Skip to case study
← Back to projects

Current work in progressUpdated

Cosmonauts: designing and testing the sponsor experience

My work at Cosmonauts spans research and early design for an AI-supported tool that helps sponsors decide who to meet at events.

I investigate sponsors’ working practices and how people understand and review the tool’s proposals. Alongside the design work, I built an unmoderated testing environment so I could iterate the prototype and study together.

My contribution: product review, sponsor discovery, prototype research and experience design.

Keeping the study and design moving together

I needed to develop and test the prototype while continuing to work on the design. Building the study around it let me revise the experience and the research activities together.

The environment brought task prompts, response capture and interaction records into the prototype journey. A separate reader made returned sessions available for review and export.

Explore the testing environment

The walkthrough follows a participant task through to the evidence available in my results reader.

Illustrative walkthrough of the environment I created. Simplified screens, fictional content and sample responses. This demonstration collects no research data.

Participant viewSample activity
Your task

Review this proposed audience. What information do you think it is based on?

Proposed audience

People working in product design

Role
Product designer
Organisation
Design studio
Inspect source

The uploaded brief describes product designers working in design studios.

Keep the task beside the experience.

The study presents a scenario and a focused activity around the prototype. Participants can refer to the task while exploring the relevant screens.

What I designed
The study structure, task prompts and activity boundaries.
Response captureSample session

What was the proposal based on?

The uploaded brief

How confident are you?

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
Recorded actionOpened “Inspect source”

Bring actions and answers together.

The environment records interactions alongside answers, confidence ratings and optional comments. These give different views of a session.

What I could examine
What someone opened or selected, what they reported and how confident they felt.
Session handoverIllustrative sequence

Session ready to return

  • Answers
  • Recorded actions
  • Timings
Results codeDEMO · SESSION · EXAMPLE
Participant sends the codeI import it into the reader

A file download was also available as a fallback.

Make the handover explicit.

At the end, participants received a results code to send back. I imported returned codes into a separate reader.

A boundary I kept visible
Results were available to me once returned. They were not streamed to a live dashboard.
Researcher viewFictional session

Session reader

Example imported
Recorded action
Opened “Inspect source”
Answer
The uploaded brief
Confidence
4 out of 5
Separate CSV exports for answers and recorded actions

Give analysis a traceable starting point.

The reader brings returned sessions together for review and export. I could compare responses with the recorded sequence of actions.

My role in the analysis
Interpret the record in context. Opening an explanation alone does not establish that someone understood it.

Following a research question into the next design decision

One focus was how people understood an AI-generated proposal: what they thought it was based on, how they checked it and what they would change.

In the study, people gave high confidence ratings even when errors deliberately included in the research prototype went uncorrected. I treated that as a reason to examine understanding more closely when developing the design and the next study.

Study context: six colleague-proxy sessions on fixed example data, including a repeat run and uneven screen exposure. These were exploratory sessions, not validation with sponsor customers.

The research question
Could people identify the information behind a proposal and review it accurately?
My interpretation
Confidence and approval needed to be considered alongside answers and recorded actions.
Direction taken forward
The team took forward a source summary and explanations beside criteria, alongside my revised testing approach.
How I refined my own testing

I reviewed the task wording, completeness of session records and supported interactions in my early research prototype. This helped me separate questions about the design from limitations of my study setup.

I narrowed the next study from a broad prototype journey to reviewing an AI proposal. A clean version and a version with deliberately introduced errors gave me a focused way to examine what people noticed and acted on.

I could not establish why some earlier records ended or assume that every record represented a separate participant. In the focused study, only one person supplied comments. Those limits shaped my interpretation.

How this informed “learn as you use”

I developed a specific design hypothesis: put an explanation beside the decision it supports, so people can inspect the reasoning while reviewing a proposal.

The source summary and explanations beside criteria give that direction a concrete form. Their effectiveness still needs testing; the study did not establish a reduction in cognitive load.

The next research question

Can people explain the information behind a proposal in their own words, then review and change it without help?

My plan separates moderated exploration from a short unmoderated test of a specific explanation. This is the next planned cycle of research.

Keeping sponsors’ working practices in view

Alongside the prototype research, I scoped and carried out sponsor discovery in a live event setting, focusing on decisions and working routines.

This provided context for early experience design. I kept participant observations separate from my interpretation and used a fieldwork logger to support structured recording.

How the fieldwork logger supported discovery

I developed a logger with separate fields for direct evidence, participant confirmation and researcher interpretation. It supported recording conversations and observations in the event setting.

The fieldwork logger and the unmoderated testing environment served different parts of my work: discovery into working practices, and evaluation of the developing experience.

How I organised the product review

I reviewed the commercial concept, product documentation and existing experience to understand the context. I used Fix / Ask / Research to organise my observations by the next step each called for.

Fix
An actionable change.
Ask
A question for discussion with the team.
Research
A question requiring evidence about people’s behaviour.

My contribution so far

I have combined sponsor discovery, product review and prototype research with experience design, and built tools to support fieldwork and unmoderated testing.

The team has taken forward the source summary, explanations beside criteria and revised testing approach.

Completed work
Discovery, product review, colleague-proxy research and the testing environment.
Direction taken forward
Explanations at the point of decision and a more focused testing approach.
Next in my prototype work
Develop alternative and error paths, then evaluate the experience with sponsor customers.

This is ongoing research and design work. No measured product improvement is claimed.

Return to the walkthrough