Tableau

1. What Tableau Is

Key points

• Data visualization = showing data as pictures (bars, lines, maps), so people understand it fast.
• A dashboard = one screen with many charts and filters together. Managers use it to track the business.
• Tableau is used for reporting (what happened), analysis (why it happened) and sharing results with a team.
• It connects to Excel, CSV, SQL databases, cloud warehouses (Snowflake, BigQuery) and many more.
• Tableau is owned by Salesforce (since 2019).
• Its query language is called VizQL. When you drag fields, Tableau writes the database query for you.

The example data used in these notes

We will use a small "Superstore-style" Orders table. Each row is one product line in one order.

2. The Tableau Product Family

Key points

• Licences (user types): Creator = can build everything (includes Desktop and Prep). Explorer = can edit and explore published content. Viewer = can only look and use filters.
• Tableau Agent is the built-in AI helper (formerly Einstein Copilot). It can suggest charts and help write calculations from plain English.
• Tableau Next is Salesforce's newer "agentic" analytics product built on the Salesforce platform. Just know the name.
• Tableau Desktop Public Edition is the free desktop app. Your work is shared on Tableau Public, so never use private or company data there.

3. The Tableau Interface

Key points

• Data pane (left): every column (Tableau calls it a field) from your data. The Analytics tab next to it has trend lines, averages and forecasts.
• Columns shelf: fields here go across the chart (left to right). Rows shelf: fields here go down (top to bottom).
• Marks card: controls how each mark looks. A mark is one drawn item: one bar, one dot, one map shape.
  • Color, Size, Label (text on the mark), Detail (split marks without changing the look), Tooltip (pop-up on hover), Shape, Path (line order).
  • The mark type drop-down: Automatic, Bar, Line, Area, Square, Circle, Shape, Text, Map, Pie, Gantt Bar, Polygon, Density.
• Filters shelf: drop a field here to keep only some data. Pages shelf: makes a "flip-book" animation, one page per value.
• Show Me: suggests chart types for the fields you selected. Grey charts need different fields; it tells you what is missing.
• A field you drop on a shelf becomes a pill (a rounded tag).

4. Workbooks, Worksheets, Dashboards and Stories

Key points

• Viz (short for visualization) = any chart Tableau draws.
• You make new sheets with the icons at the bottom tab bar: New Worksheet, New Dashboard, New Story.
• One worksheet can be used in many dashboards. Change it once, and every dashboard updates.
• The Data Source tab (bottom-left) is where you connect and prepare data.

5. Live vs Extract, and File Types

How to connect (step by step)

1. Open Tableau. On the start page, under Connect, pick the source: Microsoft Excel, Text file (CSV), or a server like MySQL.
2. Choose the file (or type the server name, user and password).
3. You land on the Data Source page. Drag the sheet or table into the canvas.
4. At the top-right, choose Live or Extract.
5. Click the Sheet 1 tab to start building charts.

File types you must know

• Extract refresh: Full = rebuild the whole copy. Incremental = add only new rows (based on a column like Order Date or an ID).
• Extract filters can keep the extract small, e.g. only the last 2 years.

6. Joins

Example tables

Orders has CA-1001, CA-1002, CA-1003. Returns has CA-1002 and CA-9999 (a typo order that is not in Orders).

How to make a join (step by step)

1. On the Data Source page, drag Orders to the canvas.
2. Double-click the Orders box. This opens the physical layer (the join area).
3. Drag Returns next to Orders. A Venn icon (two circles) appears.
4. Click the Venn icon. Pick the join type (Inner, Left, Right, Full Outer).
5. Set the join clause: Order ID = Order ID.

• Duplication problem: if one order has 3 rows in table B, the join repeats the order 3 times. Then SUM(Sales) becomes too big. Always check totals after a join.
• Null = an empty value. Left/right/full joins create nulls where there is no match.
• You can join on more than one column, and use =, <, > and so on. You can also use a calculation in a join clause.
• Cross-database join = join tables from two different sources (e.g. Excel + MySQL).

7. Relationships and Blending

Key points about relationships

• The Data Source page has two layers. Logical layer (top level) = relationships. Physical layer (inside a table) = joins and unions.
• You don't pick a join type. Tableau uses the right one per chart (this is called a context-aware join).
• Each table keeps its own level of detail, so sums are not duplicated.
• Unmatched rows are kept by default, so you don't lose data silently.
• Optional "Performance options" let you set cardinality (one / many) and referential integrity. Leave them alone unless you are sure.

Data blending (the older way)

• Blending combines data from two different data sources inside one worksheet.
• The first source you use is the primary (blue tick). The other is secondary (orange tick).
• They are linked on a common field. An orange link icon shows which field links them. You can click it to turn the link on or off.
• It works like a left join from primary to secondary, done after each source is aggregated.
• Limits: secondary fields are always aggregated, and some functions (like COUNTD on secondary) don't work well. Use it when the sources can't be related, e.g. data at different levels like daily sales vs monthly targets.

