What a Learning Analytics Dashboard actually does
Ingests xAPI, SCORM, and direct LMS event data to surface learner engagement, completion rates, and at-risk flags — then answers L&D team questions in natural language via validated SQL with FERPA-safe data routing.
The platform builds on the same text-to-SQL + visualization stack as the AI Data Visualization Tool, with three key differences specific to learning analytics: the data domain is xAPI/SCORM event streams (not operational databases), FERPA applies when K-12/HE customers are served, and GDPR Art. 22 applies to any 'at-risk learner' classification that automatically triggers academic consequences.
The AI pipeline: xAPI statements ingested from a conformant LRS or direct LMS API → Cube.dev semantic layer for L&D-specific measures (completion rate, engagement score, knowledge-gap flags) → GPT-5.4 mini ($0.75/$4.50) for natural-language queries ('which learners stalled in week 3?') → Mistral Large 3 ($0.50/$1.50) for weekly cohort summaries. Classical XGBoost flags 'at-risk learners' — always surfaced with a human-review flag, never auto-actioned. FERPA data routing through Anthropic ZDR or Bedrock for K-12/HE customers.
The market gap: Watershed LRS (enterprise), Yet Analytics (enterprise), Anthology Illuminate (Blackboard, enterprise), and Civitas Learning (HE-focused, enterprise) all target 50K+ learner deployments. No SMB-targeted white-label learning analytics platform exists for the L&D consultant serving 5–50 corporate clients or the boutique EdTech company.
AI capabilities involved
xAPI/SCORM/direct LMS event ingestion and normalization
Natural-language queries over learning records
At-risk learner classification (GDPR Art. 22 human-review flag)
Weekly cohort engagement summary generation
Question-text semantic search across learning content
Who uses this
- L&D consultants serving 5–50 corporate clients who need branded analytics dashboards
- EdTech founders embedding completion and engagement analytics for district or employer customers
- Corporate university leads who need cohort-level learning analytics without enterprise LRS pricing
- Regulated CE providers needing audit-ready completion reports for state board accreditation
SaaS alternatives on the market
Real products you can sign up for today — with current 2026 pricing, honest pros and cons.
Watershed LRS
Enterprise L&D teams with 10K+ learners and dedicated data analysts who need a production-grade xAPI LRS.
Enterprise quote
Pros
- +xAPI 1.0 conformant LRS with full statement storage and retrieval.
- +Advanced visualization of learning data with custom dashboards.
- +Strong L&D professional community and open-source compatibility.
- +Integration with major LMS platforms (Moodle, Canvas, Cornerstone).
Cons
- −Enterprise quote — no transparent SMB pricing.
- −No white-label or agency reseller path.
- −No AI NL-query feature — visualization-only analytics.
- −Designed for large enterprises, not boutique L&D consultancies.
Yet Analytics
Academic researchers and L&D teams with DevOps capacity who want open-source LRS with self-hosting flexibility.
Enterprise quote
Pros
- +Open-source SQL LRS with xAPI conformance and open-source availability.
- +Strong standards compliance and academic research use case.
- +Self-hostable for full data ownership.
- +Active open-source community.
Cons
- −Enterprise support and SLA require quote-based contracts.
- −No white-label dashboard for L&D agencies.
- −No AI analytics or NL-query features in the base product.
- −Self-hosting requires DevOps capacity.
Looker for Education (Google Cloud)
Enterprise corporate universities or large university systems already running Google Cloud data infrastructure.
~$5,000+/mo (Looker tied to Google Cloud)
Pros
- +Powerful Looker ML for learning-data exploration.
- +Google Cloud BigQuery integration for large-scale xAPI data.
- +Generative AI features (LookML, Looker Studio AI).
- +Strong enterprise trust.
Cons
- −Google Cloud dependency adds $500–$2,000/mo in BigQuery costs beyond Looker licensing.
- −No white-label or agency reseller path.
- −Overkill for L&D consultancies under $500K/yr in revenue.
- −$5K+/mo minimum is prohibitive for the SMB segment.
