ACADEMIC PROJECT / HUMAN FACTORS

Discord Conversation Search Redesign

Finding past conversations with the details you remember.

A guided search concept for Discord users who remember a person, a server, or a topic—but not the exact wording of a message.

My role

UX researcher & interaction designer

Context

Academic human factors project

Deliverable

An iterated interface prototype

Independent academic project; not affiliated with Discord. Informed by early needfinding, a three-participant concept evaluation, and a four-participant usability test of the high-fidelity prototype.

01 / PROBLEM & USER NEEDS

The Task

People who communicate across multiple servers and direct messages may remember who they spoke with, where a conversation happened, or what it was about. Locating the original messages can still require unnecessary cognitive effort.

Target users

Discord users who frequently communicate across servers and DMs and need to retrieve information from older conversations.

Design opportunity

Let users narrow a search through partial memories instead of relying exclusively on exact keywords.

THE DESIGN QUESTION

How might search work with what people remember, rather than asking them to reconstruct what they typed?

02 / RESEARCH & KEY INSIGHTS

Research & key insights

At Project Check-In 2, I had completed two interviews and think-aloud navigation sessions with frequent Discord mobile users, alongside a heuristic evaluation. Participants discussed their usual habits, then completed tasks on their phones: switching between DMs and servers, recovering older conversations and mentions, returning to threads, finding unread activity, and locating channels.

Familiar navigation was not the main difficulty

Both participants moved between servers, channels, and DMs with little difficulty. This shifted the early design focus away from rebuilding basic navigation and toward recovering information.

Recovering older conversations depended on search

One participant described older messages as taking “forever” to locate without the right keywords or context. The other found mobile search easier than expected, but struggled and scrolled extensively when asked to recover a conversation without it.

Hidden threads and unread volume weakened context

One participant expected threads to be accessible more directly from a channel rather than through several menus. The other said they would not realistically scroll through hundreds of unread messages. The heuristic evaluation also highlighted limited visibility of location and conversation context.

Early evidence, not a validation result: this check-in reports two experienced participants. Additional sessions and a wider range of experience levels were planned. These observations informed design directions; they do not establish that the redesigned prototype is more effective.

Project Check-In 2, page 1
Project Check-In 2, page 2
Project Check-In 2, page 3
Project Check-In 2, page 4

03 / HUMAN FACTORS

Human factors

I used human factors principles to frame the alternative search experience. These principles explain the design intent; they do not establish that the redesign is more effective.

Recognition versus recall

Use recognizable categories such as people and servers to support partial memories, rather than asking users to recall exact message wording.

Cognitive load

Guide the narrowing process and reduce redundant explanatory text so attention stays on the search task.

Mental models & interface clarity

Use familiar Discord colors and conventions while communicating filter requirements and search coverage consistently.

04 / DESIGN PROCESS & ITERATIONS

Design process & iterations

Three early design directions

The check-in proposed alternatives grounded in the needfinding, rather than assuming Discord’s entire navigation needed redesigning. These were exploratory directions at that stage—not three tested concepts or a documented winning concept.

01 / Recent and Important Conversations Hub

Bring recent conversations, unread activity, mentions, and recently visited threads into one mobile view. Intended benefit: reduce the need to remember whether a conversation lived in a DM, server, or thread.

02 / Improved Thread Navigation

Surface active threads directly within parent channels through visible indicators or expandable sections. Intended benefit: reduce the extra menu steps observed when returning to a thread.

03 / Improved Message Recovery and Catch-Up

Explore clearer unread markers, easier access to mentions and replies, or a summarized catch-up view. Intended benefit: help people recover important activity without scrolling through hundreds of messages.

The paper prototypes developed the hub and catch-up directions, while introducing a context-based conversation finder in place of a standalone thread-navigation concept. The following artifacts show the sketches, the evaluation plan, and the subsequent three-participant concept evaluation.

From initial alternatives to paper prototypes

01 / Recent and Important Conversations Hub

Annotated original sketch: 01 / Recent and Important Conversations Hub

A unified view of DMs, channels, mentions, and threads. The intended test: can people return to important conversations without remembering exactly where they were?

02 / Context-Based Conversation Finder

Annotated original sketch: 02 / Context-Based Conversation Finder

Filters for person, topic, date, and server turn partial memories into search criteria. The intended test: can context-based retrieval help when exact wording or location is forgotten?

03 / Message Recovery and Catch-Up

Annotated original sketch: 03 / Message Recovery and Catch-Up

Separate mentions, replies, and unread activity into clear categories. The intended test: does organizing activity help people identify what needs attention?

Planning how to compare the concepts

The evaluation plan proposed approximately 3–5 Discord mobile users with varied usage frequency. Each would think aloud through a scenario for every paper prototype: recover an important conversation, search using partial context, or identify unread activity requiring attention.

Planned observations included task success, mistakes, hesitation or backtracking, approximate steps, ease-of-use ratings from 1–5, and preference rankings. Follow-up questions would probe expectations, confusion, usefulness, missing features, and desired changes. The analysis would guide which features to keep, change, combine, or remove.

The original pages above document the evaluation plan. The subsequent check-in below reports the completed qualitative evaluation with three participants. Select any artifact to read it full size.

Original evaluation plan: participants, procedure, and qualitative questions
Original evaluation plan: quantitative measures and analysis

What I learned from testing

I walked three participants through the paper prototypes and asked what made sense, what felt confusing, and which ideas they would use.