8. Unions, Data Interpreter, Pivot and Split

Union (stack rows)

• Use it when files have the same columns: Sales_2024.csv and Sales_2025.csv.
• Join = adds columns (side by side). Union = adds rows (top to bottom).
• Tableau adds a Table Name (or Sheet) column so you know where each row came from.
• Wildcard union = union all files matching a pattern, like Sales_*.csv.

1. Drag the first table to the canvas and double-click it (physical layer).
2. Drag the second table under the first until it says "Drag table to union". Drop it.
3. Or right-click the table → Convert to Union to set a wildcard pattern.

Data Interpreter

• Excel files often have titles, blank rows and notes above the real table. Tick Use Data Interpreter on the Data Source page.
• Tableau finds the real header row and removes the extra stuff. Click "Review the results" to check what it did.

Pivot (wide to long)

Data like this is "wide" and hard to chart:

1. On the Data Source page, select the Jan, Feb, Mar columns (Ctrl/Cmd-click).
2. Right-click → Pivot.
3. You get two columns: Pivot Field Names (Jan, Feb, Mar) and Pivot Field Values (500, 800, 650). Rename them to Month and Sales.

Split

• Right-click a column like Customer = "Asha Rao" → Split. Tableau makes "Asha" and "Rao" columns.
• Custom Split lets you choose the separator (e.g. - or ,) and which parts to keep.

9. Dimensions vs Measures

Key points

• Dimensions are usually text, dates or IDs. They slice the data into groups.
• Measures are numbers that can be aggregated (summed, averaged, counted).
• In the data pane, dimensions are listed at the top, measures below a line.
• Tableau guesses the role when it connects. You can change it: right-click the field → Convert to Dimension / Convert to Measure.
• Example: Postal Code is a number, but you never add postal codes. So make it a dimension.
• Measure Names and Measure Values are special fields. They let you show many measures in one chart (e.g. Sales, Profit and Quantity side by side in a text table).

10. Discrete vs Continuous (Blue vs Green)

Key points

• Dimension/measure and discrete/continuous are two different ideas. Don't mix them up.
• Usually: dimensions are discrete (blue) and measures are continuous (green). But not always.
• A measure can be discrete. Example: make SUM(Sales) discrete and it shows as header text instead of an axis.
• A date can be both. Right-click a date pill: the top list (YEAR, QUARTER, MONTH) gives date parts (blue, discrete). The lower list gives date values (green, continuous, e.g. a timeline of each month).
• To switch: right-click the pill → Discrete or Continuous.
• Colour legend: discrete field on Color gives separate colours; continuous field gives a colour gradient (light to dark).

11. Aggregation

Key points

• Default aggregation for measures is SUM. Change it: right-click the pill → Measure (Sum) → pick another.
• Common ones: SUM, AVG, MIN, MAX, COUNT, COUNTD (count distinct), MEDIAN.
• COUNT([Order ID]) counts rows. COUNTD([Order ID]) counts unique orders. One order with 3 products: COUNT = 3, COUNTD = 1.
• ATTR([Field]) shows the value if all rows in the group have the same value, otherwise *.
• The level of detail of a chart = the dimensions in it. "Sales by Region" has 4 marks, one per region.
• Disaggregate: menu Analysis → untick Aggregate Measures. Now you see one mark per row. Used for some scatter plots.

Data: East rows have Sales 500 and 800.

SUM(Sales)   for East = 1300
AVG(Sales)   for East = 650
COUNT(Sales) for East = 2

12. Bar, Line, Text Table and Highlight Table

Bar chart: Sales by Category

1. Drag Category to Columns.
2. Drag Sales to Rows. Tableau draws bars with SUM(Sales).
3. Click the sort icon in the toolbar to sort high to low.
4. Drag Region to Color for a stacked bar. Drag Region to Columns instead for side-by-side bars.
5. Drag Sales to Label to show numbers on the bars.

Line chart: Sales over time

1. Drag Order Date to Columns. It shows YEAR(Order Date).
2. Right-click the pill → choose the green Month (date value, e.g. "May 2025") for a continuous timeline.
3. Drag Sales to Rows. You get a line.
4. Drag Category to Color for one line per category.

Text table (crosstab)

1. Drag Region to Rows and Category to Columns.
2. Drag Sales to Text on the Marks card.
3. Menu Analysis → Totals → Show Row / Column Grand Totals.

Highlight table

1. Build the text table above.
2. Drag Sales (or Profit) to Color. Change mark type to Square.
3. Now big numbers have darker cells. Use a diverging colour (red–green) for Profit, so losses stand out.

13. Maps and Scatter Plots

Maps: how to build

1. Make sure the location field has a geographic role. Right-click State → Geographic Role → State/Province. A globe icon appears.
2. Double-click State. Tableau adds Latitude (generated) and Longitude (generated) and draws a map.
3. Symbol map: drag Sales to Size. Bigger circle = more sales.
4. Filled map (choropleth): change mark type to Map and drag Profit to Color. Each state is shaded.

