How We Find the Real Bottleneck

The loudest problem is rarely the constraint

Every business has a process everyone complains about. It is usually annoying rather than constraining — irritating to do, but not the thing limiting how much work gets through.

Improving a non-constraint produces almost nothing. You make a step faster and the work simply queues slightly longer somewhere else. This is the single most common way automation budgets get spent without a measurable result.

Look for the queue, not the complaint

The bottleneck is the step with work waiting in front of it. In a physical operation this is visible — a pile of paper, a full pallet bay. In an information business it hides in inboxes, in “awaiting approval” statuses and in the phrase “I am waiting on Dave”.

• Where does work sit longest between being ready and being done?
• Which step do people chase most often?
• Which single person's absence stops throughput entirely?
• Where does the phrase “we are waiting on” point?

If you want to find the constraint in a services business in ten minutes, read the last fifty internal messages containing the word “chasing”.

Measuring it without instrumentation

You rarely need a monitoring system. Two weeks of a shared spreadsheet with three columns — item, arrived, completed — gives you waiting time per step, which is all the analysis requires.

The result is frequently uncomfortable. In one client's quoting process the actual work took 40 minutes and the average elapsed time was nine days. Every hour of that gap was waiting: for information, for approval, for someone to pick it up.

Three bottlenecks we find repeatedly

1. One person who approves everything. Usually a founder or a technical lead. Fixing it is an authority problem, not a software problem, though software can make delegation safe by showing what was approved and why.
2. Waiting for information from outside. Customers, suppliers, partners. Automation helps a lot here — structured requests, automated chasing, clear status — because the constraint is the round trip rather than the work.
3. Batching. Work that could flow continuously is done once a week because that is when someone sits down to it. Often the cheapest fix in existence: change the cadence, no software required.

Fixing it moves it — and that is the point

Relieve a constraint and a new one appears. This alarms people, and it should not. Each move means throughput increased until something else became the limit; that is what progress looks like.

What matters is knowing it will happen, so the second project is planned rather than experienced as a disappointment. We usually map two constraints ahead for exactly this reason.

When the bottleneck is not automatable

Sometimes the constraint is a skilled person doing skilled work, and the honest answer is that the fix is hiring, not software. We say so.

Even then software often helps around the edges — making sure the expensive person only sees work that is genuinely ready for them, with everything prepared. Protecting a constraint's time is frequently worth more than trying to replace it.

Frequently asked questions