Search by what you remember

All three saw value in searching by person, topic, date, or server instead of an exact keyword. The hub was easiest to understand, so I kept its familiar organization.

Make the rules clear

Participants wanted to know what counted as “important,” which filters were required, and whether search included both DMs and servers.

I moved forward with contextual search inside a shared hub for recent conversations, mentions, and catch-up activity. This feedback helped me choose a direction; I still needed to test the combined flow.

Read original evaluation notes

The proposed search interface lets users narrow results using remembered people, servers, and contextual clues. I developed a prototype, then used feedback to refine its instructions and visual language.

01 / Align the instructions with the requirement

What changed

Reconsidered wording that described all filters as optional.

Why I made the decision

A search that requires at least one filter needs instructions that match that behavior.

02 / Remove unnecessary explanation

What changed

Removed the sentence telling users to enter only what they remembered.

Why I made the decision

The guided interface should communicate how to begin without another instruction competing for attention.

03 / Evaluate how search coverage is explained

What changed

Revisited wording about DMs and servers the user can access.

Why I made the decision

Users should understand the scope of the search without the interface implying access beyond their permissions.

04 / Refine the visual language

What changed

Adjusted the visual design to align more closely with Discord’s colors and conventions.

Why I made the decision

The alternative search experience should feel connected to the product people already know.

05 / LOW-FIDELITY WIREFRAMES

Original low-fidelity wireframes

01 — Main Hub

Original Discord conversation recovery wireframe: 01 — Main Hub

02 — Important

Original Discord conversation recovery wireframe: 02 — Important

03 — Context Search

Original Discord conversation recovery wireframe: 03 — Context Search

04 — Search Results

Original Discord conversation recovery wireframe: 04 — Search Results

05 — Recovered Conversation

Original Discord conversation recovery wireframe: 05 — Recovered Conversation

Original five-screen exploration, from the main hub to a recovered conversation. Select any screen to view it full size.

06 / FINAL PROTOTYPE

A search flow built around remembered details.

The final prototype brought the conversation hub and contextual search together. People could narrow a search using the details they remembered, then return to the conversation around a matching message.

01 / Main hub

Final post-testing Discord design: 01 / Main Hub

02 / Important

Final post-testing Discord design: 02 / Mentions

03 / Context search

Final post-testing Discord design: 03A / Empty Search

04 / Search results

Final post-testing Discord design: 03B / Person Selected

05 / Recovered conversation

Final post-testing Discord design: 04 / Closest Matches

The high-fidelity prototype used in usability testing. Select any screen to view the original 2× export at full size.

07 / USABILITY TESTING

Usability testing

I tested the prototype with four participants to see how they found old conversations, used the filters, and understood Recents, Mentions, and Catch Up. I also asked whether search across DMs and servers was clear and what felt confusing or repetitive.

Key findings

Keep the cross-conversation search

Participants valued finding conversations across DMs and servers using the details they remembered.

Catch Up meant something else

Several expected unread notifications or missed activity—not another way into search.

One way into search was enough

The search bar, shortcuts, and Catch Up felt repetitive. The dedicated search screen was clearer.

Explain the labels

Recents, Topic, and the filter instructions needed clearer wording. DM and server results also needed more context.

08 / POST-TESTING REFINEMENTS

Refining the design after testing

After testing the final prototype with four participants, I simplified the navigation and made the search flow clearer. These six screens show the changes I made in response.

Before testing

Discord main screen before testing

Catch Up and filter shortcuts competed with the search bar.

After testing

Discord main screen after testing

One search entry point, with Recents and Mentions kept separate.

01 / Main Hub

Final post-testing Discord design: 01 / Main Hub

I removed Catch Up and the filter shortcuts. Recents now explains that it shows conversations recently viewed or participated in.

02 / Mentions

Final post-testing Discord design: 02 / Mentions

Mentions now explains that it contains messages where someone mentioned or replied to you.

03A / Empty Search

Final post-testing Discord design: 03A / Empty Search

I moved all filters into Find a Conversation, renamed Topic to “Topic or keyword,” and added examples and guidance to start with any remembered detail.

03B / Person Selected

Final post-testing Discord design: 03B / Person Selected

The selected-person state shows how one detail can enable search without filling every field.

04 / Closest Matches

Final post-testing Discord design: 04 / Closest Matches

Results use closest-match language and make the DM or server context easier to distinguish.

05 / Recovered Conversation

Final post-testing Discord design: 05 / Recovered Conversation

A highlighted matching message confirms where the search has taken the user.

Final post-testing screens. Select any screen to view the original 2× export at full size. This revised version has not yet been retested.

09 / OUTCOMES & REFLECTIONS

Outcome

Completed work

Early research, three paper concepts, a three-participant concept evaluation, and a four-participant usability test led to the final revised screens above.

What the testing showed

The search concept was useful to participants, but the navigation added unnecessary complexity. I simplified it around one search flow, with Recents and Mentions as separate ways to return to conversations.

What I would test next

I’d retest the revised flow to see whether people can start a search without help, understand the filters, and distinguish DM results from server activity.

What I’m taking forward

The biggest lesson was that a useful feature can still be hard to find. I don’t need more search options on the home screen; I need one clear way in, with the right guidance once someone starts.