Review of Data Visualisation with Tableau Specialisation from Coursera & UC Davis (Courses 1 & 2)
- Tat Yuen
- Mar 4, 2025
- 8 min read
The five course specialisation is a useful starting point for understanding and getting into business intelligence, analytics and data visualisation.
No programming is required but understanding of conditional logic is required to get the most out of the courses. Some experience building reports would also be useful but is not necessary.
This course is not just about the mechanics (features and functions) of Tableau. What I found immensely valuable was that they cover a lot of the principals and best practices for optimising data visualisation for effectiveness (which can apply to any data visualisation platform and not just Tableau).
Students will start with the free Tableau Public version and then a six month trial version of Tableau Desktop as you progress in a later course. Useful links are provided to additional readings and other online resources throughout the course. I found most of them very useful though a few were dead links. You have to do peer reviews where you review and confirm successful completion of other student's projects and your projects are reviewed by peers. This part I did not like but it does allow you to see who is and who is not taking the learning seriously.
Course 1 - Fundamentals of Visualisation with Tableau
Suggested pace: 10 hours over 4 weeks
Content: 25 videos (Total 105 minutes)
2 suggested readings
2 quizzes
1 practice exercise
Introduction to the course, it's instructors (all good instructors!) as well as how to install Tableau public. The what, why, and how of data visualisation that serves as a foundation for the rest of the course. Can be completed in less than a few hours.
Week1: Data Visualisation and Its Importance
Standard intro with who what and when. What is and is not data visualisation. Why is it important.
The interview with Dr. Ben Shneiderman, a early pioneer of data visualisation and human computer interaction, is worthwhile. He famously used treemaps and network analysis (which is hardly covered in this specialisation) for finding clustering patterns and linkages.
You will install Tableau Public.
After that, you may treat yourself to some eye and brain candy by checking out some awesome visualisations at this Viz Gallery.
Week 2: Getting to Know Tableau for Data Visualisation
You'll get a high level overview of the user interface but not much on how you can customise it to your liken. Go to Youtube for that.
Week two continues with getting you familiar with the component that you'll be working with like a worksheet which is how your data is stored. It'll look like a spreadsheet because it came from a spreadsheet. But it's not. It's more like a dataframe.
They introduce you to the Tableau dashboard (a dashboard that you'll create in a later course) and then you import a prepared dataset in the form of an orders table from a fictitious America office supply company. So it's American postal codes, states, etc.
Really simple prep that you finish in less than an hour
Week 3: Let's Make a Viz
In week 3 you'll use the data that was imported to make three visualisations and a simple dashboard.
Business analytics like data science starts with a good question. A good question usually offers insights that are actionable or guides the viewer to ask more specific questions for further exploration.
A simple yet powerful and versatile visualisation uses any number of bar charts. Using bar charts will be your first visualisation. Bars are great for grouping and comparisons and are pre-attentive when used with colour. You can even stack them to see percentage-like visualisations.
The second is the line graph. Line graphs are important for time-series data where you want to show trends. The slope of the line is in and of itself a pre-attentive attribute. Rising is increasing and downward sloping is decreasing.
The thrid one uses treemaps. You see treemaps often used it find out how disk space is being used on your hard drive.
You end this week by creating a very simple dashboard. A dashboard is a single page view that generally uses a group of related charts.
Summary:
I spent a lot time during week 3 viewing Youtube tutorials on charting. There are just too many features to cover in their lectures. It's worthwhile to try things out and make changes to see the resulting output. My tableau public site became a messy amalgam of such test visualisations. I went to www.kaggle.com to find other datasets to try as well.
Week 4: Tableau Community Projects and Visualisation Best Practices
The five best practices you'll learn about:
Know your audience (can be more than one & strategies for that situation)
Know your data (explore and ask)
Use color purposefully. Use gestalt principlses. It shouldn't have a hallucinatory effect.
Less is more.
Get feedback early and often
Know your audience
This is critical to get right. C-level executives like higher level summary data often visual. CFOs like high level summary data but in numbers. Marketing types like segmentation and percentages with the ability to drill down. Know what types of question your audience tends to ask.
Know your data.
The most time consuming part of business analytics and data science. The data sets you get online are mostly tidy already. In the real world, production data is often quite a mess and ingesting it for analysis is tedious and time consulting. (I like to use Python with Pandas to clean up data.)
You'll want to know how the data was obtained and its source. You'll need to find incorrect data for it's type (like NaN), missing data, and duplicates. Noting missing and incorrect data is an important part of the data inventory analysis. Ask about how granular is the data. Knowing your data provides answers to questions of accountability, transparency, explainability, and bias. Finally, knowing your data makes you more credible.
Use Color Purposefully
Golden rule; use no more than five colours. It's the same guiding principle for the number of bullet point per presentation slide.
Using white space can be very effective and is an application of a gestalt principle. Separation de-clutters what otherwise may be a crowded chart. Contrast is good when you want it and similar hues are useful for like-groupings.
