# Usability Testing Basics

> Watch real people try real tasks and see where they get stuck. Tests with a host and without one, talking out loud, and why five people is often enough.

**Course:** [UX Design and Research](https://learn.benibix.com/learn/design-user-experience)  
**Time:** 12 min

**After this lesson you can:**

- Plan a short usability test
- Choose moderated or unmoderated
- Run a think-aloud session

## How to watch people use a product

Usability testing means watching real people try real tasks to see where they get stuck. You give them a task, such as find the return policy, and you watch. A well known study found that five users reveal about 85 in 100 of the problems in one round. Small tests, run often, beat one big test.

A moderated test is led live by a host. The host sits beside the user, or joins a video call, and can ask why did you click there? It catches the reasons behind a stumble. It takes more time per person, so it suits early designs and small groups.

An unmoderated test is one people do on their own. A tool gives them the tasks and records the screen. No host is there to help or to ask. You can run 50 people overnight, but you cannot ask a follow-up question when something odd happens.

The think-aloud protocol asks people to say what they are thinking as they go. I am looking for returns. Maybe it is under Help. No, that is about orders. Those words show the mental model at work, which a click log alone never shows.

Write tasks as goals, not as steps. Find out how much it costs to send a parcel to Spain works. Click the Shipping menu gives the answer away. And never help. When the user is stuck, the silence is the finding.

Five people is enough to find the big problems, not to measure. If four of five fail a task, you have a problem worth fixing. You do not have a success rate you can quote. Numbers need more people, and a later lesson covers them.

The common mistake is to test with colleagues. They know the product, the words and the shortcuts. They will sail through tasks that real users fail. Test with people who match your users, even if that means only three.

![Two scenes side by side. On the left, a host sits beside a user at a laptop while the user talks out loud, a speech bubble reading: where would returns be? On the right, a user sits alone at a laptop with a small recording dot, and a task card on screen reads: find the return policy.](https://learn.benibix.com/diagrams/des3-test-kinds.svg)

## Which kind of test to run

1. **Moderated.** Pick it for an early design, when you need to ask why people got stuck.
2. **Unmoderated.** Pick it for a finished flow, when you need many people quickly and the tasks are clear.

### Example: Five people find the hidden button

A travel site tested its new booking page with 5 people. Each one said out loud what they were looking for.

Four of the five could not find the button to add a bag, because it was labeled extras. Renaming it to add a bag took one hour.

**Words in this lesson:** [Usability Testing](https://learn.benibix.com/term/design/usability-testing), [Moderated Testing](https://learn.benibix.com/term/design/moderated-testing), [Unmoderated Testing](https://learn.benibix.com/term/design/unmoderated-testing), [Think-Aloud Protocol](https://learn.benibix.com/term/design/think-aloud-protocol)

> **The one difference.** A moderated test has a host who can ask why. An unmoderated test has no host, so it reaches more people but cannot ask anything.

**Question:** What should you do when a test user gets stuck?

**Answer:** Stay quiet and note where they are stuck. The stuck moment is what the test is for. Helping hides the problem, and real users will have nobody beside them.

---

Source: https://learn.benibix.com/learn/design-user-experience/usability-testing
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-19
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