The AI stack
Learning analytics AI extends the text-to-SQL visualization stack with two L&D-specific additions: an xAPI event normalization layer and GDPR Art. 22-compliant at-risk classification that always surfaces as a human-review flag, never as an automated action.
Natural-language queries over learning records
Answer L&D team questions like 'which learners stalled in week 3?' or 'what is the completion rate for the compliance training cohort?'
GPT-5.4 mini + Cube.dev semantic layer
$0.75/$4.50 per M tokens (~$0.003 per simple query)Standard L&D analytics queries (completion rates, module performance, learner progress) on structured semantic layer.
GPT-5.4 for complex multi-cohort queries
$2.50/$15 per M tokens (~$0.0035 per complex query)Comparative cohort analytics ('compare Week 1 completion rates across Q1 vs Q2 cohorts by department').
Our pick: GPT-5.4 mini for standard L&D metrics queries. GPT-5.4 for complex multi-cohort comparisons. All queries must go through Cube.dev semantic layer + AST validation + RLS injection before execution.
At-risk learner classification
Identify learners at risk of non-completion or low performance — always as a recommendation for human review, never as an automated action.
XGBoost on engagement + completion features
~$0 (Trigger.dev managed compute)Any at-risk classification where GDPR Art. 22 explanation requirements apply — SHAP values are the audit trail.
Our pick: XGBoost universally for at-risk classification — always surface as 'At-Risk Flag (Review Required)' with the SHAP-based reason. Never auto-email, auto-downgrade, or auto-exclude learners based on the AI flag alone.
Weekly cohort summary generation
Generate a plain-English summary of cohort performance for L&D team distribution — replacing manual weekly analytics reports.
Mistral Large 3
$0.50/$1.50 per M tokens (~$0.003 per cohort summary)High-volume weekly summaries for multiple corporate clients where cost-per-summary is material.
Our pick: Mistral Large 3 for all weekly cohort summaries — the quality is adequate for L&D professional audiences and the cost savings at 50+ summaries/week are meaningful.
Reference architecture
xAPI event ingestion → Cube.dev L&D semantic layer → GPT-5.4 mini NL queries + XGBoost at-risk classification + Mistral cohort summaries → L&D branded dashboard. The most complex piece is the xAPI normalization ETL: xAPI statements from Moodle, Canvas, and Cornerstone use different verb/object patterns for the same learning activity, requiring per-source mapping tables.
xAPI statement ingestion from LRS or LMS API
Trigger.dev job → xAPI normalization ETL → Supabase learning_events tableEach statement normalized to: {actor_id, verb (completed/attempted/passed/failed), object_id (course/module/assessment), result (score, completion, success), timestamp}. Source-specific verb mappings defined in a normalization_rules table.
Cube.dev L&D semantic layer configuration
Cube.dev OSS (self-hosted)Defines L&D-specific measures: completion_rate, avg_score, engagement_minutes, at_risk_flag. Dimensions: cohort, department, module, date. Joins: learning_events to learner_profiles. No PII columns exposed to LLM.
L&D team submits NL query
Next.js analytics dashboard → Edge Function → Cube.dev → GPT-5.4 miniQuery submitted to Edge Function; Cube.dev semantic metadata retrieved; GPT-5.4 mini generates SQL constrained to L&D measures; AST validated; RLS clause injected (WHERE tenant_id = {id}); executed; results returned as chart + optional Mistral summary.
Nightly XGBoost at-risk classification
Trigger.dev nightly cron → XGBoost → at_risk_flags tableScores each active learner on: login_frequency (7-day), module_completion_rate, assessment_score_trend, days_since_last_activity. Outputs: risk_level (low/medium/high), risk_reason (SHAP). Stored with human_review_required=true.
L&D admin reviews at-risk flags
Next.js L&D dashboardAt-risk learners shown with flag reason and suggested intervention (reach out, offer tutoring, schedule check-in). No automated action taken — all interventions initiated by L&D admin. Art. 22 audit log records all views and actions taken.
Weekly cohort summary generated by Mistral Large 3
Trigger.dev weekly schedule → Mistral Large 3 → ResendFor each L&D client cohort, Mistral Large 3 generates a 3-paragraph summary: overall completion, top-performing module, at-risk learner count (not names). Delivered to L&D team by email.