Less is More
Have a clean aesthetic that won't trigger cognitive over load. Add labels only when necessary and leave out if the chart is self-explanatory. Use numbers only when they add useful or necessary information. Use abbreviations that make sense. Anything that is unnecessary can be a distraction.
Getting Feedback Early
Make sure you have a plan and be Agile. Work on a bit and then seek feedback. With feedback, you are making sure that you are working on the right question and working in the right direction. Your audience knows things you don't so they are your best resource. Keep them involved along the way to project completion.
Course 2 - Essential Design Principles for Tableau
Suggested pace: 12 hours over 4 week
Content: 34 videos (Total 3 hrs 42 minutes)
6 suggested readings - very worthwhile
3 quizzes
All about the Tableau and getting data into Tableau for building your viz (industry jargon for visualisation). Getting course 2 right is critical to smooth sailing as you continue through the course.
Week 1: Getting Started in Effective and Ineffective Visuals
Very good content the delves into how people perceive (seeing and interpreting) and how people know (thinking). This week is about designing effective visual presentations using various types of visualisation that caters to the way our brains work. Also covered is the area of ethics because we all know how data can be framed and presented in a way that is biased. Examples of bad visualisations are also provided but the reader can find many such examples with ease online.
Week 2: Visual Perception and Cognitive Load
We've all experienced it when sitting through Powerpoint presentations. Often called, "TMI" (too much information), where the information overloads our cognitive abilities and the message is lost in a morass of colour, charts, and textual descriptions. Worse is that it has no emotional impact so you will forget the message not long after viewing it unless you make special cognitive efforts to anchor it.
This weeks most important concept is that of pre-attentive attributes (obvious at-a-glance), sparse design (uncluttered) and appropriate use of color in making visualisations that have an acceptable cognitive load that does not overwhelm the senses.
Pre-attentive attributes convey information immediately. I call it recognition before cognition. For instance, though a stop sign only has one word, I doubt any of us actually read it. We recognise the red symbolic octogon and we know. We know that a rising graph means increasing and decreasing is downward sloping. Cognition comes when you read the charts labels. With that said, size matters not only for legibility but also for emphasis.
Week 3: Design Best Practices and Exploratory Analysis
More design tips leveraging on design principles in the previous lessons and adding the six Gestalt Principles of huan of similarity (common region), continuation, closure (rectification), proximity (emergence), figure/ground (multi-stability), and symmetry and order. It's about understanding what the user sees and seeing the sum of the parts as a whole. Using these design principles will enable the creator to ensure that visualisations allow the viewer to get the meaning of otherwise seemingly chaotic sets of data. A seventh principle, focal point, is not covered but often alluded to in speaking to what may be distracting elements in a visualisation.
Using data for exploratory analysis is explained. This is a critical part of any business analysis so you can see what your data is to identify anomalies and outliers. The dataset used in this course is already clean and tidy but this is rarely the case in the real world. Anscombe's Quartet is used as an example how data must be explained as the visuals can be deceiving similar even if the underlying data is not
Scatter plots are often used to visualise categorisation. This exploratory chart would provide insight as to whether supervised or unsupervised learning is appropriate for a dataset.
This is a critically important chapter and knowing how to apply the principles so far is the first step to separating the children from the adults.
Here's useful site about the psychology of UX design.
Week 4: Design for Understanding
Patterns are easy to see and does not require the audience to read. That's the goal of a good visualisation. The initial understanding is visceral and immediate. The cognitive kicks in only afterwards and involves reading. The last week goes into designing for your audience so an understanding of personas is useful.
Dives deeper into the relationship between date and the design for the type of data you want to present to the audience you're presenting to. Labeling can be a distraction or an information must-have. Know the difference. Tableau allows for interactive visualisation so the user can sort and filter dynamically but this has to done right and not confuse the user.
Labelling and scale of your visualisations are important for clarity and understanding but can be applied in ways intented and unintended to introduce bias that can be misleading and cause the audience to interpret the data in an incorrect way. This can be done using what are known as visual lies that cause cognitive bias.
Example: Dual axis charts and graphs must have scales that are equal or the data looks skewed. Highlighting what you want to show as important versus what is actually important using size and/or color. Biases of omission are also explained.
Summary
Quite a few students lamented on the lack of "how to" content with regards to Tableau features. I though this was a very good primer on what constitutes a good visual presentation of data. There are so many more up-to-date video tutorial on how to create a certain graph or chart, but the second gives one the goal when applying the Tableau feature sets.
This more academic approach continues throughout the courses in the rest of the specialisation and is an important and welcomed aspect of these courses. The metrics and data are the science part of data science, but the artful component is how to present insight that has emotional impact, is memorable, and actionable.
And because data can used to make important decisions, accountability, transparency and explainability are required so as to ensure that the underlying premises can be audited for any biases.
Comments