• If some places are not found, click the "unknown" note at the bottom-right of the map and fix the locations.
• You can also use your own latitude/longitude columns, or spatial files (shapefile, GeoJSON).
• Map layers let you stack several layers (e.g. filled states plus city dots) on one map.

Scatter plot: Sales vs Profit by Customer

1. Drag Sales to Columns and Profit to Rows. You get just one dot (the grand total).
2. Drag Customer to Detail. Now there is one dot per customer.
3. Drag Category to Color.
4. Optional: Analytics pane → drag Trend Line onto the view.

14. Heatmap, Treemap, Histogram and Box Plot

Heatmap: Sales by Month and Weekday

1. Drag Order Date to Columns, set it to Month (discrete, blue).
2. Drag Order Date to Rows, set it to Weekday (More → Weekday).
3. Mark type Square. Drag Sales to Color.
4. Dark squares show busy days. (Mark type Density makes a "glow" heatmap for dense scatter or map points.)

Treemap: Sales by Sub-Category

1. Drag Sub-Category to Label (or Detail).
2. Drag Sales to Size and Profit to Color. Mark type Square.
3. Big rectangle = big sales. Red = loss. Or select the fields and click Treemap in Show Me.

Histogram: how big are our orders?

1. Click Sales in the data pane, open Show Me, pick Histogram.
2. Tableau creates a new field Sales (bin) and puts CNT(Sales) on Rows.
3. Edit the bin size: right-click Sales (bin) → Edit → e.g. size 100.
4. Result: how many order lines fall in 0–100, 100–200, and so on.

Box plot (box-and-whisker)

1. Drag Category to Columns, Sales to Rows.
2. Drag Order ID to Detail, so there is one dot per order. Mark type Circle.
3. Analytics pane → drag Box Plot onto the view (or pick box-and-whisker in Show Me).

• The box covers the middle 50% of values (from the 25th to the 75th percentile). The line inside is the median.
• The whiskers reach out to 1.5 × the box height by default. Dots past them are outliers (unusual values).

15. Dual-Axis and Combo Charts

How to build Sales (bars) + Profit (line) by Month

1. Drag Order Date to Columns and set it to continuous Month.
2. Drag Sales to Rows. Drag Profit to Rows too. You now have two charts stacked.
3. Right-click the SUM(Profit) pill → Dual Axis. The charts merge.
4. On the Marks card there is now one tab per measure. Set SUM(Sales) to Bar and SUM(Profit) to Line.
5. If both use the same unit, right-click the right axis → Synchronize Axis.

• Synchronize only works when both measures have the same data type (both whole numbers or both decimals).
• Different scales can mislead. Always label both axes clearly.
• Blended axis (shared axis) is different: drag a second measure onto the existing axis. Both share one axis, using Measure Names/Values.
• Dual axis is also used for donut charts (pie + a smaller white circle on top) and map layers.

16. Gantt and Bullet Charts

Gantt chart: how long does shipping take?

1. Create a field Days to Ship: DATEDIFF('day', [Order Date], [Ship Date]).
2. Drag Order Date to Columns (continuous, exact Day).
3. Drag Order ID to Rows.
4. Set mark type to Gantt Bar. Drag Days to Ship to Size. Each bar starts at the order date and its length is the shipping time.

Bullet chart: Sales vs Target

1. You need two measures: Sales (actual) and Target.
2. Select both in the data pane (Ctrl/Cmd-click), plus Region. In Show Me, click Bullet Graph.
3. The bar is actual sales. A reference line marks the target. Shaded bands show e.g. 60% and 80% of target.
4. If Tableau puts them the wrong way round, right-click the axis → Swap Reference Line Fields.

17. Filters and the Order of Operations

Types of filters

Key points

• Show Filter: right-click a filter pill → Show Filter. Users get a control (list, drop-down, slider) on the sheet.
• Dimension filters have tabs: General (pick values), Wildcard (text match), Condition (e.g. Sales > 1000), Top (Top N).
• Date filters: relative dates (last 3 months), range of dates, or discrete parts (year = 2025).
• Classic interview trap: You set "Top 10 customers by Sales" and also filter Region = East. You get fewer than 10 rows! Why? Top N runs before the Region filter. Fix: add Region to context.
• Filters can apply to one sheet, several sheets, or all sheets using the same data source (right-click → Apply to Worksheets).
• "Only relevant values" on a filter control shows only values allowed by other filters. It is handy but slower.

18. Sets, Groups and Bins

Groups

1. Right-click Sub-Category → Create → Group.
2. Select "Phones" and "Accessories" → click Group. Name it "Mobile".
3. Tick "Include Other" to put everything else in one "Other" group.

• A new field appears with a paperclip icon. You can also select marks in a chart and click the paperclip to group them.
• Groups are good for fixing spelling variants: "Bengaluru", "Bangalore" → one value.

