Throughput Accounting

Definition

Throughput Accounting (TA) defines three simple, global measures: Throughput (T) — sales revenue minus totally variable cost; Inventory/Investment (I) — money tied up in things the system intends to sell; and Operating Expense (OE) — the recurring money spent turning inventory into throughput. Goldratt: "Tell me how you measure me and I will tell you how I will behave." Unlike cost accounting, TA does not allocate overhead to products or treat inventory buildup as profit; it evaluates every decision by its effect on T, I, and OE together, with priority given to protecting Throughput at the system's constraint.

In the Book

Chapter 1 traces TA's origin: traditional cost accounting "encourages any production, even on a non-bottleneck," because it treats absorbed cost as increased profit, so TOC's first implementations immediately clashed with it and TA was developed as the necessary alternative, alongside distinguishing the "Cost World" from the "Throughput World." Chapter 13 grounds this in Goldratt's classic P-Q example: a company producing three products (X, Y, Z) through shared resources, where Resource 2 is the constraint. Ranking products by Throughput-per-minute-of-constraint-time (Z at $26.60/min, X at $10.00/min, Y at $9.15/min) yields a different, more profitable production sequence and a different weekly income ($12,858) than ranking by traditional gross margin or activity-based costing, which pick worse sequences because they ignore which resource is actually scarce. The chapter extends the logic to a quality-improvement example: which of three defective products to fix first is answered correctly only by tracing the cost of lost time specifically at the constraint, not by looking at scrap rates or unit profitability in isolation.

Why It Matters

Throughput Accounting shows that a measurement system is not neutral — it encodes what "doing well" means and therefore drives behavior, and a measure that ignores the system's actual bottleneck will systematically reward locally rational, globally harmful choices (running non-constraint machines flat out, prioritizing the "most profitable" product by the wrong yardstick). The general lesson transfers to any domain where local metrics obscure a shared bottleneck: measurement, not intent, determines whether people optimize the part or the whole.