Improving Performance

Once people are familiar with Tableau’s functionality, they soon run into another type of issue: performance. You might all have experienced staring at the screen while reading Loading Data or Executing Query. But don’t worry, we’ve got you! If designed accordingly, Tableau dashboards can perform very well, even with large amounts of data.

This chapter will address various aspects of performance with the intent of empowering you with techniques to create workbooks that load quickly and respond snappily to end user interaction.

In this chapter, we will discuss the following topics:

• Understanding the performance-recording dashboard
• Hardware and on-the-fly techniques
• Connecting to data sources
• Working with extracts
• Using filters wisely
• Efficient calculations
• Other ways to improve performance

As you can see, there are many topics we have to cover with regard to performance improvement...

Understanding the performance-recording dashboard

Tableau includes a performance-recording feature as part of the installation package and it ships as a dashboard named PerformanceRecording.twb. The dashboard gives the Tableau author an easy way to understand and troubleshoot performance problems. The following exercises and associated discussion points will review various aspects of the performance-recording dashboard, including how to generate it, how to use it to improve performance, and how it’s constructed.

Perform the following steps:

1. Navigate to https://public.tableau.com/profile/marleen.meier to locate and download the workbook associated with this chapter.
2. Navigate to the Types of Events worksheet.
3. Select Help | Settings and Performance | Start Performance Recording.
4. Press F5 on Windows or Command + R on macOS to refresh the view.
5. Select Help | Settings and Performance | Stop Performance Recording. A new dashboard will open:
...

Hardware and on-the-fly techniques

Even though Tableau Desktop does not need a lot in terms of hardware, it still happens that company desktops or laptops are underpowered, especially if you are a so-called business user who does not require programming tools or lots of processing power in your day-to-day work.

Therefore, in this section, we will discuss the technical requirements. As per www.tableau.com, the latest technical requirements are:

Windows	Microsoft Windows 8/8.1, Windows 10 (x64), Windows 11
	2 GB memory
	1.5 GB minimum free disk space
	CPUs must support SSE4.2 and POPCNT instruction sets
Mac	macOS Catalina 10.15, macOS...

Connecting to data sources

One of the beauties of Tableau is the ease with which you can connect to many different data sources in various ways. As mentioned earlier in this book, there are many connectors defined in Tableau for interfacing with a variety of data sources. Furthermore, this flexibility extends beyond simply connecting to single tables or files.

Although Tableau makes it easy to connect to various data sources, it should be stressed that Tableau is not an Extract, Transform, and Load (ETL) tool. If complex joins and complex data blending are required to generate useful results, it may be advisable to perform ETL work outside of Tableau, for example, in Tableau Prep Builder (see Chapter 3, Using Tableau Prep Builder, for more information on this service). ETL work will ideally lead to better data modeling and thus easier authoring and quicker performance in Tableau.

The four ways in which Tableau connects to data are as follows:

• Tableau...

Working with extracts

This section will discuss what a Tableau data extract is as well as how to efficiently construct an extract. A colleague of mine recently consulted with a relatively small mobile phone service provider. Even though the company was small, the volume could be in excess of 1,000,000 calls per day. Management at the company insisted on the ability to interface with detailed visualizations of individual calls in Tableau workbooks. The performance of the workbooks was, understandably, a problem. Was such low-level detail necessary? Might less detail and snappier workbooks have led to better business decisions?

In order to balance business needs with practical performance requirements, businesses often need to ascertain what level of detail is genuinely helpful for reporting. Often, detailed granularity is not necessary. When such is the case, a summary table may provide sufficient business insight while enabling quick performance. In the case of the mobile phone...

Using filters wisely

Filters generally improve performance in Tableau. For example, when using a dimension filter to view only the West region, a query is passed to the underlying data source, resulting in information being returned for just that region. By reducing the amount of data returned, performance improves. This is because less data means reduced network bandwidth load, reduced database processing requirements, and reduced processing requirements for the local computer.

Filters can also negatively impact Tableau’s performance. For example, using only relevant values causes additional queries to be sent to the underlying data source, thereby slowing down the response time. Also, creating quick filters from high-cardinality dimensions can impair performance.

Tableau’s filters are executed in a specific order, so keep this in mind when using them. The following flowchart, along with a link to an hour-long presentation, may help you grasp the concept fully...

Efficient calculations

Calculations may be constructed differently and yet accomplish the same thing. Look, for instance, at the following example, which shows that an IF statement can be replaced by simpler code:

Scenario I	Scenario II
Create a calculated field with the following code: IF SUM (Profit) > 0 THEN 'Profitable' ELSE 'Unprofitable' END Place the calculated field on the Color shelf.	Create a calculated field with the following code: SUM (Profit) > 0 Place the calculated field on the Color shelf. Right-click on True and False in the resulting legend and rename Profitable and Unprofitable.

Since either of these scenarios will return the desired results, which...

Other ways to improve performance

To conclude this chapter, let’s consider a few other possibilities for improving performance.

Avoid overcrowding a dashboard

Often, end users want to see everything at once on a dashboard. Although this may be perceived as beneficial, it often is not. Consider the inclusion of a large crosstab on a dashboard. Does scrolling through pages of details add to the analytical value of the dashboard? Perhaps the answer is “no.” Furthermore, an excess of information on a dashboard may obscure important insights. Diplomatically arguing for leaner dashboards may lead to better decision making as well as better performance.

Fixing dashboard sizing

Dashboards can be set to an exact size or to Range or Automatic. Exact size results in quicker performance because once Tableau Server has rendered a view for one end user, that render stays in the cache and can be reused for the next end user that accesses that dashboard. Automatic...

Summary

We began this chapter with a discussion of the performance-recording dashboard. This was important because many of the subsequent exercises utilized the performance-recording dashboard to examine underlying queries. Next, we discussed hardware and on-the-fly techniques, where the intent was to communicate hardware considerations for good Tableau performance and, in the absence of optimal hardware, techniques for squeezing the best possible performance out of any computer.

Then we covered working with data sources, including joining, blending, and efficiently working with data sources. This was followed by a discussion on generating and using extracts as efficiently as possible. By focusing on data sources for these three sections, we learned best practices and what to avoid when working with either remote datasets or extracts. The next sections explored performance implications for various types of filters and calculations. Lastly, we looked at additional performance considerations...