Sets

• Fixed set: you pick members by hand. It doesn't change when data changes.
• Dynamic set: built from a condition (e.g. SUM(Sales) > 5000) or a Top N rule (top 10 customers by sales). It updates when data changes.
• Use a set on Color to see "Top customers vs everyone else", or on Filters to show only IN members.
• Combined set: join two sets on the same dimension (all members, shared members, or except shared).
• Set controls let users add or remove members from a drop-down. Set actions change set members by clicking a chart (see Dashboard Actions).

1. Right-click Customer → Create → Set.
2. Go to the Top tab → By field: Top 10 by Sales Sum.
3. Drag the new set to Color. You see IN (top 10) vs OUT.

Bins

• Right-click a measure like Sales → Create → Bins. Choose a bin size (e.g. 100).
• The bin field is a dimension. Used for histograms or customer age bands.

19. Hierarchies and Sorting

Hierarchies: how to make one

1. In the data pane, drag State on top of Country. Name the hierarchy "Location".
2. Drag City under State in the hierarchy.
3. Drag "Country" to a shelf. A small + icon appears on the pill. Click it to drill down to State; click − to roll up.

• Dates have a built-in hierarchy: Year → Quarter → Month → Day.
• Common example: Category → Sub-Category → Product Name.

Sorting

• Quick sort: click the sort icon on an axis or the toolbar.
• Field sort: right-click a dimension pill → Sort. Options: Data source order, Alphabetic, Field (by another measure, e.g. by SUM(Sales)), Manual (drag into your own order), Nested.
• Nested sort sorts inside each group. Example: sub-categories sorted by sales within each category.
• Manual sort is "frozen": it won't change when data changes. Sorting by a field is "computed" and updates with the data.

20. Calculated Fields

How to create one

1. Menu Analysis → Create Calculated Field (or right-click in the data pane → Create → Calculated Field).
2. Give it a name, like Profit Ratio.
3. Type the formula. Field names go in square brackets: [Sales].
4. The bottom of the editor says "The calculation is valid" or shows the error. Click OK.

Basic syntax

// Comments start with two slashes
[Sales] - [Profit]                      // row-level math

// Ratio: aggregate FIRST, then divide
SUM([Profit]) / SUM([Sales])            // correct profit ratio

// IF ... THEN ... ELSEIF ... ELSE ... END
IF [Profit] > 0 THEN "Profitable"
ELSEIF [Profit] = 0 THEN "Break-even"
ELSE "Loss"
END

// CASE: compare one field to fixed values
CASE [Region]
    WHEN "East" THEN "Zone A"
    WHEN "West" THEN "Zone B"
    ELSE "Zone C"
END

// IIF: short IF (test, then, else)
IIF([Sales] >= 1000, "Big order", "Small order")

// Handling nulls (empty values)
ZN(SUM([Sales]))                        // null becomes 0
IFNULL([Discount], 0)                   // if null, use 0
ISNULL([Ship Date])                     // TRUE if empty

// Logic
[Region] = "East" AND [Profit] < 0       // TRUE or FALSE

Key points

• Calculations can be row-level (run on every row, like [Sales] - [Profit]) or aggregate (run on groups, like SUM([Sales])).
• You can't mix them in one expression. [Sales] / SUM([Sales]) gives the error "Cannot mix aggregate and non-aggregate arguments". Fix: wrap both, or use an LOD.
• Why SUM([Profit]) / SUM([Sales]) and not [Profit] / [Sales]? The second makes a ratio per row, and then Tableau sums those ratios. Adding percentages gives nonsense.
• Strings use double or single quotes. Every IF and CASE needs an END.
• CASE is often faster and cleaner than many ELSEIFs when checking one field against fixed values.
• Calculated fields show an = sign next to their icon in the data pane.

21. String and Date Functions

String functions

LEFT([Order ID], 2)                      // "CA"
RIGHT([Order ID], 4)                     // "1001"
MID([Order ID], 4, 2)                    // "10"  (start at position 4, take 2)
LEN([Customer])                          // number of characters
UPPER([Region])                          // "EAST"
TRIM([Customer])                         // remove spaces at both ends
CONTAINS([Product], "Chair")             // TRUE / FALSE
REPLACE([Region], "Central", "Middle")   // swap text
SPLIT([Customer], " ", 1)                // "Asha"  (first part before a space)
[Region] + " - " + [Category]            // join text: "East - Furniture"
STR([Sales]) + " INR"                    // turn a number into text first

Date functions

// Example: [Order Date] = 2025-02-14, [Ship Date] = 2025-02-16

DATEPART('year', [Order Date])           // 2025       (a number)
DATEPART('month', [Order Date])          // 2          (a number)
DATENAME('month', [Order Date])          // "February" (text)
DATETRUNC('month', [Order Date])         // 2025-02-01 (date, rounded down)
DATEDIFF('day', [Order Date], [Ship Date])   // 2      (days between)
DATEADD('month', 3, [Order Date])        // 2025-05-14
YEAR([Order Date])                       // 2025 (shortcut for DATEPART year)
TODAY()                                  // today's date
NOW()                                    // date and time now
MAKEDATE(2025, 1, 1)                     // builds a date
DATEPARSE('dd/MM/yyyy', [Date Text])     // turn text into a date

