A learning platform that keeps instructors in the loop
A learning management platform that makes course delivery, progress tracking and instructor workflows simple for learners and administrators alike.
The shape of the work
- Industry
- Professional Services
- Duration
- 18 weeks
- Cooperation model
- Fixed price, phased
- Services
- Product designFrontend & backend buildAnalytics
- Integrations
- HubSpotDocuSignXeroGoogle Workspace
- Technologies
- ReactNext.jsNode.jsPostgreSQLTailwind CSSWebSockets
- Team
- 1 Project lead1 Product designer2 Frontend engineers1 Backend engineer1 QA engineer
Client name withheld under NDA. Engagement details are shown to the extent our agreement permits.
What we were brought in to do
Course content, grading, and progress data lived across disconnected tools. We consolidated enrollment, lessons, assessments, and instructor tools into one coherent experience.
A vocational provider running eighty courses across three campuses, with completion rates that had been drifting down for four years and no way to see it happening until final assessment. Instructors were tracking at-risk students in personal spreadsheets, which meant the knowledge left when they did. The engagement was scoped around one question the provider couldn't answer: who is about to drop out?
Product Design & Build
Where the old way broke
Instructors couldn't see where students stood or intervene early, and learners bounced between multiple disconnected tools to find their next lesson.
Assignments lived in one tool, peer review in another, and the gradebook in a spreadsheet reconciled at end of term, so a student's real position was never visible in one place during the term, only after it. The lesson player was a video embed on a page with a sidebar of links, and session recordings showed students leaving for the quiz screen and not coming back.
We unified course delivery, progress tracking, and grading into a single platform, with clear dashboards for instructors and a distraction-free learning view for students.
Instructor Intervention Cockpit
Surfaces struggling students in week two, long before final assessment.
Risk is computed from behavior (session gaps, abandoned lesson players, quiz retries, time to first submission), which is why it fires in week two instead of at final assessment. Grades don't exist yet in week two. The cockpit ranks a cohort by who is drifting, not by who is currently failing, and every name opens the evidence behind its own flag, so an instructor can judge the signal instead of taking it on trust.
- Behavioral signals, not grades, drive the flag
- Cohort ranked by drift, not current mark
- Evidence behind every flag, one click away
The instructor cockpit for the Spring intake, ranked by drift instead of mark. The open learner is 72% through the course and watched lesson 4.5 three times without moving on, where nine others also stalled, as module 4 drops from 28 reached to 9 completed.
Focused Lesson Player
The lesson player: module contents on the left, the transcript under the stage, and an inline check that pauses playback right after the point being made, with video, transcript and check held as one saved position.
Distraction-free learning experience with auto-resume, transcript sync, and inline quiz prompts.
The player keeps playback position, transcript scroll and quiz state on one synchronized timeline, so resuming on a phone lands exactly where the laptop stopped. Transcripts are word-timed and searchable, and clicking a line seeks the video. Quizzes are inline prompts that pause playback: students had been bouncing out to a separate quiz screen and not coming back.
- One timeline for video, transcript and quiz state
- Word-timed searchable transcripts that seek playback
- Inline quiz prompts that pause playback in place
Unified Assessment Suite
Consolidates assignments, peer reviews, and automated gradebook tracking into one dashboard.
Assignments, peer review and the gradebook were three tools with three deadlines. They now share a single submission record, so a peer review is a state that record passes through, with no separate artifact to reconcile afterward. Weighting is declared once per course and the gradebook derives from it, which removed the end-of-term spreadsheet reconciliation entirely.
- One submission record from upload to final mark
- Peer review as a state, not a parallel system
- Weighting declared once; the gradebook derives from it
The marking queue: weights adding to 100 across four assignments, 95 hand-ins and 76 marked, eight submissions waiting, and the open one beside its five-criterion rubric with a mark of 71 that stays a draft until it is released.
What we built together
- 01
Shadowed instructors to map their real grading and intervention flow
Shadowing a full grading cycle showed instructors tracked who had gone quiet, not who was failing, and the risk model was built on exactly that.
- 02
Designed a single progress model shared by learners and staff
Learner progress and staff reporting read one model, so a student's position is the same number wherever it appears. No two systems agreeing by luck.
- 03
Built a focused lesson player and an instructor cockpit
The player holds playback, transcript and quiz state on one timeline, and quizzes pause it inline, so a student never leaves for another page.
- 04
Instrumented completion analytics from day one
Completion analytics shipped in the first increment, so every later change could be measured against a baseline that already existed.
Operational results after launch
+28%
Course completion
−45%
Instructor admin time
−32%
Support tickets
Completion compares the two cohorts after launch against the four preceding ones, so the pre-existing downward drift is visible in the baseline instead of flattering the result. Support tickets and instructor admin time are the provider's own figures, measured per course and not in total, because course sizes vary by a factor of six.
Client name withheld under NDA. Figures are approximate, drawn from the engagement’s own reporting.
Phase by phase
Phase 1: Research & Workflow Mapping
Instructor & Student Journey Audit
Shadowed 15 instructors and 30 students to isolate workflow disconnects, grading bottlenecks, and drop-off causes.
