large-cap, US small-cap, and international classes. The Income sleeve might hold Treasuries, munis, and corporates. The sleeve tells you the strategy; the class tells you what is really inside it.
Why both layers matter for a self-directed household
Targets and drift only mean something at the level you set them. If your only target is "40% income," a portfolio that has quietly become all long-duration Treasuries still looks on-target - even though its real risk has changed completely. Set targets at both levels and the drift becomes visible where it actually happens.
A worked example: suppose your Equity sleeve target is 60%, split into 40% US and 20% international. A strong US run pushes the sleeve to 64% - a 4-point sleeve drift - but the US class inside it has drifted from 40% to 46%. The sleeve number understates what moved. Two layers catch it; one layer hides it.
How Ironlake treats it
In Ironlake the Asset Allocation Model owns this hierarchy: sleeves at the top, asset-class targets and per-class assumptions (expected return, yield, volatility, tax character) underneath. Your IPS references the active model rather than restating the targets. Rebalancing drift is detected at both levels, so a sleeve that looks fine can still flag a class that has wandered.
The model is a first-class object, not a derived view of your holdings - which is what lets you build and project one before importing a single position, and compare several models (current, retirement, post-windfall) side by side.
- See how the targets feed the projection: how the projection engine works.
- Build one: create a model and run the projection.
Honest limits
The sleeve names are a convention, not a law - you can define your own. And a model is a target, not a prediction; the per-class return and volatility assumptions are inputs you set, and the projection is only as good as they are. Ironlake shows the math behind a target and flags the drift; the allocation decision is yours.