Testing in user-centred design
In user-centred design, testing is an activity that you do with users to check how they use, perceive, or interact with something.
This might be a design concept, content, journey, or something else that you want to gain user insight on.
This guide is also available as a presentation:
When you should test
Testing is important throughout the life cycle of a product or service. This means testing a product or service from early-stage ideas to through to the live service which is available to the public.
The way people use and interact with a service is constantly changing. This means it needs continuous maintenance and feedback to make sure it:
- is up to date
- meets users’ needs (these might change over time)
- is efficient
- does what it was set up to do
Be clear about why you’re testing
Before you test something, you will need to define what you need to learn. This might come from a problem, hypothesis, or assumption.
Testing is about finding the best ways to solve a problem. It is not for trying to find evidence to match what you would like a user to say.
Usability testing helps you to understand how people use an existing product or service.
This method is most useful for:
- discovering problems in a design
- finding opportunities to improve a design
- learning about user behaviours and perceptions
Conduct usability testing
When running usability testing, you should:
- recruit people who know nothing, or very little, about your service or service constraints to get a fresh and unbiased view
- find participants with diverse experiences, backgrounds, knowledge, and abilities
- remain neutral and avoid sharing your personal opinions or how involved you were in the design
- always observe what the user does, it might be different to what they say
Chris Sutton, a user researcher who previously worked at the Welsh Local Government Association, wrote this guide for conducting a one-hour usability testing session.
Testing new ideas or concepts
Prototypes are mock-ups of an idea. They allow you to test the idea before making it live to the public.
Prototypes are not supposed to be perfect. They can be quick and rough, but they can communicate the idea well enough for you to get feedback from users.
When to prototype
You can prototype something when it’s not clear what the product requirements are, or they change rapidly.
You might need to create a prototype if you are testing an idea or concept that doesn’t exist yet.
How to prototype
You can make a prototype for
- a physical product
- a website or app
- a service
A physical product can be tested on cardboard. A website or app can be tested through a hand-drawn wireframe. A service interaction can be tested with roleplay.
It is best to start testing ideas in low-fidelity first to make sure you don’t spend lots of time building something that doesn’t meet users’ needs.
- use paper to take your team through the steps of using your app with some basic content
- make a basic digital version and test it with users
- build it into a high-fidelity version with branding and tested content, which you can have confidence in
The image below shows different levels of fidelity in prototypes to test with users and get their feedback.
When working with prototypes, you should always:
- create all components such as screens, pages, or designs. and include the content
- make sure the components fit together and the narrative for the prototype is seamless from a user perspective
- make it consistent: any mistakes or missing parts remind users that they are looking at a fake product or service
- research and collect assets such as existing content, image libraries, or similar examples to help shape the prototype
- write the interview script to conduct interviews about the prototype
- complete a trial run and go through the prototype to find and fix mistakes
Tools for prototyping
If you can access specific design software, Figma, Sketch and Adobe XD are tools that you could consider.
If you do not have access to paid-for design tools, use tools such as:
You could also use website building tools such as Squarespace, Wix, WordPress or Marvel. These tools are great for fast prototyping and testing content in a browser.
Paper-based products could include letters, brochures, flyers, reports, or marketing materials.
To prototype these, use word processing software such as Microsoft Word or Google Docs.
You can also use Canva, design software, or a presentation tool.
Always print a paper-based product to see how it would look when printed.
If you are prototyping a service, it’s a good idea to prototype with a script or a storyboard.
You can work with physical objects or use digital tools to prototype the marketing for a product.
For example, if you are working on the design of a new office, you may find using physical models useful.
You could also prototype with pictures of how the office might look and use this to gather feedback.
When prototyping physical objects, you can also change an existing object. We’ve found children’s toys, such as Lego or playdough, to be useful when creating these kinds of prototypes.
Learn more about prototyping
- The Centre for Digital Public Services’ Learn by making team created this prototype kit for labs
- GOV.UK Prototype kit
- Course: Design Thinking – The Beginner’s Guide
- Bill Buxton, What Sketches (and Prototypes) Are and Are Not (pdf)
Testing methods for something that already exists
Card sorting is a great way to understand how users group content. This method can help structure a website.
How to card sort
Card sorting can be done using physical cards or cards on a screen (such as in Mural or Miro).
Each card is labelled with a concept or word and the participant is asked to organise the cards in a way that makes sense to them.
When testing the structure of your website. If most participants place the card labelled “hours of operations” under “services,” rather than under “locations,” then it makes sense to design the site with hours as a subheading under services.
Learn more about card sorting
- Card Sorting: How Many Users to Test by Nielsen Norman Group
Tree testing – also called tree-jacking – is a great way to define the hierarchical structure of information.