Estimated cost per request
~$0.005 per cohort summary (Mistral Large 3) + ~$0.003 per NL query (GPT-5.4 mini) + ~$0 at-risk classification (XGBoost amortized). Total monthly AI COGS at 10 clients: ~$10/mo.
Cost calculator
Drag the sliders to model your actual usage. The numbers update in real time so you can stress-test economics before writing a single line of code.
Modeled at 10 L&D consultant clients, each with 200 active learners and 1 weekly cohort summary. AI costs are negligible — infra dominates.
Estimated monthly cost
$105
≈ $1,263 per year
Calculator notes
- At 10 clients × 20 NL queries/week × 4 weeks = 800 queries/mo × $0.003 = $2.40/mo AI cost for NL queries. Total monthly COGS: ~$107/mo.
- FERPA routing via Bedrock adds 10–20% to AI API costs for K-12/HE clients — route corporate L&D clients through direct Anthropic API (cheaper).
- At-risk XGBoost scoring runs nightly on all active learners at ~$0 marginal cost after Trigger.dev fixed compute — no per-learner AI cost.
- xAPI event ingestion costs are primarily ETL compute (Trigger.dev) — estimate $0.000025/run × 10 clients × 200 daily events = $0.05/day = $1.50/mo.
Build it yourself with vibe-coding tools
In a weekend, build a mock-data L&D analytics dashboard with CSV upload and basic NL queries — enough to demo to L&D consultants before investing in the full xAPI pipeline.
Time to MVP
12–16 hours (mock data demo only)
Total cost to MVP
$25 Lovable + $30 OpenAI + Metabase OSS free = L&D mock-data demo in a weekend
You'll need
Starter prompt
Build a DEMO-ONLY AI Learning Analytics Dashboard (synthetic data only). Features: - Cohort overview tiles: completion rate, avg assessment score, at-risk learner count, weekly active learners - Module performance chart: completion rate + avg score per module (bar chart, auto-generated from synthetic data) - At-risk learner list: table with risk level, risk reason, days since last activity — with 'Human Review Required' badge on every row - NL query input: 'Which learners completed module 3 but not module 4?' → shows chart + table result - Weekly summary preview: 2-paragraph AI-generated summary of cohort performance SYNTHETIC DATA: Pre-seed Supabase with: learning_events (1,000 rows, 60 learners, 10 modules, past 90 days), learners (60 rows mock names), modules (10 rows). BANNER: 'DEMO — synthetic data only. Real xAPI event ingestion and FERPA routing require professional build. GDPR Art. 22: at-risk flags are informational only — no automated actions.'
Paste this into Lovable
Follow-up prompts (run in order)
- 1
Wire up NL query: call GPT-5.4 mini with the synthetic schema (learning_events: learner_id, module_id, verb, score, timestamp) and the user's question. Return SQL → execute on Supabase → render as auto-selected Recharts chart. Add AST validator: reject if not SELECT, if tables not in allowlist, if LIMIT missing.
- 2
Wire up weekly summary: call Mistral Large 3 with: 'Write a 2-paragraph L&D analytics summary. Data: completion_rate={pct}, top_module={name} ({score}), at_risk_count={n}, new_completions_this_week={n}. Paragraph 1: overall cohort health. Paragraph 2: specific action recommendations for the L&D team. Professional tone.'
- 3
Add GDPR Art. 22 at-risk display: in the at-risk learner list, add a mandatory disclaimer header: 'These flags are AI-generated recommendations for human review. No automated action has been taken. L&D administrator must review each flag before intervention.' Log all at-risk flag views to an audit_log table: {viewer_id, learner_id, flag_reason, viewed_at, action_taken}.
Expected output
A demo L&D dashboard showing completion charts, at-risk flags with human-review badges, NL queries, and weekly summaries on synthetic data. Sufficient for L&D consultant business development. Not production-ready: no xAPI ingestion, no FERPA routing, no real learner data.