Key points

• DATEPART returns a number. DATENAME returns text. DATETRUNC returns a date.
• Date parts are written in quotes: 'year', 'quarter', 'month', 'week', 'day', 'hour'.
• DATEDIFF order is (part, start, end). If end is earlier than start, the result is negative.
• Year-to-date example: [Order Date] <= TODAY() AND YEAR([Order Date]) = YEAR(TODAY()).
• A text column like "14/02/2025" is not a date yet. Change its type (click the Abc icon → Date) or use DATEPARSE.

22. LOD Expressions (FIXED, INCLUDE, EXCLUDE)

Syntax

{ FIXED [dimension] : AGG([measure]) }
{ INCLUDE [dimension] : AGG([measure]) }
{ EXCLUDE [dimension] : AGG([measure]) }
{ SUM([Sales]) }        // no keyword = whole table (same as FIXED with no dimension)

Simple examples

// FIXED: each customer's first order date (cohort analysis)
{ FIXED [Customer] : MIN([Order Date]) }

// FIXED: sales per region, whatever the chart shows
{ FIXED [Region] : SUM([Sales]) }

// Percent of region: each category's share of its region's total
SUM([Sales]) / MIN({ FIXED [Region] : SUM([Sales]) })

// INCLUDE: average sales per customer, shown by Region
//   view = Region; first sum per customer, then AVG on the shelf
AVG({ INCLUDE [Customer] : SUM([Sales]) })

// EXCLUDE: region total next to each category in a Region + Category chart
{ EXCLUDE [Category] : SUM([Sales]) }

Why MIN() around the FIXED part? A FIXED result is copied onto every row of that region. SUM() would add those copies together and give a number that is too big. MIN() (or MAX() / AVG()) just picks the one region total.

Key points

• FIXED ignores the view's dimensions. It is computed before dimension filters, so normal filters don't change it. Context filters and data source filters do.
• INCLUDE and EXCLUDE are relative to the view and are computed after dimension filters.
• FIXED results can be used as dimensions (e.g. "first order year" for cohorts). INCLUDE/EXCLUDE results are measures.
• Common uses: customer first purchase date, percent of total, compare each item to its group's average, count customers by number of orders.
• Count customers by order count: { FIXED [Customer] : COUNTD([Order ID]) }, then drag it to Columns as a dimension and COUNTD([Customer]) to Rows.

23. Table Calculations

Quick table calculations (the easy way)

1. Build a chart: MONTH(Order Date) on Columns, SUM(Sales) on Rows.
2. Right-click the SUM(Sales) pill → Quick Table Calculation.
3. Pick one: Running Total, Difference, Percent Difference, Percent of Total, Rank, Percentile, Moving Average, YTD Total, Compound Growth Rate, Year over Year Growth, YTD Growth.
4. A small triangle (Δ) appears on the pill. It means "table calculation".

Worked example

Writing them yourself

RUNNING_SUM(SUM([Sales]))                    // running total
SUM([Sales]) / TOTAL(SUM([Sales]))           // percent of total
RANK(SUM([Sales]))                           // 1 = highest (descending by default)
SUM([Sales]) - LOOKUP(SUM([Sales]), -1)      // difference from previous mark
(SUM([Sales]) - LOOKUP(SUM([Sales]), -1))
    / ABS(LOOKUP(SUM([Sales]), -1))          // percent change
WINDOW_AVG(SUM([Sales]), -2, 0)              // 3-month moving average (2 back + current)
WINDOW_SUM(SUM([Sales]))                     // total of all marks in the window
INDEX()                                      // 1, 2, 3 ... position of each mark
FIRST()  /  LAST()                           // offset from the first / last mark

Compute Using (direction) — the simple version

• A table calc needs to know which way to move through the marks. Right-click the pill → Compute Using.
• Table (across) = left to right. Table (down) = top to bottom. Pane = inside each pane only. Cell = each cell alone.
• Specific Dimensions: you tick which dimensions to move along.
• Addressing = the dimensions it moves along (e.g. Month). Partitioning = the dimensions that make separate groups where it restarts (e.g. Region). Example: running total along Month, restarting for each Region.
• Table calcs run late (after measure filters). So filtering a table calc only hides marks; it does not change results.

24. Parameters

Example 1: Top N with a slider

1. Right-click in the data pane → Create Parameter. Name: Top N. Data type: Integer. Allowable values: Range 1 to 20. Current value: 10.
2. Right-click the parameter → Show Parameter. A slider appears.
3. Drag Customer to Filters → Top tab → By field: Top Top N by SUM(Sales).
4. Now moving the slider changes how many customers are shown.

