"A great dashboard doesn't show everything — it shows the next decision."
Learning Objectives
By the end of this module, you will be able to:
- Build core visualizations and dashboards in Tableau
- Model data and write measures in Power BI using DAX concepts
- Explore open-source BI with Metabase for quick team analytics
- Compare BI platforms and select the right tool for a given context
- Prepare datasets so BI tools stay fast, accurate, and trustworthy
Why BI Tools?
Spreadsheets don’t scale to shared truth. BI tools connect to databases, encode business logic in calculated fields, and publish interactive views to executives without emailing CSVs.
| Need | BI answer |
|---|---|
| One source of truth | Certified data connection |
| Self-serve slicing | Filters, parameters |
| Governance | Row-level security, permissions |
1. Tableau
Getting Started
Workbooks = sheets + dashboards. Dimensions slice; measures aggregate. Drag, drop, then refine.
Connecting Data
Live vs extract: extract speeds dashboards and reduces database load; live stays perfectly current. Blend only when you must — prefer joins in the database or dbt.
Building Visualizations
Sheets → dashboard layout. Use consistent colors and filters across tiles so the story feels like one app, not four charts.
| Viz goal | Tableau-ish move |
|---|---|
| Compare parts | Bar / sorted bar |
| Trend | Line with date on continuous axis |
| Filter deep dives | Context filters + dashboard actions |
| Explain one KPI | Reference line + annotation |
Calculated Fields
Row-level calcs vs aggregate calcs — Tableau’s grain matters. LOD expressions (FIXED, INCLUDE) solve “sum of max per customer” style puzzles.
Table Calculations
Running total, percent of table, rank — computed along a partition (e.g. within each region).
Building Dashboards
Layout containers, padding, and actions (filter on click) turn a grid into a workflow. Mobile layouts need fewer tiles and bigger text.
Try This! One dashboard: KPI strip + trend + breakdown; one filter applies to all three.
2. Power BI
Getting Started
Power Query shapes data; model defines relationships; report paints visuals.
Data Modeling — The Star Schema
Fact table (events, transactions) surrounded by dimension tables (product, date, customer). Avoid ambiguous many-to-many unless you know bridge patterns.
DAX (Data Analysis Expressions)
Measures evaluate in filter context. CALCULATE modifies that context — the Swiss Army knife.
| Concept | Plain English |
|---|---|
| Measure | Aggregates at query time |
| Column | Stored per row |
RELATED | Pull from dimension on the many side |
Power BI Report Design
Sync slicers, use bookmarks for storytelling modes, theme JSON for brand colors.
Concept: A bad model makes every DAX measure painful — invest in relationships early.
3. Metabase — Open Source BI
Setup
Docker or cloud; connect to Postgres, BigQuery, etc. Great for teams who want SQL + GUI.
Key Features
Questions (GUI or SQL), dashboards, pulses (scheduled email), collections for org.
Building in Metabase
Start with a question, add filters, pin to a dashboard. Keep SQL mode readable with comments.
SQL Mode with Variables
Template variables for dates or segments — safer than pasting literals.
4. Choosing the Right Tool
| Factor | Tableau | Power BI | Metabase |
|---|---|---|---|
| Sweet spot | Visual analytics culture | Microsoft shops | Fast SQL-first sharing |
| Licensing | Per user / role | Often bundled with M365 | OSS + paid hosting |
| Learning curve | Medium | Medium (DAX depth) | Low for SQL users |
Decision Framework
- Stack: Already on Azure/Office365? Lean Power BI.
- Analyst-heavy visuals: Tableau remains strong in exploration.
- Engineers + internal metrics: Metabase punches above its weight.
5. Preparing Data for BI Tools
| Practice | Why |
|---|---|
| Surrogate keys | Stable joins |
| Conformed dimensions | Same “Product” everywhere |
| Pre-aggregate huge facts | Snappier dashboards |
| Document grain | “One row per order line” |
Wide vs narrow: BI often likes star schemas; avoid 200-column monoliths without documentation.
Fun Fact: The best “BI fix” is sometimes a view in the warehouse, not a clever calculated field.
Practice Exercises
Exercise 1: Tableau Fundamentals (Beginner)
One workbook: bar, line, map or treemap, one calculated field.
Exercise 2: Power BI Data Model (Intermediate)
Star schema sketch + two measures with different filter behavior.
Exercise 3: Metabase Quick Dashboard (Intermediate)
SQL question with a parameter + small dashboard.
Exercise 4: Cross-Tool Comparison (Advanced)
Same metric in two tools; note differences in default aggregation.
Exercise 5: Embedded Analytics (Advanced)
Research embed tokens / SSO patterns for your stack.
Mini-Project: Executive KPI Dashboard
3–5 KPIs, trend, variance to target, drill-down dimension, mobile-friendly layout, short doc of data refresh and definitions.
Key Takeaways
- Model first (grain, keys, relationships), pretty second.
- Extract vs live is a performance vs freshness tradeoff.
- DAX and Tableau calcs both need you to think in context and grain.
- Metabase shines when SQL is the source of truth.
Resources for Further Learning
- Tableau Training Videos
- Microsoft Power BI Learn
- Metabase Documentation
- Kimball Dimensional Modeling
Key Takeaway
- Align tool choice with org stack, skills, and governance needs.
- Model facts and dimensions clearly before layering complex calcs.
- Publish extracts or aggregates when dashboards feel sluggish.
- Document metric definitions so “revenue” means the same in every chart.
- Iterate with real users — filters and actions beat extra charts.
Next up: Module 10 — Introduction to Big Data — when your laptop stops being enough.