# Theory of Constraints The theory of constraints finds the one constraint that limits throughput, and organizes the rest of the system around it. **The question.** What one constraint limits throughput, and is the rest of the system organized around it? Also known as: TOC. The improvement cycle is the five focusing steps. The accounting view is throughput, inventory, and operating expense. ## When to use it - A plant or service chain is missing output, every department is busy, and capital requests are coming from several machines at once. - Work-in-process is high, and improving a non-bottleneck has not raised shipments. - You need a rule for what to run, what to starve, and where a buffer belongs. - Cost-cutting is aimed at operating expense while the constraint sits idle for lack of material or a policy. ## When to avoid it - There is no stable system and no repeating flow, so "the" constraint cannot be named. Stabilize first. - The constraint is unknown because data on queues, downtime, and mix are missing. Collect those before running the five steps as a slogan. - The question is how much of a non-constraint item to order. That is a lot-size question, and EOQ or a pull rule may be enough. - Leadership will not change local efficiency targets. Subordination fails if every machine is still measured on utilization. ## Overview The theory of constraints says a system's throughput is limited by one constraint at a time, so improvement work should find that constraint, get more from it, and make every other resource serve it. Elevation, meaning added capacity, comes after that. It supports the choice of where to spend attention and capital, and which local efficiency targets to drop. The usual failure is to treat every busy resource as a bottleneck, or to buy a machine before the current one is fully used. Subordination is the hard step, because it requires non-constraints to run below the utilization their supervisors are used to reporting. When the constraint moves, old rules written around it have to move with it. ## The five focusing steps ### Identify Name the one constraint that most limits the system's goal, usually throughput now and in the future. In a plant it is often a machine, a skill, or a policy. If everything is called a bottleneck, you have not identified it. Rank by impact on the goal, not by how loud the complaint is. ### Exploit Get more throughput from that constraint with the resources you already have. Cut idle time on it, run the mix it should run, quality-check before it so it never works on scrap, and keep it fed. Exploitation is cheap. Do it before you buy another machine. ### Subordinate Run every other resource to serve the constraint, even if that means they sit idle. Release work at the constraint's pace. Local utilization targets on non-constraints fight this step and rebuild inventory. ### Elevate Add capacity at the constraint only after exploit and subordinate. Overtime, an extra shift, a second machine, or outsourcing are elevation. If you elevate first, you spend capital to hide a policy problem. ### Repeat When the constraint is broken, go back to identify. Goldratt's warning is not to let inertia become the constraint: rules written around the old bottleneck, such as batch sizes or efficiency bonuses, will limit the system after the machine is no longer the limit. ## What it is not - A license to ignore cost. Operating expense still matters. The claim is that raising throughput at the constraint usually beats shaving expense off a non-constraint. - The same as lean waste-hunting. Lean attacks waste everywhere. TOC attacks the one limit to throughput first. Use both, in that order, when the stream is long. - A utilization drive. Raising utilization on a non-constraint builds inventory and can starve the constraint of the right mix. ## The steps 1. **State the goal and the measures.** Write the system's goal in one sentence. For a for-profit plant Goldratt used make more money now and in the future, tracked as throughput, inventory, and operating expense. If those three are not visible, you cannot tell whether a local improvement helped. 2. **Find the constraint with queues and lost time.** Walk the flow. The constraint is where work waits, where lost time costs the whole system's output, or where a policy stops release. Name one. If you name four, keep going. 3. **Exploit, then subordinate, before you spend.** List the idle time, bad mix, and rework on the constraint, and remove them. Change release rules and local targets so non-constraints feed it. Write down the old efficiency measures you are switching off. 4. **Elevate only if it is still the constraint.** Recheck after exploit and subordinate. Add capacity only if the same resource still limits throughput. Tie the spend to a measured gap, not to a utilization chart. 5. **Watch for the constraint to move.