Portfolio Risk Analysis: VaR, CVaR, Factors, and Drawdown Explained

Direct answer: What is the difference between VaR and CVaR?

Portfolio risk analysis should combine VaR, CVaR, drawdown, factor exposures, correlation, liquidity, and data-quality checks so a risk signal becomes an actionable decision instead of a scary number.

How Celestice helps

Celestice links holdings, factors, exposures, data freshness, scenarios, and policy thresholds in a live risk workspace. In practice, Celestice helps teams move from a risk number to an evidenced alert, an explanation, and a reviewed action path.

A single risk number is never enough

Ask most investors how risky their portfolio is and they will quote a volatility figure. It is a start, but volatility alone describes the width of normal fluctuations — it says little about how bad a bad day can get, what is actually driving the risk, or whether the number you are looking at is even trustworthy. A genuine risk review answers four different questions: how much could I lose, how bad is the tail, what is causing the exposure, and is this signal real?

Value at Risk and the tail beyond it

Value at Risk (VaR) estimates the most you would expect to lose over a given horizon at a given confidence level — for example, "with 95% confidence, this portfolio should not lose more than 4% in a month." It is a useful, intuitive summary, but it has a blind spot: it tells you the threshold, not what happens when you cross it.

That is why Conditional VaR (CVaR), also called expected shortfall, matters more for serious risk work. CVaR answers: given that you breach the VaR threshold, how much do you lose on average? It measures the severity of the tail, not just its edge. Two portfolios can share the same VaR while one has a far uglier CVaR — and that difference is exactly the catastrophic-loss risk you most want to know about.

Crucially, VaR is not one number. It changes with the method (historical, parametric, Monte Carlo), the confidence level (95% vs 99%), and the horizon (one day vs one month). A complete review lets you switch among these and see how the risk picture shifts, rather than trusting a single canned figure.

The metrics that round out the picture

Alongside VaR and CVaR, a few measures give risk its shape:

• Beta — how much the portfolio moves with the broader market. High beta means market swings are amplified.
• Sharpe ratio — return earned per unit of risk taken. It reframes performance as risk-adjusted, so a high return achieved by taking enormous risk is exposed for what it is.
• Maximum drawdown — the worst peak-to-trough decline actually experienced. This is the number that tests an investor's nerve, because it is the loss they would have had to live through without selling.

Factor decomposition: what is actually driving risk

Knowing how much risk you have is less actionable than knowing where it comes from. Factor decomposition breaks the portfolio's risk into its underlying drivers — exposure to the overall market, to sectors, to styles like value or growth, to interest rates, and so on.

This is where hidden concentration surfaces. A portfolio that looks diversified by holding fifty stocks may in fact be a single large bet on one factor — say, mega-cap technology — dressed up as diversification. Correlation analysis adds to this: when supposedly independent holdings all move together in a stress event, the diversification you counted on was an illusion. Seeing the factor and correlation structure tells you whether your risk is spread or stacked.

Market risk versus data noise

Here is the discipline most tools skip. A risk dashboard will happily compute a frightening number from bad data — a reconciliation break, a stale price, an unsettled position — and present it with the same confidence as a real exposure. Acting on that is worse than doing nothing.

A trustworthy risk workflow explicitly separates genuine market-risk pressure from reconciliation and data-quality caveats. Before you treat a threshold breach as a reason to trade, the system should confirm the underlying data is current and reconciled. "Is this signal real?" is a question that belongs before "what should I do about it?"

From risk signal to the right next step

Risk review is not an end in itself; it is a routing decision. Once you have established what risk is present, why it matters, and that the signal is real, the next step might belong in optimization (re-weight to reduce the exposure), rebalancing (repair drift that created it), trading (act on an approved change), or simply documentation and committee follow-through. The value is in connecting the diagnosis to the right workflow rather than leaving a scary chart with no path forward.

This is the perceive-and-reason core of an agentic wealth platform: aggregate the position data, compute risk under the chosen method and horizon, decompose it into drivers, flag the data caveats, and surface the most appropriate next action with the evidence attached — leaving the decision, and any high-impact trade, under human approval.

The takeaway

Reading portfolio risk well means refusing to settle for a single statistic. Use VaR for the expected threshold and CVaR for the tail beyond it; use beta, Sharpe, and drawdown to give the risk shape; use factor and correlation analysis to find where it really lives; and always separate true market risk from data noise before acting. Done this way, risk analytics stops being a wall of numbers and becomes what it should be — the first step toward a better decision.