Example 2: "Choose a measure"

1. Create a parameter Choose Measure. Data type: String. Allowable values: List → add "Sales", "Profit", "Quantity".
2. Create a calculated field Selected Measure:

CASE [Choose Measure]
    WHEN "Sales"    THEN SUM([Sales])
    WHEN "Profit"   THEN SUM([Profit])
    WHEN "Quantity" THEN SUM([Quantity])
END

3. Put Selected Measure on Rows and Category on Columns.
4. Show the parameter control. Users pick a measure and the chart changes.
5. Tip: put the parameter in the title, like "<Choose Measure> by Category", using Insert in the title editor.

Key points

• Allowable values: All (type anything), List (fixed choices), Range (min, max, step).
• Parameters are global: one parameter can be used in any sheet of the workbook, even across data sources.
• Since 2020.1, a list parameter can refresh its values from a field when the workbook opens ("dynamic parameter").
• Parameter actions let users set a parameter by clicking on a chart.
• Filter vs parameter: a filter is tied to one field. A parameter is one free value you can use anywhere (in calcs, Top N, reference lines, bins).

25. Dashboards and Layout

How to build a dashboard

1. Click New Dashboard (bottom tab bar).
2. Set Size on the left: Fixed (e.g. 1200 × 800; looks the same everywhere), Automatic (fits any screen, may squash), or Range (min and max).
3. Drag a Horizontal container from Objects to the top. Drag KPI sheets into it.
4. Drag sheets onto the canvas below. Grey shading shows where they will drop.
5. Use a sheet's drop-down → Use as Filter (funnel icon) to make clicks on it filter the others.
6. Add a Text object for the title. Use Distribute Contents Evenly on a container to make items equal width.

Key points

• Tiled = items snap into a grid and never overlap. This is the default and the safest.
• Floating = items sit on top at an exact position and size. Good for legends, logos, small notes. Hold Shift while dragging to float.
• Objects you can add: Horizontal, Vertical, Text, Image, Web Page, Blank, Navigation (button to another sheet), Download (PDF, image, crosstab), Extension.
• Device layouts: click Device Preview to create Desktop, Tablet and Phone layouts. The phone layout can show fewer charts, stacked vertically.
• Good design: most important number top-left, 3–5 charts max, consistent colours, clear titles that say the message.
• The Item hierarchy in the Layout tab shows how containers are nested. Use it to select hard-to-click containers.

26. Dashboard Actions

How to add a filter action

1. On the dashboard, menu Dashboard → Actions → Add Action → Filter.
2. Source sheet: the map. Target sheets: the line chart and the table.
3. Run action on: Hover, Select (click), or Menu (a link in the tooltip).
4. Clearing the selection will: Show all values / Keep filtered values / Exclude all values.
5. Click OK and test by clicking the map.

27. Stories, Formatting and Tooltips

Stories: how to build

1. Click New Story. Drag a sheet or dashboard into the story area.
2. Write a short caption in the box at the top, like "West sales grew 30% in Q2".
3. Click New Blank Point or Duplicate for the next point.
4. You can filter or highlight differently in each point, then click Update to save it.

Formatting tips

• Menu Format → Workbook sets fonts and lines for everything at once.
• Right-click any part → Format to change only that part (axis, borders, shading).
• Number format: right-click a measure → Default Properties → Number Format (currency, %, decimals, K/M units).
• Remove clutter: hide gridlines, use fewer colours, label directly instead of using legends.
• Edit titles to say the insight, not just the field names.

Tooltips and Viz in Tooltip

• Click Tooltip on the Marks card to edit the text. Insert fields with the Insert menu, like <SUM(Sales)>.
• Drag extra fields to the Tooltip shelf to show them on hover without changing the chart.
• Viz in Tooltip: show a small chart inside the tooltip. In the tooltip editor, click Insert → Sheets → choose a sheet.

<Sheet name="Monthly Trend" maxwidth="300" maxheight="300" filter="<All Fields>">

Now hovering over "East" shows the monthly trend for East only. The filter part passes the hovered mark's values to the small chart.

28. Trend Lines, Reference Lines and Forecasting

What is in the Analytics pane

• Constant Line: a fixed value, e.g. a target of 1,000.
• Average Line, Median with Quartiles, Totals.
• Reference Line: line at a value (average, median, min, max, constant or a parameter). Can be per table, pane or cell.
• Reference Band: shaded area between two values. Distribution Band: e.g. 60%–80% of average.
• Trend Line: best-fit line. Models: linear, logarithmic, exponential, polynomial, power. Hover to see the equation, R-squared (how well the line fits, 0 to 1) and p-value (lower than 0.05 usually means the trend is real, not chance).
• Forecast: predicts future values using exponential smoothing. Needs a date on the view and enough history.
• Cluster: groups similar marks automatically (k-means).
• Box Plot: shows median, quartiles and outliers.

