Bottleneck and Constraint Analysis: Find the One Step That Sets Your Output

Bottleneck and Constraint Analysis sits in the Analyse phase of RDMAICS. It does one job. It identifies the step with the highest demand-to-capacity load, then helps you manage that constraint through the Five Focusing Steps.

The logic is simple. A process can only produce as fast as its slowest capable step. You can improve four steps out of five and see no change in output. Until you act on the constraint, you are moving work around rather than moving more work through.

Reach for this tool when your process has a clear sequence of steps, when you can state a shared rate unit such as orders per hour, and when queues are building somewhere. If work is piling up in one place and starving in another, this tool will tell you where the ceiling sits.

The example used throughout this guide is fictional. It is the sample data that loads with the tool, and it is there so you can see the mechanics before you put your own numbers in.

The process is order fulfilment. Five steps run in sequence: order entry, picking, quality check, packing and dispatch. The rate unit is orders per hour. Demand on the process is steady at 95 orders per hour across every step, because every order must pass through every step.

What differs is capacity. Order entry can handle 120 orders per hour. Picking can handle 80. Quality check can handle 55. Packing can handle 100. Dispatch can handle 110. Those five numbers decide everything that follows.

Start by naming the process and the rate unit. In this case the process is order fulfilment and the rate unit is orders per hour. The rate unit matters more than people expect. If one step is measured in orders per hour and another in cartons per shift, the comparison collapses.

Then list the steps in order. For each one, enter capacity, demand, WIP or queue, and flow time. Capacity is what the step can do at its current design and staffing. Demand is what arrives. WIP is the work sitting in front of the step. Flow time is how long work takes to clear the step.

In the sample, the WIP and flow time columns tell their own story before any calculation runs. Order entry holds 8 in queue with a flow time of 4. Picking holds 45 with a flow time of 18. Quality check holds 120 with a flow time of 37. Packing holds 22 with a flow time of 9. Dispatch holds 5 with a flow time of 5.

Queue and flow time are not used to pick the constraint. Load does that. But they act as a sanity check. If your calculated constraint does not also hold the largest queue, go back and test your capacity figures.

The tool divides demand by capacity at each step and returns a load percentage. Order entry sits at 79.2 per cent. Picking sits at 118.8 per cent. Quality check sits at 172.7 per cent. Packing sits at 95.0 per cent. Dispatch sits at 86.4 per cent.

The result names the constraint as quality check, at the highest load of 172.7 per cent, from 95 demand divided by 55 capacity. The chart plots capacity, demand and WIP for every step on one axis, so you can see where the demand line crosses above the capacity bar.

Two steps are above 100 per cent. Picking at 118.8 per cent cannot keep up either. That is worth noting, but it is not where you start. Quality check is the binding constraint. If you fixed picking alone, the extra work would simply arrive faster at a step that can only process 55 orders per hour.

The 120 units of WIP sitting at quality check and the flow time of 37 confirm the finding. After the constraint, queues collapse. Packing holds 22 and dispatch holds 5, because quality check is only ever feeding them 55 orders per hour.

The tool also offers an AI review that explains the result in plain language if you want a second read on what the numbers are showing.

Naming the constraint is only half the tool. The other half is the Five Focusing Steps, which turn the finding into assigned work. Each step carries an action, an owner and a status, so the analysis does not stop at a chart.

In the sample case the five steps are populated as follows. Identify the constraint is set as protecting quality-check time from interruptions, owned by the quality lead, in progress. Exploit the constraint is set as sequencing work before quality check, owned by the warehouse lead, in progress. Subordinate everything else is set as releasing orders to match the constraint rate, owned by the planning lead, planned. Elevate the constraint is set as adding a parallel verification station, owned by the operations manager, planned. Repeat after the constraint moves is set as recalculating utilisation after the pilot, owned by the improvement lead, planned.

Note the sequence of effort. Exploit and subordinate come before elevate. You get more from the existing 55 orders per hour before you spend money adding a station. Releasing orders to match the constraint rate is the subordination move that stops the 120 unit queue from growing further.

The fifth step is the one people skip. Once quality check is lifted, the constraint moves. Picking at 118.8 per cent is the obvious next candidate. The tool expects you to run the analysis again.

The most common error is treating the busiest step as the constraint. Busy and constrained are not the same thing. Order entry at 79.2 per cent may feel flat out to the people working there. It is not capping output.

The second error is mixing units. Capacity and demand must share one rate unit. In this case both are orders per hour. If they are not aligned, the load percentages are meaningless even though they will still calculate.

The third error is using theoretical capacity rather than real capacity. If quality check can process 55 orders per hour only when fully staffed and uninterrupted, then 55 is an optimistic number, not a conservative one.

This tool will not tell you why quality check can only handle 55 orders per hour. It reports the gap. It does not diagnose the cause. The cause might be staffing, equipment, rework arriving from picking, or a verification method that takes longer than it needs to. You have to find that out through observation and other Analyse phase tools.

It will not validate your inputs. If you enter a capacity of 55 when the true figure is 40, the tool will calculate a load percentage with complete confidence and point you at the wrong number. The arithmetic is reliable. The data is your responsibility.

It will not account for variation. Demand of 95 orders per hour is treated as a flat figure. Real demand arrives in bursts. A step at 95.0 per cent load, like packing, will still form queues when arrivals cluster, even though it looks like it has headroom.

It will not handle processes that are not sequential, and it will not model steps that share people or equipment with other processes. It also will not tell you whether adding a parallel verification station is worth the cost. That is a business decision with inputs this tool never sees.

Finally, it will not implement anything. The Five Focusing Steps are a structure for assigning owners and tracking status. Three of the five actions in the sample case are still at planned. The tool records that honestly. Moving them forward is work that happens outside the screen.