Throughput accounting is a decision method from the theory of constraints that judges every choice by three numbers, throughput (T), investment (I), and operating expense (OE), rather than by allocated unit cost. A decision is good if it raises T, or lowers I or OE, without hurting the other two.
Eliyahu Goldratt introduced the idea in The Goal (1984) and formalized it as an accounting alternative in later work. The point was blunt: standard cost accounting was built to value inventory for the balance sheet, not to make good operating decisions, and on the plant floor the two goals fight. This post covers what T, I, and OE mean, how the derived measures (net profit, ROI) fall out of them, and the one calculation that changes how a plant picks its product mix.
What is throughput accounting?
Throughput accounting is a way of measuring a business by cash generation rather than by cost allocation. It does not spread fixed costs and overhead across units to produce a per-unit cost. Instead it asks how fast the whole system turns raw material into sold product, and treats almost everything else as a fixed pool of operating expense that you spend to make that conversion happen.
That is a real departure from cost per unit produced which absorbs labor and overhead into every part whether or not making that part actually consumed more cash. Throughput accounting refuses the absorption step. Its claim is that most of the numbers a standard cost system prints, unit cost, absorption variances, efficiency against a standard, are accurate bookkeeping and terrible decision inputs, because they reward local activity that does nothing for the system.
The friction shows up in ordinary decisions. Standard costing tells a plant to keep every machine busy, because idle machines absorb no overhead and generate an unfavorable variance. Throughput accounting says keeping a non-constraint machine busy just builds inventory in front of the bottleneck, it raises I and does nothing for T. Two accounting systems, two opposite instructions, from the same floor. That gap is the reason the method exists.
What do throughput, investment, and operating expense mean?
The whole framework runs on three definitions, and the discipline is refusing to blur them.
| Measure | Definition | What counts |
|---|---|---|
| Throughput (T) | The rate the system generates money through sales | Sales revenue minus truly variable cost (usually raw material, sales commission, freight). Not a unit count, a money rate. |
| Investment / Inventory (I) | Money tied up in things the system intends to sell | Raw material, work-in-process, finished goods, and the money sunk into equipment and buildings. |
| Operating Expense (OE) | Money spent turning I into T | Direct labor, salaries, rent, utilities, most overhead. Largely fixed in the short run. |
The word that trips people up is throughput. In operational throughput it means good units per hour. In throughput accounting it means money, revenue minus the costs that genuinely vary with each unit sold. Labor is deliberately not a variable cost here: you pay the crew whether the line runs one part or one thousand, so their wages live in OE, not in T. That single reclassification is what stops the system from pretending overproduction is profitable.
How do you calculate net profit and ROI?
Both derived measures come straight out of the three primary ones, with no allocation step in between. Net profit is throughput minus operating expense: NP = T − OE. Return on investment is that net profit divided by the money tied up: ROI = (T − OE) / I. Two more measures round it out, productivity (T / OE) and inventory turns (T / I), but net profit and ROI carry most decisions.
Because there is no absorbed unit cost, there is also no way to look more profitable by building inventory you cannot sell. Under standard costing, a plant that overproduces parks labor and overhead in finished-goods inventory on the balance sheet, and the income statement smiles. Throughput accounting closes that door: unsold product adds to I and generates zero T, so overproduction looks exactly as bad as it is. That single property is why lean shops gravitate toward it, it agrees with the shop-floor gut feeling that a warehouse full of unsold parts is a problem, not an asset.
Why does product cost per unit mislead?
Because the scarcest resource in the plant is not a dollar of material, it is an hour at the constraint. One step sets the pace for the whole system, and every product you make spends some of that constraint's limited time. So the honest measure of a product is not its margin per unit; it is its throughput per unit of constraint time.
Consider two products competing for the same bottleneck. Product A earns $40 of throughput per unit but takes 8 minutes at the constraint. Product B earns $25 per unit but takes 2 minutes. Standard costing, looking at gross margin, prefers A. Throughput accounting divides by constraint time: A earns $5 per constraint-minute, B earns $12.50. Sell every minute of the constraint to B and you make more than twice the money, even though B is the "cheaper," lower-margin part.
This is not a trick of the example. It is the general result whenever products compete for the same bottleneck at different speeds, which is nearly always. Plants that reorder their mix this way often find their least glamorous, thinnest-margin SKU is quietly their best earner per constraint-hour, while the flagship product is a constraint hog. Standard costing had been telling them the opposite for years.
How do you use throughput accounting on the floor?
Run every operating decision through the same short filter, in order. This sequence turns the three measures into a habit a plant manager can use in a scheduling meeting without a spreadsheet.
- Find the constraint first. Throughput accounting is meaningless without knowing which resource limits the system. Look for the step buried in work-in-process with starved steps behind it; the constraint's time is the currency every product spends.
- State the decision as T, I, OE. For any proposal, a new order, a price, a make-or-buy, a capital request, ask what it does to throughput, to inventory/investment, and to operating expense. Ignore allocated unit cost entirely.
- Rank products by throughput per constraint-minute. Not by margin, not by revenue. The product that earns the most per unit of the bottleneck's time earns the most money for the plant, full stop.
- Say yes to any order that raises T without raising OE. If a low-price order fills otherwise-idle constraint time and covers its own truly variable cost, it adds throughput to the bottom line even at a price standard costing would reject.
- Treat every hour lost at the constraint as lost T forever. A breakdown, a changeover, or a quality escape at the bottleneck is not a local efficiency problem, it is money the whole system will never earn back. Track it as machine downtime at the constraint specifically.
- Judge cost cuts by their effect on throughput. A cut that trims OE but slows the constraint destroys more T than it saves. Only cut where throughput is untouched.
What do the numbers say about idle capacity?
Throughput accounting matters most when a plant has spare capacity it is not selling, and most plants do. Per the Federal Reserve's G.17 Industrial Production and Capacity Utilization release U.S. manufacturing capacity utilization ran about 75.8% in spring 2026, roughly 2.4 percentage points below its 1972–2025 long-run average. When a quarter of installed capacity sits idle, the standard-cost instinct to reject "unprofitable" low-margin orders can leave money on the table: any order that covers its truly variable cost and uses idle constraint time adds throughput. The idea traces to Goldratt's throughput-accounting work the accounting companion to the theory of constraints.
How does throughput accounting differ from operational throughput?
They share a name and a philosophy but measure different things. Operational throughput is a physical rate, good, sellable units per hour, the subject of our guide to throughput in manufacturing. Throughput accounting is a financial rate, dollars of contribution per period. Both come from the same insight that the constraint governs the system, which is also why both sit alongside the plant's manufacturing KPIs rather than replacing them. Raise physical throughput at the constraint and, in throughput-accounting terms, you have raised T without touching OE, the best kind of decision there is. The link runs the other way too: a throughput-accounting product ranking tells the scheduler which parts deserve the constraint's next free hour.
Where does measurement fit in?
Throughput accounting needs two numbers the standard system rarely captures cleanly: real constraint time per product, and honest downtime at the constraint. Neither survives being reconstructed from memory at month-end. That is the case for capturing production data where the work happens, the same shift from paper logs to real-time capture that plants like CLS made, so the throughput number in the morning meeting is the number the floor actually ran. To put a dollar figure on the constraint's losses before you start, run its numbers through a free OEE calculator.