How to make sense of them
When doing date visualisations in Tableau, it can occur quite frequently that when the date is added on to a graph, there a continuous date line appears when the needed outcome was to compare the different quarters of the data as bars. The reason for this is that dates can either be described as discrete or continuous, this blog will explain the implication of the different date descriptions and when you can use each one.
Discrete vs Continuous
Tableau describes data fields as either discrete (represented as the BLUE pill) or continuous (represented as the GREEN pill) as described in this (hyperlink) blog post. Dates are a particular data type within this where it is important to understand what each format implies as the visualisation outcomes can be drastically different. It is not simply the way in which the data looks in the view but also how it behaves.
To summarise,
- BLUE FIELDS = DISCRETE as they contain a FINITE number of values; they break down and add more dimensions to the view. An example of this would be the different types of departments a company might have. This can be broken down into categories and subcategories but no further.
- GREEN FIELDS = CONTINUOUS as they could contain an infinite number of values. Tableau will naturally aggregate them when placed in the main view. An example of this would be the number of sales made by a company. Although it is highly unlikely that it would be infinite, this value could be any number along a number axis.
How does this relate to dates?
Within Tableau, dates represent a specific data type that can exist as either Date Values or Date and Time Values. For the purpose of this blog, we will focus exclusively on how the date itself is expressed rather than discussing the time element. When a date field is added to a view, it can be defined as either discrete or continuous. It is important to understand that Tableau will handle the identical underlying data in a completely different way depending on which description you choose.
Discrete Dates
When dates are discrete, they are nested within hierarchies of YEAR QUARTER MONTH DAY and each becomes its own individual label/block as opposed to a point in a scale. This is useful for any type of analysis where one needs to view how the passage of time has affected one or several variables. These date parts can be used to compare time periods against each other and they are the date parts of a calculation. An example of this would be the way in which the sales in Q3 have been in the past decade.
Continuous Dates
Continuous dates, these can be thought of as an axis of time which is connected from a starting point onward. This allows values to be shown over time, the ups and downs of sales over all the quarters in the past decade.
If they’re both dates in the same visual format, why do they behave so differently? The main difference in this situation becomes apparent when there is an input of data which has gaps in the data, so if for example there were no sales in a specific month then with:
Discrete Dates - the gap does not appear, tableau simply skips the value of the fields which are ‘missing’ and puts the available values next to each other.
Continuous Dates - the axis keeps an even spacing even though there is no sales data for that month, in a line graph this will be represented as a dip or break in the line as the axis’ priority is to represent the time field not only the values plotted.
Troubleshooting: what goes wrong
A common mistake is using a discrete month field when you're trying to spot a trend. Because discrete only creates headers for periods that actually have data, a gap in your sales (ie. a month with no activity) just won't show up. The chart ends up looking like a smooth run of sales, when really there was a dip.
The opposite mistake happens too, using a continuous quarter field when what you actually wanted was to compare quarters against each other. Instead of four bars grouped neatly by year, you'll get one long timeline, which makes it much harder to line up Q3 this year against Q3 last year.
Knowing this, if a chart looks off in tableau, looking at the format of the date pill can be a starting point to solve the issue.