New Data Visualization – Event Time Series
Recently we invented a new type of data visualization. Hey its not often you can say that. And no it is not some crazy complex visualization that takes you weeks to work out just what is going on, but something, I suppose, which is a little more prosaic.
Like a lot of our design work, this one was inspired by a customer realestate.com. They wanted to be able to visualize the volume of traffic for a particular property advertised whilst overlaying the advertising campaigns associated with that property.
Sounds simple – but actually its not. The reason is that you are comparing two distinct time series data, which may or may not be correlated. In addition it is possible to have multiple campaigns running simultaneously so the “campaign” part of the chart could have many components spanning different time dimensions (events span time) whilst the volume is a spot figure per day every day.
We started to design how this may look at the difficulty was overlaying all the data points in a way that was understandable to the end user.
The problem was that we initially had a mind set that a single chart was required. The breakthrough was realizing that separate charts’ sharing a common time axis was the solution. This would let the user draw their own conclusion from the data in the same way that layers on a GIS map allow the user to create joins.
So what we came up with was a standard time series line chart to show volume and sitting below that a gannt style chart to visualize the events that took place over time. The example below highlights the changes in share price of various stocks are correlated with external events such as elections, natural disasters and recessions.
A nice outcome, and a really easy way to visualize and correlate data such as: marketing campaigns to web traffic; or historical events to share price movements.
Like a lot of our design work, this one was inspired by a customer realestate.com. They wanted to be able to visualize the volume of traffic for a particular property advertised whilst overlaying the advertising campaigns associated with that property.
Sounds simple – but actually its not. The reason is that you are comparing two distinct time series data, which may or may not be correlated. In addition it is possible to have multiple campaigns running simultaneously so the “campaign” part of the chart could have many components spanning different time dimensions (events span time) whilst the volume is a spot figure per day every day.
We started to design how this may look at the difficulty was overlaying all the data points in a way that was understandable to the end user.
The problem was that we initially had a mind set that a single chart was required. The breakthrough was realizing that separate charts’ sharing a common time axis was the solution. This would let the user draw their own conclusion from the data in the same way that layers on a GIS map allow the user to create joins.
So what we came up with was a standard time series line chart to show volume and sitting below that a gannt style chart to visualize the events that took place over time. The example below highlights the changes in share price of various stocks are correlated with external events such as elections, natural disasters and recessions.
A nice outcome, and a really easy way to visualize and correlate data such as: marketing campaigns to web traffic; or historical events to share price movements.