- Instructor Workflow Map
- Student Friction Analysis
- EdTech UI Benchmark
Phase 2: Platform UX Architecture
Unified Interface & Design System
Designed a distraction-free student lesson player and a comprehensive instructor cockpit for real-time progress tracking.
- Figma Design System
- Interactive Lesson Player Prototype
- Usability Audit
Phase 3: Full-Stack Engineering
Real-Time Platform & Assessment Engine
Built the web application with Next.js and WebSockets for real-time progress syncing, automated grading, and instant messaging.
- Platform Application Engine
- Gradebook API
- WebSocket Event Service
Phase 4: Pilot Rollout & Analytics
Course Pilot & Performance Verification
Launched first on 12 of the provider's eighty courses, monitoring completion rates and instructor time savings to validate business impact.
- Completion Analytics Dashboard
- Instructor Feedback Report
- Full Production Release
The course builder: six modules and twenty-two lessons totaling 3 hr 39 min, added up from the lessons, never typed, and the one publish rule the course still fails.
About our collaboration
A cross-functional team of 5 worked on a fixed price, phased basis over 18 weeks, covering Product design, Frontend & backend build, Analytics. We ran a standing mid-week checkpoint and written decisions in place of status meetings. Nothing shipped without a live demo first.
We shadowed instructors through a full grading cycle before designing anything, and that's where the intervention cockpit's ranking came from: the spreadsheets they kept were sorted by who had gone quiet, not by who was failing. Every increment went to two courses before the eighty, and the analytics shipped in the first increment, not the last.
What we'd carry into the next one
- 01
Early instructor intervention indicators boosted course completion rates by 28%.
Week two is early enough to act and week ten isn't. That's the entire argument for reading behavior over grades: grades don't exist yet in week two.
- 02
A distraction-free lesson player reduced platform support tickets by 32%.
The tickets that vanished were navigation questions, not technical ones. A player with nothing else on screen removed the thing people were asking about.
- 03
Consolidating course tools into one interface reduced administrative overhead by 45%.
One submission record from upload to final mark removed the end-of-term reconciliation entirely, which is where most of the reclaimed admin time actually came from.
One hand-in, marked and released
The mark adds up from the rubric, and the learner sees none of it until it is released.
A VAT return handed in on a bookkeeping course, followed through the four states of its one submission record, with the marker’s screen and the learner’s side by side. Switch tabs, or use the arrow keys once one is focused.
The hand-in arrives already past peer review, because peer review is a state this record passed through, not a separate tool. The learner handed it in against the same five criteria, at the same weights, that the marker is about to use.
- Uploaded
- Peer review
- Marking
- Released
- Output VAT30%
- Input VAT25%
- Return boxes 1–920%
- Reconciliation15%
- Audit trail10%
Why nothing leaks early: the learner’s view reads the submission record’s released state, not the score. Scoring pace in this replay is illustrative.
From a paused lesson to the instructor who needs to know about it
Built on Next.js and Node.js with a WebSocket event service. Learners and staff read one progress model, so the learner’s screen and the cockpit never have to agree by luck.
- 01 · SourceLesson player and hand-insPlayback position, transcript scroll and quiz state sit on one timeline, so a pause, a seek and an inline answer are a single position.
- 02 · TransportWebSocket event serviceProgress syncs in real time, which is why resuming on a phone lands exactly where the laptop stopped.
- 03 · EngineProgress model and risk signalsRisk is read from behavior: session gaps, abandoned players, quiz retries and time to first submission. It can fire in week two, before any grade exists.
- 04 · StateSubmission record in PostgreSQLOne record from upload through peer review to final mark. Weighting is declared once per course and the gradebook derives from it.
- 05 · DeliveryLearner view and instructor cockpitBoth read the one progress model through the Gradebook API, so a learner's position is the same number wherever it appears.
What a term can no longer hide
Marks released right, learners seen early, admin kept out of teaching
A mark is a draft until someone releases it
The mark is summed from the rubric's weighted criteria, never typed, and it stays a draft the learner can't see. Their screen shows nothing at all until release, then the whole mark at once.
Stalling shows up in week two, with its evidence
Risk is read from session gaps, abandoned lesson players, quiz retries and time to first submission, not grades. The cohort is ranked by drift, and every flag opens the behaviour behind it.
No end-of-term reconciliation to do
Assignments, peer review and the gradebook share one submission record, and weighting is declared once per course. The end-of-term spreadsheet reconciliation isn't faster. It's gone.
Finding out who dropped out only at final assessment? Scope your build in 3 minutes.
Scope your buildNearby engagements
AI & AutomationA private legal assistant grounded in verified precedents
A private knowledge assistant that searches internal case files and precedents, providing cited answers legal teams can verify in seconds.
Legal & Law Firms · 14 weeks
Product DesignAn onboarding flow that guides trial users to value
A redesigned SaaS trial onboarding experience with progressive checklists, sample data, and inline guidance that turns signups into active subscribers.
Professional Services · 10 weeks
Product DesignA design system that brought speed and consistency to 4 product teams
A token-based design system in Figma and React that eliminated component duplication across 4 product squads and cut the time from design handoff to merged frontend.
Professional Services · 14 weeks
Let's talk
Running a large platform, shaping a first MVP, or getting a product ready for a funding round? Tell us where you are. We'll shape the process around it, and stay with you after launch.
