Best Match
Make ranking visible.
The first result shows a recommendation reason, transcript evidence, title, summary, and timecode so the student can judge why this moment comes first.
02 / Project
Course-recording retrieval tool
CuePoint
Turns a vague memory of a technical lesson into a verified, timecoded lecture moment.
- Timeline / scope
- 10 weeks · solo, one-course MVP
- My Role
- Solo Product Designer
- Outcome
- Tested functional MVP
My contribution
Problem framing, one-course MVP scope, search and clarification logic, the six-entity model, interface design, and two-stage testing.
Tools & workflow
Figma for the Alpha prototype → Lovable for the functional MVP; Google Sheets for the data model, GitHub for implementation review, and Vimeo for timecoded playback.
Prepared examples and simulated matching; no live API, transcript pipeline, or production backend.
Interaction premise
A small answer can hide inside hours of video.
Students often remember the task an instructor demonstrated, but not the lecture title, wording, or timestamp. CuePoint begins with that natural memory and reduces the cost of returning to the exact teaching moment.
A timestamp is not enough. Users need evidence.
Primary interaction path
Best Match → Moment Detail → Timecode
CuePoint explains why a result matches before the user spends time watching.
Moment Detail
Confirm before committing attention.
The detail view connects the selected moment to its key step, transcript context, match reason, and preview. It is the confirmation layer between the search result and the video.
Original recording
Return at the selected timecode.
The final handoff opens the original course video at a prepared Vimeo timecode, preserving the instructor's demonstration instead of replacing it with a detached text answer.
Behavior behind the screens
Specific questions move. Broad questions pause once.
Specific questions should move directly to a result.
Clarify only when multiple intents could produce different moments.
Admit when the prepared course data cannot support a reliable answer.
System evidence
The segment is the bridge.
Six entities connect a natural question to a saved instructional moment: Course, Recording, Segment, SearchQuery, SearchResult, and SavedResult. The segment turns a long video into a retrievable unit without losing its source.
Explore the full data model
Course contains RecordingsRecording contains SegmentsSearchQuery produces SearchResultsSearchResult references SegmentSavedResult preserves the chosen Segment and context
Research & iteration
Test the structure, then the working behavior.
The Figma Alpha tested whether students could understand and compare results. The functional MVP tested when to clarify, what evidence a moment needed, and whether save, reset, and no-match states were clear.
View testing details
Figma Alpha
Could people understand the structure?
Testing exposed result cards that looked too similar, leading to a clearer Best Match and stronger evidence hierarchy.
Functional MVP
Could the decision logic hold together?
The second stage tested conditional clarification, Moment Detail evidence, save feedback, reset behavior, and no-match handling.
| Before / finding | Design response |
|---|---|
| Original action label: “Open moment.” | Renamed it “View moment details.” |
| Moment Detail needed more evidence. | Expanded the transcript context. |
| Save confirmation and reset needed clearer feedback. | Strengthened saved confirmation, added removal, and renamed “Refine again” to “Start over.” |
Ownership
An end-to-end product arc, centered on retrieval trust.
I owned problem framing, scope, system logic, information architecture, interaction design, the six-entity data model, two-stage testing, iteration, and final product communication. AI-assisted tools accelerated prototyping; product decisions and final direction remained mine.