A self-contained pipeline for tracking attendance, engagement, pre-class work, and skills assessment across the six-session DIFC cohort. No coding required — the recommended way to run it is through Claude Cowork.
Each time you process a session, Claude will:
- Classify each student as Present, Partial, or Absent based on time-in-session
- Update the Excel master tracker with the new session row
- Regenerate three static attendance charts (heatmap, trend, histogram)
- Rebuild
dashboard.html— a fully interactive, browser-based dashboard with two tabs
The dashboard is entirely self-contained. Open it in any browser. No server or internet connection required (except for fonts on first open).
The primary working view for day-to-day course management.
Top KPIs
| KPI | What it shows |
|---|---|
| Course Progress | Session X of Y with a progress bar and sessions-remaining sub-label |
| Attendance Rate | Weighted rate for the latest session, delta vs prior, cumulative avg across all sessions |
| Student Talk Time | % of session voice that was student-led, delta vs prior, and benchmark vs target |
| Needs Attention | Students with cert risk OR high engagement risk (deduplicated count) |
Attendance Heatmap — one row per student, one column per session. Colour-coded Present / Partial / Absent. Hover cells for late-arrival and early-departure details. Search bar to filter by name.
Session Attendance Detail — table showing per-session headcount, late arrivals, early departures, and attendance rate.
Session Activity — dual-axis chart showing reactions, hand raises, and chat messages per enrolled student across sessions.
Student Talk Time — trend chart showing student vs instructor voice split per session.
Session Engagement Detail — expandable table with per-session raw engagement counts.
Certificate Eligibility — one row per student showing sessions attended, missed, remaining, and how many more they can afford to miss. Missed count shows a hover tooltip with specific session labels. Pre-class work completion badges shown per session. Status light (green / amber / monitoring / red) based on eligibility rules. A skill-gap risk notice appears here when students are missing proficiency on assessed outcomes.
A student needs to attend 4 of 6 sessions and complete pre-class work for any session they missed in order to receive a certificate.
A secondary view for exploring skills development across the cohort. Five sub-tabs, in order:
Top-level KPIs and the Cohort Competency Matrix — all 11 Durable Skills learning outcomes arranged in the full framework hierarchy (Competency → Sub-competency → LO). Columns use a 3-row header:
- Row 1: Competency (solid brand colour)
- Row 2: Sub-competency (tinted)
- Row 3: Individual
#lo-namein monospace
Columns not yet assessed appear as ghost/dimmed cells. Hover any header for full name or breadcrumb path. The COHORT AVG footer row shows mean score per outcome, overall average, and cohort average talk time. Click any student row to open the Student Detail drilldown. Search bar filters by name.
Score legend: 0 No submission · 1 Doesn't apply · 2 Partial · 3 Proficient · 4 Exemplary
Cohort-level radar chart showing mean scores across the four competencies, plus competency mini-cards with proficiency rates, trend arrows, and sub-competency labels.
Master-detail view for individual student analysis.
Left sidebar — scrollable list of all students with coloured status dots. Search by name or sort by name / status. Scales to any cohort size.
Right panel — opens when a student is selected:
- Radar chart (left, 45% width) showing student scores vs cohort average (dashed line)
- Four competency cards (right, 2×2 grid) showing mean score, proficiency rate, and delta vs cohort. Click a card to expand outcome-level detail below — showing per-LO scores grouped by sub-competency with session history. Click again to collapse.
- Pre-Class Work — collapsible section showing per-session PCW scores with attendance context.
Score distribution charts — by competency (horizontal bar) and by individual learning outcome.
Trajectory charts showing how cohort scores have changed across sessions — overall and by individual LO.
| Status | Rule |
|---|---|
| Present | Attended ≥ 75% of the official session duration |
| Partial | Attended 40–74% of the official session duration |
| Absent | Attended < 40%, or no record in the attendance CSV |
The denominator is always the official session duration you provide, not the recorded platform time. Late arrivals (> 10 min after session start) and early departures (> 10 min before end) are flagged in the heatmap.
| Score | Meaning |
|---|---|
| 0 | No submission |
| 1 | Doesn't apply the skill, or mostly inaccurate |
| 2 | Recalls / uses the skill only somewhat accurately |
| 3 | Accurately and effectively uses the skill |
| 4 | Accurately uses the skill AND demonstrates deep grasp through analysis or justification |
Score 0 is excluded from the Cohort Average computation to avoid conflating non-participation with low performance. The submission rate (shown in the sub-label) captures participation separately.