** When output rises, the queue will appear somewhere else, including in the market. Repeat identify. Review policies written around the old constraint, or they become the new one. ## In practice A food plant cannot hit the weekly ship plan. Three packaging lines report 85 percent utilization, and the capital request is a fourth line. Walking the flow shows the constraint is not packaging. It is a single cooker with 14 hours of planned production a day, 90 minutes of which are lost to waiting for the next batch of mix, and 40 minutes to changeovers clustered on Tuesdays. Mix is released in large MRP batches, so the cooker alternates between idle and a queue of the wrong product. Packaging is busy because it runs ahead on easy SKUs and then waits. Exploit is a standing mix buffer in front of the cooker, a changeover sequence that groups similar recipes, and a rule that QC samples the mix before the cooker, not after. Subordinate is releasing mix at the cooker's pace and measuring packaging on feed to the cooker and on on-time cases, not on line utilization. After four weeks cooker output is up 11 percent and finished-goods inventory is down. Elevate, a second cooker, is taken off the list. The next constraint is a filler on one SKU family. ## What you leave with - A one-sentence goal and the three measures, or an agreed substitute in a non-profit setting. - A named constraint, with the evidence that it, not another resource, limits throughput. - An exploit list and the local targets that will be subordinated or dropped. - A decision on whether elevation is still needed, with a number attached. - A review date for where the constraint has moved. ## Failure points ### Every resource called a bottleneck Departments bring their own capacity charts and all get funded. The system's throughput does not move, because only one of them was limiting it. ### Elevation first A new machine is bought while the current one waits for material, runs scrap, or changes over twice a shift. Capital hides the policy. ### Non-constraints kept at high utilization Supervisors are still scored on keeping people and machines busy. They release work early, the constraint gets the wrong mix, and inventory rises. ### Inertia after a win The old bottleneck is broken and the batch-size rules, bonus, and schedule written around it stay in place. Those rules become the constraint. ### Improving a machine that is not the constraint OEE on a non-constraint goes up, extra output sits in a pile, and shipments are unchanged. Measure the constraint's throughput, not every machine's efficiency. The failures above are usually [Framing](https://usefulconsultingframeworks.com/your-edge/#framing): the wrong problem was chosen before this framework was applied. ## Origin Eliyahu Goldratt set the theory out in The Goal, published by North River Press in 1984 and written with Jeff Cox as a plant-manager novel. The book argues that a system has at least one constraint, that throughput is governed by that constraint, and that local efficiency elsewhere can make the system worse. The five focusing steps, also called the process of ongoing improvement, are the working procedure. Goldratt later extended the ideas into critical-chain project management, thinking processes, and throughput accounting. Nobody needs a license to use the five steps. Goldratt's books and the TOC materials sold by successor firms are copyrighted. This note describes the focusing steps and the three measures. It does not reproduce Goldratt's diagrams or the later thinking-process tools. ## Further reading - [The Five Focusing Steps](https://northriverpress.com/wp-content/uploads/2018/01/Free-download-5FS.pdf). Goldratt's own short account of the five steps, hosted by North River Press, the publisher of The Goal. Free PDF. - [Theory of constraints](https://en.wikipedia.org/wiki/Theory_of_constraints). Wikipedia's overview of the five steps, throughput accounting, and later extensions such as critical chain. Check its sources before quoting. ## Related frameworks - [Overall Equipment Effectiveness](https://usefulconsultingframeworks.com/frameworks/oee/): Splits losses on a single machine. Pick it to exploit a machine constraint, not to rank every line in the plant. - [Value Stream Mapping](https://usefulconsultingframeworks.com/frameworks/value-stream-mapping/): Shows where inventory and waiting sit door to door. Pick it when you need the picture of the stream before naming the constraint. - [Economic Order Quantity](https://usefulconsultingframeworks.com/frameworks/eoq/): Sizes a repeat order when demand is steady. Pick it for a non-constraint item. Do not use it to load the constraint in large batches. - [PDCA](https://usefulconsultingframeworks.com/frameworks/pdca/): The cycle to run on each exploit change. Pick it when the constraint is named and you are testing a small fix on it. Last updated October 20, 2026.