This method of testing can be useful to:
- see where people may get lost during a journey
- understand the best way to help people complete their task in the easiest way possible
You could use it to test a whole website or help you to understand the information architecture of a service.
How to do tree testing
Tree testing involves:
- showing someone a starting point to complete a task
- asking them where they would click to help them reach their goal
- track the journey the person would choose to travel, from the first page to complete the task
Learn more about tree testing
- Tree Testing guide by the Nielsen Norman Group
- What is tree testing and how can it improve your site’s usability? on UserZoom.com
- Tree Testing on userinterviews.com
Wizard of Oz testing
The Wizard of Oz method involves interacting with a mock interface which is controlled behind the scenes by a human.
You can use it to test costly concepts inexpensively and to define a problem.
A/B testing (sometimes called split testing) can be an experimental way to test two ways of completing a task or viewing information.
You can show version A of a design to half of your audience and show version B to the other half.
This allows you to test and measure which design idea is most effective.
Learn more about A/B testing
- What is A/B Testing? A Beginner’s Guide by Neil Patel
- A Refresher on A/B Testing on the Harvard Business Review
- What is A/B Testing? The Complete Guide: From Beginner to Pro on CXL
- Complete Guide to A/B Testing Design, Implementation and Pitfalls on Towards Data Science
- A/B Testing Guide on VWO
Highlighter testing for content
Test content with highlighter testing. It is a quick and easy way to get feedback on your content and how it makes people feel.
How to do highlighter testing
To test the content, you give a piece of content to a group of users and ask them to highlight the text.
You can use different colours of highlighter to understand different emotions, for example:
- blue could be used to underline content that makes them feel confident about the service
- pink could be used to show words or sentences that were confusing or misleading
When you have collected all the responses, you have a clear idea of how the content has made people feel and should have some clear themes and areas to work on.
This type of testing can be done online using a tool such as Microsoft Word or Google Docs.
It is also effective when the page is printed, and people complete the task physically.
Learn more about highlighter testing
- Designing remote content testing by Lexie Claridge
- 3 Effective Methods for Content Tests (Beyond Usability Testing) on dscout
A heuristics evaluation is when a digital expert uses knowledge of best practices to evaluate a product or service.
This can be useful to learn about well-known issues, for example:
- content issues (use of “click here” and “read more”)
- accessibility problems (such as checking for colour contrast) using a tool that automatically picks up issues for devices or browsers
Only use this type of testing if you have the right experts to guide and complete testing
Be mindful that this testing is never as good as testing with real people.
Check-box testing using heuristics only gives you some feedback on the interface and content. It will not give you insight into the overall user experience.
Learn more about heuristics evaluations
- Heuristic Evaluation by the Interaction Design Foundation
- How to Conduct a Heuristic Evaluation by the Nielsen Norman Group
- 10 Usability Heuristics for User Interface Design by Nielsen Norman Group
- Your Step-by-Step Guide to Heuristic Evaluation in UX Design by Career Foundry
- Heuristic Evaluation on Wikipedia
Using quantitative data to see how your product or service is performing is a good way to learn about the behaviours, issues, or pain points of people using your digital services.
How to use do an analytics review
When using analytics for testing, it is important to consider contextual information to support the data.
For example, a bounce rate shows how many people come to your website and ‘bounce’ back off again, without clicking any other pages on your site.
A high bounce rate, meaning lots of people only visit one page of your website could be good or bad, depending on the context and goal.
- A high bounce rate could be very good if you have a single page for people to complete a specific task that people have found directly.
- If many people are coming to your website and leaving without finding what they are looking for, or going elsewhere for similar services, that could be a metric to track and understand the problems people are facing.
For public services, these could be services that are highly competitive such as leisure centre memberships or fostering.
Tools for doing analytics reviews
With tools such as Google Analytics, and the many other analytics software packages available, you can learn about:
- the user journeys people take on your site
- how long people may spend on each page or component
- when people leave your site
- other useful pieces of data to help you understand user behaviour
Nielsen Norman Group has published advice and guidance about three ways to use analytics in user-centred design.
A crit is when a group of people get together to talk about, share and improve a product.
They are used to:
- help teams review their own work
- create consistency
- create or iterate a style guide
- build a team
- improve the product or service
Rules of a crit session
- Remember everyone did the best job they could with the information they had at the time.
- Never talk about the person, only the work.
- Be kind and honest.
- Only share constructive criticism, so if it cannot be changed, move on.
- Nobody should ever feel like they need to defend something.
- It is ok to stop at any time.
Learn more about crits
- Content crits: what they are and how to run them on Content Design London
- Content crits: they’re not scary! on the GDS blog