Tableau for Advanced Users: Mastering Data Visualization and Analytics

Introduction

Tableau has become the gold standard tool for data visualization, used by over 86,000 organizations worldwide including 96% of the Fortune 500. It‘s loved by data professionals for its versatility and ease-of-use in quickly creating interactive dashboards to explore data and share insights.

But while anyone can learn the basics of Tableau in a few hours, true mastery requires leveling up your skills with advanced techniques. According to expert Tableau trainer Andy Kriebel, only about 10% of Tableau users take full advantage of the tool‘s capabilities.

In this comprehensive guide, we‘ll dive into a range of advanced tactics – from complex calculations to R/Python integration, innovative chart types to dashboard design principles – to take your Tableau abilities to the next level. Whether you‘re an analyst, data scientist or business leader, these techniques will help you get more value from your data and tell impactful data stories.

Advanced Chart Types

Let‘s start with some chart types beyond the standard bar, line and pie charts. While you can make many of these through "Show Me", they require some additional customization:

Dumbbell Charts

Dumbbell charts are an alternative to paired bars for comparing two data points, like a starting and ending value or a forecast vs. actual. They make it easy to spot the difference between the points. Here‘s how to make one:

Create a scatterplot with the two measures on Rows (e.g. Sales 2019 and Sales 2020)
Change both to Circle marks and adjust size
Create a Measure Names/Values calculated field: <Measure Name> + SUM(<Measure Value>)
Put the new field on Detail and Path to connect the circles
Add labels, average lines, and formatting

Dotplots

While bar charts are the go-to for comparing categories, they take up a lot of space. Dotplots are a more compact alternative that encode the values based on position on a scale. To create one in Tableau:

Put the categorical dimension on Rows
Create a calculated field for the X position based on the measure, e.g. WINDOW_AVG(AVG([Profit]))
Add the new field to Columns and change to Circle mark
Make a dual axis, synchronize and hide the original axis
Size the circles, add a text label, and format

Spark Bars

For comparing metrics across multiple categories and time periods, spark bars (or spark lines) are a space-efficient option popularized by Edward Tufte. To make spark bars in Tableau:

Put Category on Rows and a discrete date field on Columns
Add the measure to Rows and change to Bar mark
Drag the measure to Size and reduce to fit
Add reference lines, labels, and highlighting as needed

According to Tableau Visionary Ryan Sleeper, spark bars "allow you to evaluate trends across several categories and compare their performance – something that‘s very difficult with traditional bar charts."

Hex Tile Maps

When mapping data, hex tile maps provide an alternative to filled maps that avoids visual distortion from geographic size differences. Here‘s the process in Tableau:

Install the Hex Tile Maps extension
Create a new sheet with the geographic field on Detail
Assign the measure to Color, and adjust the palette
Go to the Map Layers pane, add the Hex Tile Map
Customize the hex size, color, opacity, borders, etc.

Hex maps work especially well for mapping election results, as shown by the New York Times. They provide an equal-sized geographic unit and a cleaner look than a traditional filled map.

Calculation Deep Dives

Calculated fields are where Tableau really shines for answering complex questions about your data. Let‘s explore a few advanced examples:

Hexbin Plot

Scatterplots are great for showing relationships between two measures, but they can get crowded with many points. Hexbin plots (2D histograms) are a way to visualize the density of points. Here‘s how to make one with LODs:

Create a scatterplot with the two measures
Create a calculated field for the X/Y coordinates:
[X Coordinate]: INT([Measure 1])
[Y Coordinate]: INT([Measure 2])
Create a hex count calc: {FIXED [X Coord],[Y Coord] : COUNT([Measure 1]) }
Drop X/Y on Columns/Rows, change to Polygon mark
Put Hex Count on Color, adjust padding

This example from Tableau Zen Master Ann Jackson shows the relationship between efficiency and production numbers for NBA players. The hexagons show where the densest clusters of players are.

Logistic Regression

Sometimes you want to predict a binary outcome (e.g. purchase or not) based on a set of input variables. Logistic regression is a statistical method for this kind of classification. Here‘s how it works in Tableau:

Connect to the data source with the target variable and predictors
Create calculated fields for each predictor multiplied by a parameter coefficient (e.g. [Age]*[Age Coef])
Create a calculated field summing the predictor terms
Create a final calculated field to convert the logistic function:
1/(1+EXP(-[Logit]))
Show the coefficient parameters and probabilities in the view

In this example from Playfair Data, logistic regression is used to predict employee attrition based on factors like age, job level, and years at the company. The coefficient parameters can be adjusted to fine-tune the model.