How to add a forecast

1. Build a line chart: continuous MONTH(Order Date) on Columns, SUM(Sales) on Rows.
2. Analytics pane → drag Forecast onto the view.
3. The forecast appears in a lighter colour with a shaded band. The band is the prediction interval (95% by default): the likely range.
4. Right-click → Forecast → Forecast Options to set how far ahead (e.g. 6 months) and the model.

29. Performance Tips

Key points

• Use extracts for slow sources. Use extract filters and Hide All Unused Fields to make them small.
• Aggregate the extract to visible dimensions (e.g. monthly, not every row) if you don't need row detail.
• Fewer marks: a table with 100,000 cells is slow and nobody reads it. Show summaries; drill into details with actions.
• Fewer quick filters, especially with "Only relevant values" or long lists. Use filter actions or parameters instead.
• Fewer sheets per dashboard. Each sheet is at least one query.
• Calculations: numbers and Booleans are faster than strings. Prefer CASE or IN over long IF … OR … chains. Move heavy logic to the database or Prep.
• Use relationships rather than big joins when tables are at different levels.
• Use context filters to shrink data early for slow Top N or FIXED calcs (but not too many).
• Use fixed dashboard size so results can be cached and reused.

How to find the slow part

1. Menu Help → Settings and Performance → Start Performance Recording.
2. Use the dashboard: click filters, change sheets.
3. Stop the recording. Tableau opens a workbook showing which events (queries, layout, rendering) took the longest.
4. Also try Workbook Optimizer (Server menu → Run Optimizer). It checks your workbook against best-practice rules.

30. Publishing and Sharing

How to publish (Server or Cloud)

1. Menu Server → Sign In to your Server or Cloud site.
2. Server → Publish Workbook. Pick a project (a folder) and a name.
3. Set permissions (or keep the project's defaults).
4. Choose how data is published: embedded in the workbook, or as a separate published data source others can reuse.
5. Set authentication for the data (embed password, or prompt users).
6. Click Publish. For Tableau Public: File → Save to Tableau Public.

Key points

• Site = a separate space on a server for one team or company. Project = a folder that holds workbooks and data sources.
• Site roles: Creator, Explorer, Viewer (plus admin roles). They set the maximum a user can do.
• Permissions are set on projects, workbooks and data sources: view, filter, download, web edit, and so on. Best practice: give permissions to groups, not single users, at the project level.
• Refresh schedules: published extracts refresh on a schedule (e.g. daily at 6 am). Full or incremental.
• Tableau Bridge: lets Tableau Cloud refresh from data inside your company network (on-premises databases).
• Subscriptions email a snapshot on a schedule. Data-driven alerts email you when a number crosses a threshold.
• Row-level security (RLS): each user sees only their own rows. Simple way: a calculated filter like [Sales Rep Email] = USERNAME(), or ISMEMBEROF("East Team").
• Tableau Pulse (Cloud): define a metric once (e.g. weekly Sales); users follow it and get AI insight digests.
• Tableau Public is free but everything is public. Never upload private or company data.

31. Tableau vs Power BI

32. Build a Sales Dashboard (End to End)

Step 1 — Connect and prepare

1. Connect to Superstore.xlsx. Drag Orders to the canvas. Relate Returns on Order ID and People on Region.
2. Check data types (dates are dates, Postal Code is a string dimension). Set State geographic role.
3. Choose Extract for speed.

Step 2 — Create calculations

// Profit Ratio
SUM([Profit]) / SUM([Sales])

// Is Returned?
IF ISNULL([Returned]) THEN "No" ELSE "Yes" END

// Sales Current Year (for KPI)
IF YEAR([Order Date]) = { MAX(YEAR([Order Date])) } THEN [Sales] END

// Sales Previous Year
IF YEAR([Order Date]) = { MAX(YEAR([Order Date])) } - 1 THEN [Sales] END

// Year-over-year growth
(SUM([Sales Current Year]) - SUM([Sales Previous Year]))
    / SUM([Sales Previous Year])

Step 3 — Build the sheets

1. KPI cards: one sheet each for Sales, Profit, Profit Ratio, Orders (COUNTD([Order ID])). Put the measure on Text, make it big. Add YoY growth below it.
2. Map: State, filled map, Color = Profit Ratio (diverging red–blue).
3. Trend: continuous Month of Order Date vs Sales, with a Forecast or an average reference line.
4. Category bars: Sub-Category by Sales, sorted, Color = Profit (losses show red).
5. Top 10 customers: a Top N filter driven by a Top N parameter.

Step 4 — Assemble the dashboard

1. New dashboard, Fixed size 1200 × 800.
2. Title text at the top. Horizontal container with the 4 KPI cards.
3. Map and trend in a vertical container on the left; category bars and top customers on the right.
4. Add filters: Region, Category, and a relative date filter. Apply them to all sheets using this data source.

Step 5 — Make it interactive and share

1. Map → Use as Filter (filter action): click a state to filter everything.
2. Highlight action on Category. Viz in Tooltip on the map showing the monthly trend.
3. Add a Phone layout. Run the Performance Recorder.
4. Publish to Tableau Cloud (or Public for a portfolio) and schedule a daily extract refresh.

