Most Portfolios Are Assembled. Yours Should Be Constructed.

Here is how most portfolios get built. You fill out a questionnaire. It spits out a risk score. The score maps you to a model: conservative, moderate, or aggressive. You are dropped into the matching basket of funds, the same basket held by a thousand other clients with the same score, and that is called portfolio construction.

That is assembly, not construction. The investments came off a shelf, and the only decision anyone made about you was which shelf.

Assembly is efficient for the advisor. A handful of models can run hundreds of households, get rebalanced on a schedule, and never require anyone to look at your actual situation. What it is not is optimal for you, because the things that drive your after-tax outcome are exactly the things a model cannot see.

Construction Has Layers a Model Skips

Real construction begins before a single fund is chosen, with the whole picture: every account across every custodian, your cost basis lot by lot, your marginal rates this year and the years ahead, your concentrated positions, your liquidity needs, the holdings you already own, and what you actually want the money to do. The portfolio is built around that. It is not selected from a menu and adjusted at the edges.

Allocation is the first layer, not the last. What you own still matters, but it is the beginning of the work.

Location decides where each asset sits. The same holding can be tax-efficient or tax-poisonous depending on which account holds it. Bonds and REITs throw off ordinary income and belong in tax-deferred accounts; your highest-growth assets belong in a Roth where the growth is never taxed. A model that manages accounts in silos cannot make this call. We covered the full logic in All the Right Things in All the Wrong Places.

Tax-lot construction is where most of the hidden return lives. Owning the index as individual stocks rather than a fund, through direct indexing, turns every position into a tax tool. Losses can be harvested at the security level all year, even when the index is up, and banked to offset gains elsewhere. A fund buries that opportunity where you can never reach it.

Risk is engineered in deliberately, not inferred from a five-question quiz. Real risk management means understanding the correlations, factor exposures, and concentration already sitting in the portfolio, including the ones a client did not know they had, and constructing around them. That is the discipline of institutional portfolio management, and it does not fit on a questionnaire.

Concentrated positions and financing are part of construction, not afterthoughts bolted on later. A large low-basis holding shapes everything around it, and liquidity is better raised by borrowing against the portfolio than by selling and triggering tax. These belong in the design from the start.

It Is One System, Not Five Features

The layers are not a checklist run in sequence. They feed each other. Direct indexing fills a loss bank; those losses offset the gain when a concentrated position is finally diversified; box spread borrowing raises cash without realizing gains, so the loss bank stays intact; location decides where each piece lives so the tax treatment is never wasted. Run as a system, each decision makes the others more powerful. Run as isolated products, most of the value leaks out between them.

Why Almost Nobody Does It This Way

The honest answer is scale. Bespoke construction does not scale the way models do, so the industry standardized. Most advisors run hundreds of relationships on a few model portfolios because it is efficient for the firm, and the after-tax cost of that efficiency is quietly paid by the client, every year, in a number nobody puts on the statement.

The whole approach sits in our guide to tax-aware investing. The short version: a portfolio should be built around the person who owns it, engineered for after-tax outcome, and run as one system. Anything less is just assembly with a nicer name.