Execution Cost
The total cost of converting a trading decision into a position — spread, commissions and fees, slippage, market impact, financing, and the opportunity cost of orders that were rejected or never filled.
Execution cost is everything the market charges between the decision and the position. The framework that takes it seriously is implementation shortfall, proposed by Perold in 1988: compare the paper portfolio, which transacts instantly at decision prices in unlimited size, against the real one. The difference is the true cost of execution, and it is always larger than the visible line items.
The components differ sharply in visibility. Spread is the most visible and the most advertised — though advertised spreads are averages drawn from calm regimes, not from the moments a trader’s orders actually cluster in. Commissions and financing appear on statements. Slippage appears only if decision prices were recorded. Market impact — the price moving because the order itself consumed liquidity — is invisible without careful measurement. Least visible of all is the cost of orders that never became trades: fills refused under last look, requotes declined, entries missed while the price ran. A cost accounting that omits the unfilled orders is a survivor-only accounting.
The backtest implication is direct. Most simulations model execution cost as a single flat number — a fixed spread, a fixed slippage allowance — when the real quantity is regime-dependent and correlated with exactly the events a strategy trades. A stop triggered during spread widening does not pay the average; it pays the event regime. Flattering the cost assumption is one of the quietest ways a dead strategy stays alive on paper.
The honest property of execution cost is that it is measurable from records the trader already owns: order timestamps, quoted prices, fill prices, and statements are sufficient to reconstruct the distribution — not the anecdote — of what execution actually cost. The Execution Cost Auditor we are building is intended to compute exactly that from imported statements, because retail transaction-cost analysis effectively does not exist.