Monte Carlo Simulation

Monte Carlo methods are used to model uncertainty by generating many random trials to determine the range of possible outcomes. Tableau can handle Monte Carlo with a few table calculations. Here‘s the basic process:

Create a calculated field for a random value between 0 and 1
Create a calculated field to generate the random trial values based on an assumption (e.g. NORM_INV([Rand],100,20))
Create a table calc for the trial number RANK_UNIQUE(RANDOM())
Show the MIN, MAX and desired percentiles

This example from The Data School shows a Monte Carlo simulation of sales data. By generating 1000 random trials from a normal distribution, it estimates the best and worst case sales scenarios.

According to a survey by the Business Application Research Center, advanced analytics like predictive modeling are one of the top trends shaping BI, used by 45% of companies. Tableau‘s calculation language provides a flexible foundation to implement many of these methods.

R & Python Integration

For even more advanced capabilities, Tableau can integrate with programming languages like R and Python. This taps into their ecosystems of packages for statistical analysis, machine learning, and more. Some common use cases include:

Forecasting with time series methods (e.g. ARIMA, exponential smoothing)
Clustering customers or transactions (e.g. k-means, hierarchical)
Predictive modeling with machine learning (e.g. logistic regression, random forests)
Text mining and sentiment analysis
Network graphs and spatial analysis

Here‘s an example of integrating a random forest model in Python:

SCRIPT_REAL("
import pandas as pd
from sklearn.ensemble import RandomForestRegressor

df = pd.DataFrame(_arg1)  
X = df[[‘store‘,‘year‘,‘month‘]]
y = df[‘sales‘]

rf = RandomForestRegressor(n_estimators=100)
rf.fit(X, y)

return rf.predict(X)
", 
ATTR([Store ID]),ATTR([Year]),ATTR([Month]))

And the same model in R:

SCRIPT_REAL("
library(randomForest)

df <- data.frame(.arg1,.arg2,.arg3,.arg4)
colnames(df) <- c(‘Sales‘,‘Store‘,‘Month‘,‘Year‘)

rf <- randomForest(Sales ~ Store + Month + Year, df) 
predict(rf, df)  
",
AVG([Sales]), ATTR([Store ID]), ATTR([Month]), ATTR([Year])
)

Both scripts take in the relevant variables, train a random forest model, and output the predictions to visualize in Tableau. The R and Python connections open up endless opportunities to extend Tableau through code.

Data science continues to be one of the hottest fields, with demand for data scientists growing 650% since 2012. Integrating data science techniques with the visualization power of Tableau is a killer combination.

Dashboard Design Principles

Building effective Tableau dashboards is both an art and a science. Let‘s walk through an example focused on diversity & inclusion metrics and callout some key design principles:

1. Thoughtful Layout

Group related views together (e.g. candidate funnel, hire demographics, performance)
Use a clear hierarchy, with the most important views on top
Create a logical flow, with filters/parameters on the left and details on demand

2. Informative Headlines

Include a succinct, action-oriented title (e.g. "Diversity Dashboard – 2020 in Review")
Add short subtitles to introduce each section
Include key takeaways and reference lines in the titles where possible (e.g. "Diversity Hires Fall 10% Below Target")

3. Contextual Color

Use color minimally to highlight key insights
Choose a cohesive, brand-relevant palette
Make sure color use is consistent across metrics

4. White Space

Don‘t feel the need to fill every pixel
Let the most important elements stand out
Leave some breathing room, especially in dense dashboards

5. Interactivity

Let users adjust date ranges and other key parameters
Cross-highlight views based on selections
Provide tooltips with additional context and commentary

Award-winning dashboard designers like Anya A‘Hearn recommend storyboarding your dashboard before building. Think carefully about the key questions to answer, how users need to interact, and what context to provide. A thoughtful, intentional design process leads to more effective dashboards.

Learning Resources

Continuing to level up your Tableau skills takes constant practice and learning. Here are some of the best resources to expand your knowledge:

Conclusion

Mastering advanced Tableau techniques is an ongoing journey that combines technical skills, design thinking, and domain knowledge. As the field of analytics continues to evolve, it‘s important to stay on top of new features and approaches.

The real key is practice – building lots of dashboards with real data, gathering feedback, and refining your work. Look for opportunities to apply advanced techniques like complex calculations, statistical models, and custom chart types. Participate in the Tableau community to be inspired by the work of others and share your own creations.

Equipped with these advanced tools and a commitment to honing your craft, you‘ll be poised to tackle the toughest data challenges and bring valuable insights to your organization. Here‘s to your journey to Tableau excellence!

Tableau for Advanced Users: Mastering Data Visualization and Analytics

Introduction