Vizible Difference

Note first that here I loosely define data densification as what includes both interpolation and extrapolation of data marks as well as their associated values.

Below is a charting example in Tableau with a particular interest in using multiple data densification techniques.The use case

I was given a minimum data set such as a series of task-date pairs.

In its Tableau visualization, we used 4 densification techniques:

1.Densification of dates

The dates of tasks in the data table are discrete. Many dates in between are missing. By turning on Show Missing Dates, we are able to view all the dates. This way we densified the dates to be contiguous. We added a lot of extra data marks to the original 4 marks.

[ This is a guest post by Hans Romeijn. This is a followup post to my recent post on calculating YTD/YoY, QTD/QoQ, MTD/MoM and WTD/WoW. Hans shows a practitioner's approach to the calculation with performance in mind. The key is to replace conditionals by multiplications with pre-calculated indicator 1 or Null. This can accelerate the calculations significantly, especially when the data set is big.

The companion workbook courtesy by Hans is available for download.

I was asked a question: How to find out the IDs that showed up consecutively 5 times during the last 14 days?

How would you solve it?

Here I came up with 2 solutions. The 2nd one is a little simpler. Note that in either solution, we need to Show Missing Dates to make the calculations be based on contiguous dates. BTW, the last 14 days is irrelevant to the solution.

1. Solution with Running Sum

The idea is, CountD per Day for each ID equals 1 if it shows up, 0 if it doesn't.

In corporate finance, bridge chart is often used to visualize itemized sales/revenue performance during a particular period, such as a quarter or a year.

Bridge chart can be designed using waterfall chart. But we will use a different approach. Here we add itemized AOPs to the mix so that we can visualize the performance against goals for each item.

AOP means annual operation plan which is equivalent to the target in annual financial planning, usually on a quarterly basis.

A colleague posted this: "Hi Team, may I ask if you have any good idea to show the % difference of two randomly selected data points on a line chart?"

I found a solution to it, which is as follows.

- Click to set one point as a reference

- Hover over another point 

- Show the difference between the two points in tooltips

Both click and hover are actions that allow us to save a point's attributes into parameters. Then we can calculate the difference easily.

The show/hide buttons in containers and also in sheets allow us to create drill down functionalities in Tableau dashboards. Actually they make it simple to drill down in more ways than before.

Drill down with fixed sized containers

Here is an example. Given a simple bar chart by category.

Then with the click of a button, we can look into more details such Quarterly Profit and Regional Quarterly Sales.

Subtitle: Sunburst Chart with Labels Inside and Categorical Sequential Colors

Here I am presenting how to design Sunburst Chart with practical considerations, such as:

Labels insideCategorical sequential colors with dynamic data.The design will be based on map layers, a new feature since Tableau desktop version 2020.4. This post by Sebastian Depalla is a great source of inspiration.

First, here are two very important techniques when dealing with map layers.

[ Followup guest post by Hans Romeijn: Calculating Period-To-Date/PoP with Indicators for Better Performance ]

Year to Date (YTD) and Year over Year (YoY) calculations are very important in business dashboards. Jim Dehner recently wrote a great post on the topic. It inspired me to present additional approaches for the issues.

Calculating YTD and YoY without LOD

We have been using YTD and YoY long before LOD (level of details) appeared. So we are not dependent on LOD at all.

This topic came as a question from a colleague. I found the same question on Tableau community site which was unanswered. There is a similar post on the topic, a method first proposed by Zen Master Mike Cisneros. But it is about a single measure while the question is about two measures. So I am writing here a short answer to it. Note that this approach works for both single-measure and two-measure cases. That is, split the single one into two measures if necessary.