Capacity Planning for Execution Strategies: When the Strategy Hits Its Own Ceiling

A strategy has capacity. The capacity is a real, finite number, set by the relationship between the strategy's expected return per unit of notional and the market impact it generates per unit of notional. Below capacity, the strategy compounds at its risk-adjusted return; above capacity, it does not — by definition, the moment the strategy hits its capacity ceiling is the moment the marginal trade ceases to add to the strategy's total expected return.

Discovering capacity by hitting it is expensive. Estimating capacity in advance is straightforward when the right inputs are available, and most of the difficulty is in the inputs rather than the model.

Why capacity is finite

Every alpha-generating strategy works by exploiting a price dislocation. The act of trading on the dislocation tightens it. A small amount of trading tightens it a little; a large amount tightens it enough that the strategy's edge is consumed by the same trading that is supposed to extract it.

Mathematically, the strategy's return per unit notional declines as notional rises, while the strategy's market impact per unit notional rises (slowly at first, then faster). The intersection of those two curves is the capacity. Up to that intersection, every additional dollar adds expected return; past it, every additional dollar consumes the return of the dollars that came before.

The market-impact function

Most quantitative work uses a square-root market-impact model: impact ∝ σ × sqrt(participation rate). This is the Almgren-Chriss family and it is approximately right for execution-style strategies trading order-of-day-volume fractions. For strategies trading in smaller fractions or holding for longer than a day, more elaborate models apply (linear-then-square-root, propagator models), but the square-root version is the right place to start.

The model has two empirical parameters: the daily volatility σ of the instrument, and a venue-specific impact coefficient that captures how aggressively the local book pushes back against participation. The volatility is observable; the coefficient is estimated from your own historical fills against your own historical TCA. Drovix institutional clients receive their estimated coefficient with each weekly TCA report.

The half-life function

Strategy capacity also depends on how fast the alpha decays. A strategy with a short information half-life can be scaled higher than one with a long half-life, because the impact is concentrated within the half-life and the residual market-information leak is bounded.

Operationally, the half-life enters the capacity calculation through the maximum participation rate the strategy can sustain. A 1-hour half-life strategy can participate at 5-10% of intraday volume; a 1-day half-life strategy can participate at 1-3%. The capacity is the half-life-appropriate participation rate, multiplied by the volume of the instrument, integrated over the strategy's trading window.

Estimating capacity in three steps

1. Estimate the alpha decay

Measure the cumulative alpha of the strategy as a function of position-held duration. The shape is typically rising, peaking, then declining. The peak is the optimal holding period; the slope of the rise determines how quickly the alpha would dilute if the position were scaled.

2. Estimate the market-impact coefficient

Compute the volume-weighted average implementation shortfall on a representative sample of historical fills. Regress the shortfall against sqrt(participation rate × volatility) on a per-instrument basis. The regression slope is the impact coefficient.

3. Solve the intersection

With the alpha and the impact in hand, solve for the notional N at which the marginal alpha equals the marginal impact. That is the capacity. Translate it into more useful units — daily notional, monthly notional, percentage of average daily volume — depending on how the strategy will be managed.

What undermines a capacity estimate

• Backtest survivorship bias. The capacity estimated on a historical sample reflects the market conditions that prevailed in the sample. Capacity in a regime change is generally lower; a strategy estimated to have $200m capacity in 2024 may have $80m capacity in 2026 because the underlying liquidity has shifted.
• Crowding. If the same strategy is being run by other desks, the effective capacity is the total capacity divided across operators, not your standalone share. Crowding is hardest to estimate directly; the indirect indicator is whether the strategy's recent decay rate is faster than its long-run average.
• Venue dependence. The impact coefficient is venue-specific. A strategy with $200m capacity on Drovix's wholesale book may have $40m capacity on a thin venue, and the difference dominates the strategy's profitability when scaled.
• Style drift. A strategy whose half-life shortens over time has higher capacity but lower margin per dollar. A strategy whose half-life lengthens has the opposite. Treat capacity as conditional on the current half-life, not the long-run average.

What Drovix surfaces

Each institutional account receives an estimated impact coefficient per instrument, recomputed monthly from the account's own historical fills. The coefficient is exposed in the post-trade report alongside the implementation shortfall. A risk officer can use the coefficient to compute capacity for any strategy spec without re-running TCA from scratch.

We do not estimate the alpha side of the capacity calculation — that is properly the operator's responsibility, and the alpha is something the operator should not be sharing with their venue anyway. What we do is provide the impact side with enough precision that the operator can plug it into their own capacity model and get a number that survives a peer review.

Operational discipline

• Set the strategy's daily notional cap at no more than 60% of estimated capacity. The cushion absorbs the difference between estimated and realised impact during regime changes.
• Tag each fill with the instantaneous participation rate; alert when the rate exceeds the strategy's design participation by more than 50%.
• Recompute capacity quarterly. The number drifts; the calendar reminder is what keeps the strategy from quietly exceeding its own design.
• Treat 'we are running a strategy at capacity' as a P0 condition to be reviewed at the next risk meeting, not as a permanent state.

Where to go next

→ RFQ vs Streaming by Notional — once you know your capacity, the next question is how the resulting notional is best executed against the venue's pricing modes.

→ Decomposing Execution Cost — the cost components a capacity calculation has to forecast accurately.