Critical Thinking
├── Reasoning → #evidence-based
├── Problem Solving → #right-problem · #break-it-down · #constraints · #gap-analysis
└── Decision Making → #purpose · #decision-selection
Effective Communication
└── Structure & Clarity → #message-construction
Effective Interactions
└── Behavior Strategy → #self-awareness
Values & Moral Reasoning
└── Integrity & Ethics → #discernment · #accountability
1. Attendance CSV — exported from your virtual classroom platform. Save to DIFC Attendence Tracker/.
Required columns: student name, email, time first entered, time last exited (or minutes attended).
2. Engagement CSV — exported from the Forum platform session log. Save to DIFC Session Engagement Tracker/.
Required columns: session duration, attendance count, reactions, hand raises, chat messages, student talk time %, instructor talk time %.
Place any of these in the Session Data/ subfolder. The pipeline auto-discovers the most recent file for each type by filename prefix.
| File prefix | What it adds |
|---|---|
ForumEngagementData- |
Per-student talk time and engagement risk level |
student_points_per_page_ |
Pre-class work scores per student per session |
student_scores_per_outcome_ |
Durable Skills rubric scores (0–4) per student per outcome |
These files are optional. If absent, the relevant sections of the dashboard show placeholder states rather than erroring.
The Export All button (top right of the dashboard) downloads a ZIP file containing:
Charts (PNG)
cohort-profile-radar.png— cohort radar across the four competenciesgrowth-trajectories.png— overall score growth across sessionsgrowth-by-lo.png— per-LO growth trajectoriesscore-dist-by-competency.png— score distribution by competencyscore-dist-by-lo.png— score distribution by individual learning outcomeattendance-trend.pngtalk-time.pngsession-activity-per-learner.png
Tables (CSV)
attendance-heatmap.csv— student × session attendance statusessession-attendance.csv— per-session headcount summarysession-engagement.csv— per-session engagement metricscertificate-tracking.csv— certificate eligibility per student with PCW badgescompetency-matrix.csv— all 11 LOs × all students with rubric scorescompetency-summary.csv— mean score and proficiency rate per competencypcw-scores.csv— pre-class work numeric scores, student × session matrixlo-growth.csv— per-LO mean score and proficiency rate across sessions
Raw data
dashboard_data.json— complete data snapshot including skill snapshots, engagement history, and all session records
Step 1 — Download the shared folder to your computer. Keep all files together — do not rename or move anything inside it.
Step 2 — Open the Claude desktop app. In the bottom-left, click Select a folder and choose the downloaded folder. This gives Claude read/write access to all files inside it.
Tell Claude something like:
"Please process Session 3. The attendance CSV is Session_3.csv (90 minutes) and the engagement CSV is ForumSessionLog-Session3.csv."
Claude will classify attendance, update the Excel tracker and charts, and rebuild dashboard.html. It will confirm how many students were Present, Partial, and Absent, and note if any data anomalies were detected.
If you have Session Data files (engagement, PCW, outcome scores), drop them into the Session Data/ folder before asking Claude to process — they will be picked up automatically.
DIFC Data Project/
│
├── README.md ← This file
├── run_weekly.py ← Main pipeline (do not edit)
│
├── dashboard.html ← Interactive dashboard (auto-generated)
├── dashboard_data.json ← Persistent data store — do not delete
│
├── DIFC Attendence Tracker/
│ ├── run_session.py ← Attendance sub-pipeline (do not edit)
│ ├── requirements.txt ← Python dependencies (reference only)
│ ├── session_data.json ← Attendance data store — do not delete
│ ├── Attendance_Master_Tracker.xlsx ← Updated each session
│ ├── chart_1_heatmap.png ← Updated each session
│ ├── chart_2_trend.png ← Updated each session
│ └── chart_3_histogram.png ← Updated each session
│
├── DIFC Session Engagement Tracker/
│ └── (engagement CSVs go here)
│
└── Session Data/ ← Optional enrichment files
├── ForumEngagementData-*.csv ← Per-student talk time + risk
├── student_points_per_page_*.csv ← Pre-class work scores
└── student_scores_per_outcome_*.csv ← Durable Skills rubric scores
Important: Do not delete
dashboard_data.jsonorsession_data.json. These files hold all historical session data. Deleting them means re-running every session from scratch.
Claude says it can't find a file Confirm the CSV is in the correct subfolder and that the filename you gave Claude matches exactly, including capitalisation.
The dashboard hasn't updated Ask Claude to confirm outputs were saved to the folder. If the folder wasn't selected before you started, close the conversation, reselect the folder, and try again.
The data looks wrong Check the session number and duration you provided. Wrong session number is the most common cause of unexpected data.
Something unexpected Paste the error or the unexpected output into Claude and ask for help.
Sean Hughes — shughes@minervaproject.com