Known gotchas
- !xAPI statement normalization is the highest-effort piece of the real build: Moodle uses 'http://adlnet.gov/expapi/verbs/completed' while Cornerstone may use a proprietary verb URI for the same activity — every LMS source requires its own mapping table.
- !GDPR Art. 22 at-risk flags must include a human-review gate in the L&D dashboard UI — if an automated email or Slack alert is sent to a learner flagged as at-risk without human approval, this is a potential Art. 22 violation for EU learners.
- !Learner name and email are PII under GDPR and FERPA — the Cube.dev semantic layer must exclude these fields from all NL-queryable columns. Always refer to learners by anonymized ID in AI calls; de-anonymize only in the frontend display layer with appropriate access controls.
- !Metabase's white-label feature on Pro ($85/mo Cloud) removes branding but does not include xAPI ingestion or AI NL queries — you still need to build the analytics pipeline on top of Metabase's query API.
- !Weekly summary Mistral calls should include only aggregate data (completion_rate, avg_score, at_risk_count) — never individual learner names in AI prompts, even in truncated form.
- !FERPA requires that AI API calls processing K-12/HE learner records route through ZDR-enabled endpoints (Anthropic API with ZDR or AWS Bedrock) — direct Anthropic API without ZDR is not compliant for K-12/HE customers.
Compliance & risk reality check
Learning analytics compliance depends entirely on who the learners are: corporate L&D (adult employees) has light compliance load; K-12/HE has heavy FERPA load. GDPR Art. 22 applies to any at-risk classification regardless of learner type.
FERPA for K-12/HE learners (critical when applicable)
FERPA applies when learning events come from K-12 schools or universities receiving federal funding. Corporate L&D analytics (company training data) does NOT trigger FERPA. When K-12/HE customers are in scope, per-district/institution DPAs are required and all AI calls processing learner records must route through FERPA-compatible endpoints (Anthropic ZDR, AWS Bedrock).
Mitigation: Create a customer_type flag: 'corporate' vs 'k12_he'. Route AI calls for k12_he customers through Anthropic ZDR or AWS Bedrock. Execute a FERPA DPA with each K-12 district or HE institution before ingesting their learner data.
GDPR Art. 22 — at-risk learner classification
GDPR Article 22 applies to automated individual decision-making with significant effects. Flagging a learner as 'at-risk' and automatically triggering academic consequences (email to parent, grade change, exclusion from program) without human review qualifies as an automated decision with significant effects. The at-risk flag itself (an internal analytics metric) is not a violation — the automated action triggered by it is.
Mitigation: Surface all at-risk flags as 'Human Review Required' in the L&D dashboard. Never auto-send communications to EU learners based on AI flags alone. Log all views and actions taken on at-risk flags to the compliance audit trail. Include in privacy policy: 'We may use automated analysis of your learning activity to generate recommendations for your learning coordinator. A human reviews these recommendations before any action is taken.'
AB 2013 California training-data disclosure (Jan 1, 2026)
If learner interaction data is used to train AI models, AB 2013 requires disclosure. With ZDR routing (no data retention by AI providers), this obligation is eliminated.
Mitigation: Enable Anthropic ZDR and include in privacy policy: 'Learning activity data is not used to train AI models. We use AI APIs with zero data retention to ensure your learning data remains private.'
Build vs buy: the real math
8–12 weeks
Custom build time
$25,000–$50,000
One-time investment
15–25 months
Breakeven vs buying
No comparable white-label L&D analytics platform exists at SMB pricing. An L&D consultancy charging 10 clients $99/mo = $990/mo revenue vs $100/mo infra COGS = 90% gross margin. A $37.5K midpoint build recoups in 38 months at 10 clients, 19 months at 20 clients ($1,980/mo revenue). The build creates a new revenue stream — branded L&D analytics — in a market where all alternatives are enterprise-only. AI API costs at $10/mo for 10 clients become effectively zero as Mistral and GPT-5.4 mini pricing declines, pushing gross margin toward 99% over time.
Skip the DIY — RapidDev builds the production version
A Lovable MVP gets you a demo. Production needs auth that doesn't leak data, AI calls that don't bankrupt you, observability when models drift, and code you can audit. That's what we ship.
Discovery call (free)
30 minWe map your exact Learning Analytics Dashboard use case: who uses it, target volume, AI model choice, integrations, compliance scope. You get a detailed scope document and fixed-price quote within 48 hours.
AI-accelerated build
8–12 weeksOur engineers use Claude Code, Lovable, and custom tooling to ship 3–5x faster than agencies. You see weekly progress in a staging environment — not a black box.
Launch + handoff
1 weekWe deploy to your infrastructure, transfer the GitHub repo, set up CI/CD and monitoring, and train your team. You own 100% of the source code, prompts, and model configurations.
What you get
Timeline
8–12 weeks
Investment
$25,000–$50,000
vs SaaS
ROI in 15–25 months
30-min call. Fixed-price quote within 48 hours. No commitment.
Frequently asked questions
How much does it cost to build a white-label AI learning analytics dashboard?
RapidDev builds this for $25,000–$50,000 over 8–12 weeks. The lower end covers: xAPI event ingestion with normalization for 2–3 LMS sources, Cube.dev L&D semantic layer, GPT-5.4 mini NL queries with AST validation and RLS, XGBoost at-risk classification with GDPR Art. 22 human-review flags, and Mistral weekly cohort summaries. The upper end adds: FERPA routing for K-12/HE customers via Bedrock, SCORM completion tracking for CE accreditation audit logs, and 5+ LMS source integrations. This is above the standard $13K–$25K band due to xAPI normalization complexity and FERPA scaffolding.
How long does it take to ship an AI learning analytics dashboard?
8–12 weeks. A mock-data demo can be built in a Lovable weekend. The 8-week production build adds: xAPI normalization ETL for 2 LMS sources, Cube.dev semantic layer, NL-query pipeline, XGBoost at-risk classification, and weekly summary delivery. The 12-week version adds 3+ additional LMS connectors, FERPA routing via Bedrock, CE completion audit reports, and SOC 2 evidence collection.
Can RapidDev build this for my L&D agency or EdTech company?
Yes. RapidDev has built analytics platforms with xAPI event ingestion, Cube.dev semantic layers, and FERPA-compliant data routing. We start by mapping your top LMS data sources (Moodle, Canvas, Cornerstone, TalentLMS) and defining the L&D-specific Cube.dev measures before writing any NL query code. Book a free 30-minute consultation at rapidevelopers.com.
Does GDPR Art. 22 prevent me from building an at-risk learner feature?
No — GDPR Art. 22 prevents automated decisions with significant effects without human review, not the at-risk flag itself. An at-risk classification shown to an L&D administrator for their review is compliant. What is not compliant: automatically enrolling a flagged learner in remediation, automatically emailing the learner, or automatically excluding them from a program based solely on the AI flag. Build the human-review gate (L&D admin must acknowledge and act on each flag) and log those actions — this satisfies Art. 22.
What is xAPI and how is it different from SCORM?
xAPI (Experience API) is a modern learning data standard that captures any learning experience as a 'verb-noun-object' statement: 'learner completed module', 'learner scored 85 on assessment', 'learner practiced skill in simulation.' SCORM is the older standard focused on tracking structured e-learning course completion within a single LMS. xAPI works across mobile, offline, simulations, and real-world activities — SCORM is confined to online LMS-delivered courses. For a learning analytics dashboard, xAPI provides richer, more granular learner-behavior data; SCORM provides simpler completion/score signals. Most modern LMS platforms support both.
Can this platform replace a formal Learning Record Store (LRS)?
The platform includes an xAPI statement ingestion layer that stores normalized learning events — functionally similar to an LRS for analytics purposes. For formal xAPI 1.0 conformance (required by SCORM Cloud, ADL compliance programs, or enterprise LMS integrations that push statements to your endpoint), the storage layer must pass the xAPI conformance test suite. This is an additional 2–4 weeks of development. If your use case requires full xAPI conformance certification, flag this during scoping — most L&D analytics use cases do not require it, but CE accreditation programs sometimes do.
Want the production version?
- Delivered in 8–12 weeks
- You own 100% of the code
- AI cost monitoring built in
30-min call. No commitment.