# How to find bottlenecks in a growing business

Source: https://leverops.co/resources/how-to-find-bottlenecks/
Last updated: 2026-10-11

**A bottleneck is the step that limits how quickly or reliably work can move through a process.** It may be a person, a decision, missing information, or a system handoff. To find one, follow actual work from request to completion and look for the longest waits, repeated corrections, and overloaded decision points.

## Start with one process, not the whole company

Pick a process that matters: a lead becoming a customer, an order reaching fulfillment, a customer request getting resolved, or a weekly report getting produced. Choose a recent example that went smoothly and one that did not. Ask the people involved to show you where the work went, what information they needed, and when they waited.

The official process may say requests move directly from sales to operations. The real process may include an email to the owner, a private spreadsheet, and a second request for missing details. Those hidden steps are where the useful evidence lives.

## Record movement, waiting, and rework

For each step, note who owns it, what triggers it, how long the work itself takes, how long it waits, and what sends it backward. You do not need perfect time tracking to start. A sample of ten recent cases can reveal whether the same delay repeats.

<div class="table-wrap" tabindex="0" role="region" aria-label="Table">

| Observation | What it may mean | First question |
|---|---|---|
| Work waits for one approver | Decision rules are unclear or too centralized | Which cases truly need that person? |
| Details are requested twice | Intake or handoff is incomplete | Where should the detail be captured once? |
| A report is rebuilt manually | Data is scattered or the report is poorly defined | Which decision does the report support? |
| Many cases return for correction | The process allows missing or conflicting inputs | What check belongs earlier? |

</div>

Look at frequency as well as delay. A rare 20-hour exception may matter less than a ten-minute duplicate entry repeated on every order.

## Rank the problem before choosing a tool

Estimate the annual time, cost, customer effect, or missed opportunity tied to each repeated issue. Then consider the cost and risk of changing it. A small step that affects every customer may deserve attention before a dramatic but unusual failure.

Ask whether the step should exist at all. If not, remove it. If the rule is unclear, clarify it. If the same predictable action happens repeatedly, [automation](/services/workflow-automation/) may help. If a person copies information between systems, consider [integration](/services/systems-integration/). AI is relevant only when the information or judgment requires it.

## Check whether the fix worked

Write down a baseline before changing the process: time to complete, waiting time, corrections, owner escalations, or customer response time. After the change, measure the same thing for enough cases to see whether it improved. Ask the team whether the new path is easier to use. A change that moves work to another department is not a real improvement.

For a financial starting point, try the [manual work cost calculator](/tools/manual-work-cost-calculator/). If the bottleneck is still hard to locate, [process improvement](/services/process-improvement/) begins with observing the work, not prescribing software. [Book a free operations review](/book/) to discuss one process.