Skip to content
Interface language: English
Design / Signal

Cognitive walkthrough: what a new user needs to know at each step

Analyse the knowledge and feedback needed at each step of a first-use task, then test the hypotheses with users.

Article language: English

Illustration: a woman in red and golden streaks of light in an open gallery.
In this article

The designer knows that the three dots under “Assets” open the photo upload and that a collection is made from uploaded files. A new person simply wants to send a client three photographs to choose from. The whole first use of the tool lies between those sentences, including several places where the interface author's knowledge can masquerade as the user's.

A cognitive walkthrough is an expert analysis of successive actions in a chosen task. The team asks what would enable a new person to know what to do and whether they had achieved the expected effect. It does not pretend to be a participant study. The team forms reasoned hypotheses about what a person might want or understand and about possible difficulties that can later be tested. It does not present those hypotheses as thoughts actually observed in particular participants.

Let's walk through a fictional tool called Capsule. A photographer has received access to a new library and needs to prepare a set of three photographs for a client. She understands the task but not this application's terms and rules. The analysis will use descriptions of several screens and their behaviour. None of the assessments below comes from a session with a user.

A user can know their profession without knowing your vocabulary

Record who is doing the task, what they bring to it and what they have not done before. The photographer knows her files and understands selecting photographs and sending a link. We do not assume she knows what the application calls a collection, whether creating a set shares it immediately, or where recipient management lives. Make these assumptions visible to every evaluator.

Do not write “beginner user” without explaining what they are new to. Someone opening Capsule for the first time is not necessarily using a computer for the first time. Experience with similar tools may help or create the wrong expectation. If “Save” normally preserves a draft in their work, do not assume without checking that they will recognise it as publication for a client here.

The CHI95 tutorial description by the method's authors presents the walkthrough as a way to find problems in learning a new interface. It notes that a design description can be analysed before implementation and empirical testing. That is useful for Capsule: we can refine decisions before inviting a photographer to a real trial.

Prepare a detailed starting state. The account already exists, eight photographs are on the computer, and the client has supplied an address and is waiting for three to choose from. If we analyse the stage after sign-in, record that boundary. Another analysis might start with the account invitation. Do not combine the cases into one story without showing the extra requirements.

The team also needs a correct path for completing the task. Write it separately: enter the library, add files, select three photographs, create a set and share it with the right recipient. Specify the actual effect of each action. If the authors do not themselves know whether saving creates a draft, a design decision is missing first; that is not yet a finding about user behaviour.

Choose one correct path and state its conditions. If the application allows adding files by button and by dragging, both routes may need separate analysis. Do not change paths halfway through merely to bypass an awkward step. Record the alternative and ask which users could access it. The goal is not to defend the shortest demonstration prepared by the author.

The screen description should include real labels and information shown after actions. A rectangle labelled “button” does not allow an evaluator to assess recognition of its purpose. A missing message after adding files makes it impossible to assess whether the person will notice completion. Flesh out the material only to the level needed for this task. A walkthrough does not require designing every application setting before the first discussion.

Four questions for one action

In their chapter on evaluation without users, Clayton Lewis and John Rieman describe four questions: about the desired effect, noticing the action, recognising its connection to the goal, and understanding feedback after doing it. They are applied to a known sequence of actions. We use them for our own example, rather than treating the list as an automatic pass form.

Consider adding photographs. First ask whether the photographer will want to achieve the effect of that action at this moment. If an empty panel says only “No assets”, it may not suggest that she needs to start by uploading files. This is not just about spotting a button; it is about recognising the necessary step towards sending the client a set.

Next check whether the action is noticeable in the view being analysed. The control may be outside the initial screen, behind an unrecognisable icon or in a section the person has no reason to visit. In Capsule, ask how the photographer would know that three dots open the action for adding material. “Because that is where they always are” needs an explanation of the experience supporting that assumption.

The third question concerns the link between the element found and the goal. She may see “Import” but expect it to mean transferring a library from another program. “Add photographs” might better describe the intended step. Do not declare a winner on team preference alone. Record the predicted interpretation, the source of the assumption and how it could be checked.

Finally assess feedback. After files are selected, the application closes the window but the list stays empty. Are the photographs still uploading, or did the operation fail? Can the action be retried without creating duplicates? Evaluation needs a description of the transition, not just the finished screen. The user needs to recognise what was achieved and which step now makes sense.