Quick Revision

The most-asked Tableau questions, with short and simple answers. Say these in your own words.

Basics

• What is Tableau? A drag-and-drop tool that connects to data and makes interactive charts and dashboards.
• Name the Tableau products. Desktop (build), Prep (clean), Server and Cloud (share inside a company), Public (free, public), Pulse (metrics with AI insights).
• Server vs Cloud? Server: your company hosts it. Cloud: Salesforce hosts it.
• .twb vs .twbx? .twb = design only. .twbx = design plus data, packed into one file.
• What is a .hyper file? An extract: a fast, compressed snapshot of data.
• Live vs extract? Live = always fresh, depends on database speed. Extract = fast snapshot, refreshed on a schedule.
• Worksheet vs dashboard vs story? One chart; many charts on one screen; a sequence of slides.

Data

• Join vs union? Join adds columns (side by side). Union adds rows (stacked).
• Join vs relationship? Join merges tables early into one; rows can duplicate. Relationship keeps tables separate and joins per chart at the right level.
• What is data blending? Combining two data sources in one sheet, on a linking field. Works like a left join after aggregation; primary vs secondary source.
• What does Data Interpreter do? Cleans messy Excel/CSV: finds headers, removes titles and blank rows.
• What is pivot? Turning many columns (Jan, Feb, Mar) into rows (Month, Value).

Core ideas

• Dimension vs measure? Dimension = groups and labels. Measure = numbers you aggregate.
• Discrete vs continuous? Discrete (blue) = headers. Continuous (green) = axis.
• COUNT vs COUNTD? COUNT counts rows. COUNTD counts unique values.
• What does ATTR do? Shows the value if it is the same for all rows in the group, otherwise *.
• What is the Marks card? It controls how marks look: Color, Size, Label, Detail, Tooltip, Shape, Path.

Filters & organising

• Filter order of operations? Extract → Data source → Context → Dimension → Measure → Table calc.
• What is a context filter? A filter that runs before others, so Top N and FIXED respect it. Its pill is grey.
• Group vs set? Group merges values into one. Set splits data into IN and OUT.
• Fixed vs dynamic set? Fixed = hand-picked members. Dynamic = based on a condition or Top N, updates with data.
• What is a bin? Equal-sized ranges of a number, used for histograms.

Calculations

• Why SUM([Profit])/SUM([Sales]) and not [Profit]/[Sales]? The second sums row ratios, which is wrong.
• "Cannot mix aggregate and non-aggregate" error? Wrap the row-level part in an aggregate, or use an LOD.
• DATEPART vs DATETRUNC? DATEPART returns a number (2). DATETRUNC returns a date rounded down (2025-02-01).
• ZN vs IFNULL? ZN turns null into 0. IFNULL turns null into any value you give.
• FIXED vs INCLUDE vs EXCLUDE? FIXED = exact level you name. INCLUDE = finer than the view. EXCLUDE = coarser than the view.
• LOD vs table calc? LOD runs in the database on raw data. Table calc runs on the numbers already in the view.
• What is Compute Using? The direction a table calc moves through the marks (across, down, pane, specific dimensions).
• Parameter vs filter? A filter is tied to a field. A parameter is a free user input used in calcs, filters or lines.

Dashboards & sharing

• Tiled vs floating? Tiled snaps into a grid without overlap. Floating sits on top at an exact position.
• Dashboard action types? Filter, Highlight, Go to URL, Go to Sheet, Change Parameter, Change Set Values.
• What is Viz in Tooltip? A small chart shown inside a tooltip on hover.
• How do you speed up a slow dashboard? Extract, hide unused fields, fewer marks, fewer filters, fewer sheets; check with the Performance Recorder.
• How do you do row-level security? A user filter or calculated filter like [Rep Email] = USERNAME().
• What is Tableau Pulse? A Tableau Cloud feature that tracks metrics and sends AI-written insights.

Learning Path

Go step by step. Download free Tableau Public and the Superstore sample data, and build something at each step.

1. The interface — data pane, shelves, Marks card, Show Me. Make 5 basic charts.
2. Core ideas — dimensions vs measures, discrete vs continuous, aggregation.
3. Connecting data — Excel and CSV, live vs extract, relationships, joins, unions, pivot.
4. More charts — maps, scatter, dual axis, histogram, treemap, Gantt, bullet.
5. Filters — all filter types and the order of operations. Practise the Top N + context trap.
6. Calculations — IF/CASE, date functions, then table calcs, then LOD expressions.
7. Parameters — build the "choose a measure" and "Top N" examples.
8. Dashboards — containers, actions, device layouts, a story.
9. Portfolio — publish 2–3 dashboards on Tableau Public (sales, HR, or a public dataset) and add the link to your resume.
10. Then — learn basic SQL too. Most analyst roles ask for SQL plus one BI tool.