Who this is for — Readers who need to turn a list of ideas or instruments into a portfolio with a defined purpose that can be measured before and after implementation.
Portfolio construction is the process that converts objectives, constraints and assumptions into actual positions. It covers the investable universe, weights, liquidity, risk limits, benchmark, expected costs, rebalancing rule and monitoring system. Its output is more than a percentage table: it is a specification that explains why each position exists, what job it performs and when it should be reconsidered.
Sequence matters. Choosing products first and inventing the objective later usually produces a collection of preferences, not a governed portfolio. A target return is not yet investable until acceptable losses, horizon, liquidity needs and operational capacity have also been stated.
Seven decisions in the process
| Stage | Question that can be documented | Output |
|---|---|---|
| Objectives | What economic outcome should the portfolio support, and over what horizon? | objectives ranked by priority |
| Constraints | Which losses, cash flows, liquidity, leverage, tax and rules limit the choices? | policy and limits |
| Universe | Which assets and vehicles are eligible and genuinely accessible? | inclusion criteria and approved list |
| Allocation | How are capital and risk distributed? | target weights, ranges and budgets |
| Implementation | Which instruments, orders, currencies and hedges will express the allocation? | executable portfolio |
| Measurement | Which benchmark and metrics will make the outcome comparable? | performance protocol |
| Governance | Which events trigger review, rebalancing or a new policy? | calendar, bands and escalation rules |
These decisions are linked but not interchangeable. A 10% issuer limit is a constraint; a 10% target weight is an allocation decision; a 10% contribution to volatility is a measured outcome. Writing only “10% risk” makes it impossible to identify which quantity was controlled.
Objectives and liabilities come before the optimiser
An objective may concern real growth, capital preservation, income, funding a liability, absolute return or return relative to a benchmark. Several objectives can coexist, but they must be ranked and translated into horizons, cash flows and observable conditions.
The relevant definition of risk follows from the purpose. Volatility measures return dispersion; drawdown describes a peak-to-trough path; scenario loss links positions to a shock; shortfall risk compares assets with future commitments. None automatically replaces the others.
CFA Institute distinguishes asset-only, liability-relative and goals-based approaches. That distinction prevents the mean–variance frontier from being treated as a universal solution: a portfolio that is efficient among assets alone may still be unsuitable for future cash flows, obligations, taxes or liquidity needs.
Universe, data and estimates
The eligible universe should be screened for access, custody, liquidity, transparency, capacity, currency, cost and operational competence. A useful asset class in theory may not be implementable through the available vehicle. Two funds with similar labels can differ in index, replication method, securities lending, spread, tax treatment and operational risk.
Quantitative inputs are estimates rather than known properties:
- expected returns and risk premia;
- volatilities, covariances and tail dependence;
- inflation, interest rates, currencies and correlations with liabilities;
- spreads, commissions, impact, capacity and taxes;
- the frequency and size of external cash flows.
Each estimate needs an observation date, source, currency, frequency, horizon and method. Changing one coordinate can change the resulting portfolio. Sensitivity to inputs is itself a result to disclose, not something to hide behind the final weights.
Allocation, selection and implementation
Asset allocation establishes long-term exposures and, where the mandate allows them, temporary deviations. Selection chooses securities or strategies to implement those exposures. Capital allocation converts the design into amounts and weights; a risk budget distributes a specified risk measure.
Implementation tests whether the theoretical portfolio can survive:
- minimum quantities, lots, contract multipliers and rounding;
- spreads, commissions, market impact and turnover;
- collateral, leverage, margin and liquidity calls;
- hours, venues, currencies and maturities;
- concentration by issuer, factor, counterparty and exit route;
- dividends, coupons, securities lending and tax treatment;
- reconciliation of ordered, executed and booked positions.
An optimal weight that cannot be traded or maintained is not an implementable portfolio.
Example: from purpose to weight
Suppose a hypothetical portfolio has three purposes: a reserve for spending within twelve months, long-term growth and partial protection from unexpected inflation. The reserve amount and liquidity come first. Next, the designer considers which asset classes may support growth and diversification. Only then are vehicles, target weights and ranges chosen.
A 15% reserve is not “risk free”: it may carry credit, inflation, currency or reinvestment risk. A 60% growth allocation does not imply 60% of total risk. The design still needs contribution analysis, scenarios, liquidity and dependence checks. These percentages illustrate the sequence; they are not a model portfolio.
Measurement and review
The benchmark should be specified before evaluation, on a comparable basis for currency, frequency, total return, costs and investability. Portfolio return must distinguish external flows, income, cost and price movement. Attribution can then try to reconcile the outcome with the decisions actually made.
Review is not the same as chasing the latest winner. It can be triggered by weight drift, changed objectives, new liabilities, breached limits, deteriorating liquidity, structural changes in the universe or an invalidated assumption. Rebalancing returns the portfolio toward a policy that remains valid; adopting a new policy is a separate decision.
Common error — Judging the portfolio from final return alone. Without the objective, benchmark, cash flows, costs and risk taken, that number cannot identify the quality of the process.
Sources
- Harry Markowitz, Portfolio Selection, The Journal of Finance (1952) — expected return, variance, covariance and portfolio choice.
- CFA Institute, Principles of Asset Allocation — 2026 refresher reading — asset-only, liability-relative and goals-based processes, risk budgeting and mean–variance limitations.
- CFA Institute, Asset Allocation with Real-World Constraints — 2026 refresher reading — size, liquidity, horizon, tax, governance and rebalancing.
- Investor.gov, Asset Allocation and Diversification — official investor education on allocation, diversification and rebalancing.
- Victor DeMiguel, Lorenzo Garlappi and Raman Uppal, Optimal Versus Naive Diversification, The Review of Financial Studies — out-of-sample comparison and estimation error.