Aeneia conceptual photography

The first-use empty-state design (in Polish) article develops the particular case of a panel without material. A cognitive walkthrough covers the wider decision sequence, from recognising the absence to achieving a result. Do not stop at an attractive invitation to begin: the next screen may again require knowledge the previous one did not provide.

Check default values too. After eight photographs are uploaded, the entire collection might already be selected even though the photographer needs only three. Establish how she will recognise the current selection and whether the available description explains the scope of the next action. A visible “Create set” button does not by itself say what the set will contain. At this step the state matters, not only the button label.

A record of one barrier should be open to challenge

“Unintuitive import” is too general. For Capsule, make a more precise note: on the library screen, “Import” is intended to add individual photos; we assume a new person associates import with migration; we predict that they might overlook this button. The basis is a specific label and an explicit assumption about experience, not an alleged observation of the photographer.

Add a control question: what would change our assessment? Perhaps there is an explanation omitted from the material. Perhaps the person has used the convention before. Perhaps the term fits the vocabulary of the professional group being studied. If no possible evidence can change the opinion, the meeting becomes a discussion of preferences rather than an analysis of knowledge required for action.

Separate three kinds of result: a missing design decision, a potential barrier and knowledge whose origin needs checking. For missing feedback after upload, the team first has to describe the behaviour. For a misleading label, it might prepare an alternative. For an assumption about vocabulary, it may need contact with the intended audience. One label, “UX error”, blurs the differences.

After encountering a barrier, record the assumption used to analyse later steps. You may assume the first problem will be fixed so you can examine the next screen. That does not mean a user would actually reach it alone. In Capsule, further review of sharing may be conditional on “after the photos have been added correctly”. The report can then avoid burying an early block under a list of later positive assessments.

If several people assess a step, ask for separate reasons before group discussion. The point is not to count votes for a button. Differences may reveal two different pictures of the user: one person assumes cataloguing experience, another assumes a first encounter with a library. Agree which picture the current analysis concerns instead of averaging conflicting assumptions.

The interface author may explain the behaviour to the team, but that explanation is not evidence that the audience will learn it. If the hint is given only in the meeting, record the absence of an equivalent in the interface or the actual user preparation. Saying aloud “it is obvious” does not automatically reach a new account.

Tie priority to consequences. Failure to find file upload stops the whole task. An unclear set name may hinder only a later return. The number of challenged answers to the four questions is not an automatic measure of severity. One poorly explained sharing step can matter more than several minor section-label issues, especially when it changes who receives the material.

Completion must also be recognisable

In Capsule, selecting photographs might create a set that remains private. If the user must send it to the client, seeing thumbnails does not finish the task. Walk through choosing the recipient, the scope of access and confirmation of sharing. At each stage check the knowledge needed for the decision. You do not need to analyse all permission management to notice that draft and sent material are not distinguished.

Consider the first return too. The photographer closes the view and wants to find the prepared set. Which name will let her recognise the result? Does she return to the same library or another tab? This is a separate part of the task if the team has included it in expected use. Do not record success merely because a green confirmation appeared several screens earlier.

A first-steps guide can be part of the solution if it reaches the person before the action for which it is needed; the article on designing first-step instructions (in Polish) discusses how to make them available. Do not assume knowledge of help simply because it exists. Record where the person would find it and why they would consult it at that moment. If the task requires reading a manual beforehand, include the time and delivery of that knowledge in user preparation.

An assessment of the first five seconds of a site (in Polish) cannot answer questions about all those steps. It may address understanding the first view, but does not confirm recognition of the effect of sharing. Choose a test method suited to the uncertainty discovered by the analysis. Do not expect one method to answer questions about the whole journey.

Confront the hypotheses with a moderated usability test (in Polish) or observation of actual work (in Polish). Prepare the task around the photographer's goal without giving away tab names discovered during analysis. In the record, distinguish expert prediction from participant behaviour. If a person completes the task differently, that is material for updating your understanding, not a reason to persuade them to follow the predicted path.

A walkthrough should identify specific design decisions, their reasoning and how to check them next. Such a scope can be prepared with Aeneia Design. For Capsule, this means a journey where a new person can add their own photographs, make a set and recognise who can open it. The team knows what has not yet been checked and does not mistake that knowledge for a completed user study.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear