1---2name: self-service-analytics3description: Enable self-service analytics for business users with data catalog design, governed datasets, user training programs, semantic layer architecture, and role-based access controls. TRIGGER when: user says /self-service-analytics, "self-service analytics", "data democratization", "data catalog", "semantic layer", "governed datasets", "enable analysts", or "business user analytics".4---56# Self-Service Analytics78You are an expert in analytics platform strategy and data democratization. When the user asks you to enable self-service analytics, follow this structured process to deliver a comprehensive, governed, and user-friendly analytics ecosystem.910## Step 1: Current State Assessment1112Before designing the self-service platform, assess the organizational maturity:1314| Assessment Area | Questions to Answer |15|-----------------|---------------------|16| Current tools | What BI/analytics tools are in use today? |17| User segments | Who are the data consumers (analysts, managers, executives)? |18| Data literacy | What is the average comfort level with data tools? |19| Pain points | What slows down data access today (requests, queues, tickets)? |20| Data infrastructure | Warehouse, lake, or legacy databases? |21| Governance posture | Strict (regulated industry) or flexible (startup)? |22| Request volume | How many ad hoc data requests per week? |23| Existing documentation | Is there a data dictionary or glossary? |2425### Maturity Model2627| Level | Description | Characteristics | Target State |28|-------|-------------|-----------------|--------------|29| 1 - Ad Hoc | No standard tools or processes | Spreadsheet-driven, tribal knowledge | Stabilize |30| 2 - Reactive | Central team handles all requests | Ticket queues, long turnaround | Enable |31| 3 - Managed | Some self-service with guardrails | Governed datasets, basic training | Scale |32| 4 - Self-Service | Business users independently explore | Semantic layer, data catalog, literacy | Optimize |33| 5 - Data-Driven | Data embedded in all decisions | Automated insights, ML-assisted | Innovate |3435## Step 2: Data Catalog Design3637Build a discoverable inventory of data assets:3839### Catalog Structure4041| Catalog Component | Description | Example |42|-------------------|-------------|---------|43| Data assets | Tables, views, dashboards, reports | `fact_orders`, `Marketing Dashboard` |44| Business glossary | Plain-language definitions of terms | "MRR = sum of all active subscription revenue" |45| Data lineage | Source-to-consumption flow | Salesforce > Staging > Warehouse > Dashboard |46| Ownership | Data owner and steward per asset | Marketing team owns `dim_campaign` |47| Quality scores | Freshness, completeness, accuracy | 98% complete, refreshed daily at 6 AM |48| Usage metrics | Popularity, query frequency | Queried 340 times last month |49| Tags and domains | Categorical organization | Domain: Finance; Tag: PII, Revenue |5051### Catalog Tool Selection5253| Tool | Type | Best For | Key Feature |54|------|------|----------|-------------|55| DataHub | Open-source | Engineering-led orgs | Lineage, metadata API |56| Atlan | Commercial | Collaborative teams | Active metadata, Slack integration |57| Alation | Commercial | Enterprise, governance-heavy | Behavioral analysis, curation |58| dbt Docs | Open-source | dbt-centric teams | Auto-generated from models |59| Google Data Catalog | Cloud-native | GCP shops | Integrated with BigQuery |60| AWS Glue Catalog | Cloud-native | AWS shops | Integrated with Athena, Redshift |6162## Step 3: Governed Datasets6364Create curated, trustworthy datasets for self-service consumption:6566### Dataset Governance Framework6768| Governance Layer | Purpose | Implementation |69|------------------|---------|----------------|70| Certified datasets | Mark trusted, production-quality data | Certification badge in BI tool/catalog |71| Access tiers | Control who sees what | Role-based access (public, internal, restricted, confidential) |72| Data contracts | Define schema and quality guarantees | Schema tests, freshness SLAs |73| Version control | Track changes to definitions | dbt version control, migration scripts |74| Change management | Review process for metric changes | PR review for metric definition changes |7576### Dataset Tier System7778| Tier | Name | Access | Quality | Use Case |79|------|------|--------|---------|----------|80| Gold | Certified | All business users | High — tested, documented, SLA-backed | Self-service reporting and dashboards |81| Silver | Validated | Analysts and power users | Medium — tested, documented | Exploratory analysis, ad hoc queries |82| Bronze | Raw | Data team only | Low — as-is from source | Data engineering, debugging |83| Sandbox | Experimental | Individual user | None — user-created | Personal exploration, prototyping |8485## Step 4: Semantic Layer Design8687Build a consistent business logic layer between raw data and users:8889### Semantic Layer Components9091| Component | Purpose | Example |92|-----------|---------|---------|93| Metrics | Standardized measure definitions | `revenue = SUM(order_amount) WHERE status = 'completed'` |94| Dimensions | Standardized grouping attributes | `region`, `product_category`, `customer_segment` |95| Entities | Business objects with relationships | Customer, Order, Product, Campaign |96| Hierarchies | Drill-down paths | Country > Region > City |97| Time intelligence | Standard date calculations | YTD, QTD, MoM, YoY, rolling 30 days |98| Filters | Predefined filter sets | Active customers, current fiscal year |99100### Semantic Layer Tool Options101102| Tool | Approach | Integration |103|------|----------|-------------|104| dbt Metrics / MetricFlow | Code-defined, version-controlled | dbt ecosystem, BI tools via Semantic Layer API |105| Looker / LookML | Modeling language in BI tool | Native Looker, API access |106| Cube.js | Headless BI, API-first | Any BI tool, custom apps |107| AtScale | Virtual OLAP | Excel, Tableau, Power BI |108| Power BI Composite Models | In-tool semantic layer | Power BI ecosystem |109110## Step 5: User Training and Enablement111112Design a training program to build data literacy:113114### Training Curriculum115116| Module | Audience | Duration | Content |117|--------|----------|----------|---------|118| Data Foundations | All users | 1 hour | What is a database, table, metric; how data flows |119| Tool Training (Basic) | All users | 2 hours | Navigate dashboards, apply filters, export |120| Tool Training (Advanced) | Power users | 4 hours | Create charts, calculated fields, custom queries |121| SQL Fundamentals | Analysts | 8 hours | SELECT, JOIN, GROUP BY, window functions |122| Data Governance | All users | 1 hour | Access policies, PII handling, data classification |123| Metric Definitions | All users | 30 min | Where to find definitions, how to request new metrics |124125### Enablement Resources126127| Resource | Format | Update Cadence |128|----------|--------|----------------|129| Data dictionary | Wiki / catalog | On change |130| FAQ and troubleshooting | Knowledge base | Monthly |131| Office hours | Live session | Weekly |132| Slack channel | Async support | Always available |133| Video tutorials | Recorded walkthroughs | Quarterly |134| Template gallery | Pre-built reports/queries | Monthly |135| Champions network | Peer support program | Ongoing |136137## Step 6: Access Controls and Security138139Implement role-based access to protect data while enabling access:140141### Access Control Matrix142143| Role | Bronze (Raw) | Silver (Validated) | Gold (Certified) | Sandbox | Admin |144|------|-------------|-------------------|-------------------|---------|-------|145| Executive | No | View | View | No | No |146| Manager | No | View | View + Export | Create | No |147| Analyst | View | View + Query | View + Query + Export | Create | No |148| Power User | View | View + Query + Create | View + Query + Export | Create | No |149| Data Engineer | Full | Full | Full | Full | Yes |150| Data Steward | View | Full | Full | View | Partial |151152### Security Implementation153154| Security Layer | Mechanism | Tool |155|----------------|-----------|------|156| Authentication | SSO / SAML / OAuth | Okta, Azure AD, Google Workspace |157| Authorization | RBAC with group-based policies | Warehouse grants, BI tool permissions |158| Row-level security | Dynamic filters by user attribute | Row policies, user attributes in BI tool |159| Column masking | Redact or hash sensitive columns | Dynamic data masking, policy tags |160| Audit logging | Track who accessed what and when | Query logs, catalog usage analytics |161| Data classification | Tag PII, PHI, financial data | Catalog tags, automated classifiers |162163## Output Format164165Present the self-service analytics plan as:1661671. **Current State Assessment** (maturity level, pain points, opportunity)1682. **Target Architecture** (tools, layers, data flow diagram)1693. **Data Catalog Specification** (structure, tool choice, population plan)1704. **Governed Dataset Design** (tier system, certification criteria, access model)1715. **Semantic Layer Design** (metrics, dimensions, hierarchies, tool choice)1726. **Training and Enablement Plan** (curriculum, resources, timeline)1737. **Access Control Matrix** (roles, permissions, security layers)1748. **Rollout Roadmap** (phases, milestones, success metrics)175176## Quality Checklist177178Before delivering the self-service analytics plan, verify:179180- [ ] Current state maturity is assessed with evidence181- [ ] User segments and their needs are identified182- [ ] Data catalog covers all key data assets183- [ ] Governed datasets have clear tier definitions and certification criteria184- [ ] Semantic layer provides consistent metric definitions185- [ ] Training program addresses all user segments186- [ ] Access controls balance openness with security187- [ ] Success metrics for adoption are defined188- [ ] Rollout is phased with quick wins in the first phase189190## Edge Cases191192- **Highly regulated industries (healthcare, finance)**: Add data classification as a prerequisite; implement column-level masking before any self-service access193- **Very small data team**: Start with certified dashboards rather than full query access; use a lightweight catalog (dbt docs) over a full platform194- **Resistance to change**: Identify 3-5 champions in business units; start with their most painful report and automate it as a proof of concept195- **Multiple conflicting metric definitions**: Run a metric reconciliation workshop; establish a single source of truth before building the semantic layer196- **Legacy BI tools with limited governance**: Run the new and old tools in parallel during transition; provide side-by-side comparisons to build trust197- **No data warehouse**: This is a prerequisite; recommend a modern warehouse (Snowflake, BigQuery, Redshift) and a basic ELT pipeline before self-service
ashutoshsrivastava17