Tableau: Dimensions, Measures, Joins and Blending

1The Tableau ecosystem

Product	What it is
Desktop (paid)	The main authoring tool for visualisations, dashboards and stories
Public (free)	Same authoring, but workbooks must be saved to the public Tableau Cloud. This is the one to learn on
Creator	A licence role bundling Desktop and Prep
Server (on-premise)	Publish, share and manage workbooks inside an organisation
Cloud, software as a service (SaaS)	The hosted version of Server. View, interact, edit and collaborate from a browser or mobile
Prep	A separate app for cleaning, shaping and combining data before Desktop

Worth remembering

Public saves your work publicly. Fine for coursework and a portfolio, not for anything confidential.

The workspace

Fig 11.1The six areas, and what each one does

The Marks card is where Weeks 3 and 4 live. Colour, size and shape are the pre-attentive attributes, and the card is where you assign them to fields. Click any area for what it is for.

Dimensions and measures

	Dimension	Measure
Holds	Categories and descriptions	Quantities
Examples	Name, Date, Region	Sales, Profit, Quantity
Usually	Discrete	Continuous
Pill colour	blue	green
Role in the view	Splits the data into groups	Gets aggregated within each group

Tableau classifies fields automatically on connection, and it is sometimes wrong, so check before building.
The blue and green convention is worth internalising, since you read a view's structure from pill colours at a glance.
This maps directly onto Chapter 4's measurement scales: nominal and ordinal fields become dimensions, ratio fields become measures.

Build a view by dropping pills on shelves

Fig 11.2Click a field to put it on a shelf, and watch the view change

Data pane · click to assign

Columns

Rows

Marks · Colour

A dimension on a shelf splits the data into one mark per category. A measure on a shelf aggregates within each of those groups. Put two dimensions on and you get a grid; put a measure on Colour and the marks are shaded by value. That is the whole grammar, and Show Me is only guessing at it for you.

Connecting to data: live against extract

Fig 11.3Two connection modes, and which one each dashboard wants

Best practice: use extracts for large datasets or slow connections. This is Chapter 5's data latency problem as a concrete choice. An operational dashboard needs live; a strategic one is fine on a refreshed extract and will be far faster.

Metadata management

Metadata is data about the data, and it controls how Tableau interprets what it received.

Task	What it does	Example
Rename fields	Business-friendly names	`Cust_ID` becomes `Customer ID`
Change data type	Correct the interpretation	String, Number, Date, Boolean, Geographic Role
Create aliases	Rename values inside a dimension	`CA` becomes `California`
Create hierarchies	Group related fields for drill-down	Region → State → City

Fig 11.4Geographic Role, and the hierarchy that gives you drill-down

Geographic Role is the one that unlocks maps. Tag a field as a country, state or city and Tableau can plot it without you supplying coordinates, which is worth comparing against the work GeoPandas needed in Chapter 8. Hierarchies build drill-down: once Region → State → City exists, the view gains a plus icon, and that is Chapter 5's drill-down and Chapter 6's data story type set up in one step.

Combining data: the four join types

Fig 11.5Which rows survive each join

the result

Joins merge at row level into a single table, they are recommended when the tables come from the same source, and they are defined on the Data Source tab before you build a sheet. Watch what happens to the null cells: a left join keeps a customer who never ordered, and their order value arrives as null rather than zero.

Data blending

Fig 11.6Row-level merge against aggregate-then-combine

Blending needs a primary source, a secondary source, and a linking field shown with a chain icon. The limitation: every field from the secondary source is treated as aggregated, using ATTR or another aggregation, so you cannot get row-level detail out of it. Same source, or you need row-level detail, use a join. Different systems that cannot be joined, blend, and accept the aggregation limit.

Worksheet, dashboard, story

Object	What it is	Course concept it implements
Worksheet	A single visualisation, a viz	Everything from Modules 1 and 2
Dashboard	Several worksheets, filters and objects presented together	Chapter 5
Story	An ordered sequence conveying a guided narrative	Chapter 6

The three objects map exactly onto the three theory modules. A story is author-driven narrative made concrete: you fix the order of what the reader sees, while each dashboard inside it stays interactive.

Calculated fields and parameters

Fig 11.7A parameter puts the constant in the reader's hands

Calculated fields extend the data without changing the source, which is how you build metrics the source never stored. Parameters are what give the reader control: swapping a constant for a parameter lets the audience change a threshold, a target line or a scenario themselves. That makes parameters Tableau's version of Chapter 10's widgets, and the mechanism behind reader-driven narrative here.

Dashboard interactivity

Fig 11.8Cross-filtering, switched on per object

A selection in one chart can filter every other chart on the dashboard, switched on per object using the filter button on the worksheet's container. Filters combine, so choosing a year in one view and a category in another narrows to the intersection. Cross-filtering is the clearest example of Chapter 5's warning about data compatibility across visuals: it only works when the fields being filtered mean the same thing in every worksheet.

Key points

The Tableau products, and what separates Desktop, Public, Server, Cloud and Prep.
The six workspace areas and what each does.
Dimensions against measures, including the blue and green pill convention.
How dimensions and measures map onto Chapter 4's measurement scales.
Live connection against extract, with two advantages and two drawbacks of each.
When to prefer an extract.
The four metadata tasks, and why Geographic Role and hierarchies matter most.
Joins against blending, and the situation each is for.
The four join types and what each keeps.
The three blending requirements: primary source, secondary source, linking field.
The blending limitation on secondary-source fields.
Worksheet, dashboard and story, and the course module each corresponds to.
Calculated fields against parameters, and why parameters enable reader-driven narrative.
How cross-filtering works on a dashboard, and what it requires of the underlying data.