Introduction

Understanding our digital audience is a growing trend in museums and tools are being made to collect more information. For this project, I collected data pertaining to the Rubin Museum in New York, specifically looking at posts made by the museum on their Twitter account and how these posts were received and engaged with by their audience.

My goals in collecting this information was the gain an understanding of the size of the audience, the frequency of posts, and how engaged the audience was through this medium. I also wanted to see if there was a way to show what kinds of tweets receive more feedback than others.

Discussion

As inspiration for my charts, I looked at several visualizations. A lot of my inspiration came from seeing what was available through Tableau and trying to find the best way to express the information I wanted visible. While exploring Google Analytics and Twitter Analytics for ideas, I found their bar charts and line graphs most helpful in their visualizations, which made this my natural first choice in graphs.

I was also interested in the visualizations shown in class, including the heat map, which I reasoned would be a useful way to monitor when posts are made, and the matrix which was a useful tool in comparting different categories in a simple format.

Methodology

Collecting the data and uploading to Tableau

To collect the analytics for the visualization, I used the Google Sheets add-on, Supermetrics, to search the Rubin’s Twitter feed for data. I then downloaded the sheet as an excel to my computer and then linked that excel with Tableau to create my visualizations.

Data preparation

Before loading the excel file to Tableau, I deleted several columns of information that was either redundant – such as “date” when there was already and “date and time” column – and added two additional columns of data. In order to graph more of the content of the tweets, I used a formula to mark true or false if each tweet used a hashtag and also if they mentioned another Twitter account in their tweet.

Creating the graphs

Bar Chart

To start my visualization, I wanted to get a basic overview of the number of tweets composed by the Rubin Museum and how many of those tweets were liked or retweeted. To do this, I entered the year as a row and broke that down into months to show the frequency of tweets over time. I then added the sum number of posts, sum like count, and sum retweet count to the columns section. This displayed three separate bar charts next to each other. To combine the three charts into one so that the different metrics could be viewed side by side by month, I combined the three sum values under the heading, Measures Values. I was then able to add the Measure Values to the column section so that the data were divided by year, month and measure value, resulting in all three categories being combined to one chart instead of three for an easier comparison.

Matrix

After getting a better idea of the size of my audience, I wanted to find a visual to show the types of tweets and how much engagement they receive. To do this, I created a matrix as a simple way to show a lot of information at once. The graph divides the tweets into four categories, namely, the presence or absence of a hashtag in the tweet to the presence or absence of a “@” or mentioning another Twitter user within the tweet. The graph shows how many of each combination of tweets were made. Then there is the added element of counting how often these tweets were liked or retweeted which turns the four categories of tweets into four graphs.

To make this matrix, I first put the dimension Uses @ in the column section and Uses # in the rows section. This gave me the four categories and how many are in each section. I then added the measurements, the sum count of likes, to the column section and the sum count of retweets to the rows section. This turned the four categories into four graphs to show which category was receiving more likes and retweets than the others.

Heat Map

My final graph was a heat map showing the frequency of posts made by days of the week to see which days tended to get more posts over the course of the year. To create this map, I added the dimension, weekday, to the columns section and the dimensions year and month to the rows section. Finally, I added the measurement, sum number of records, to the body of the graph. To add a little more readability to the chart, I changed to color to be monochromatic and stepped the color so that there were 4 categories instead of a constant gradual.

Arranging the Dashboard

For consistency, I went back and edited the graphs so that all data showing the sum number of records or tweets was blue, as this was the only overlapping element used in all three graphs. I also noted that my bar chart was not fitting in the dashboard without a scrollbar. To make scrolling easier, I changed the rows and columns so that scrolling would be vertical instead of horizontal.

Discussion

My first graph (see Bar graph) shows the audience size and level of engagement. The second graph (see Matrix) shows the types of tweets posted, and the third graph (see Heat map), show the frequency of posts made by the museum by weekday.

Future Direction

I made the mistake of exporting the information from Google Sheets to an excel file and connecting the data for Tableau to the excel file. This option is much less portable and requires me to have the same information that I collected on one computer to still me available on another computer. Connecting Tableau to my Google Sheet would have allowed for more portability to my project as well as the ability to update the information whenever the Google Sheet was refreshed or edited. I also learned that, after downloading the visualizations from Tableau Public to my computer, I am not allowed to edit the data again until Tableau is connected to the original datasets. The graphs are still able to be edited which was a pleasant relief.

Connecting the data to Tableau from the Google Sheet would have also allowed the option to make that page shareable and post a URL to the dashboard so that users can see the raw data and read the tweets that were used and analyze further what my visualizations were not able to fully capture.

The data I chose for this project turned out to be rather limiting in how deeply I could look at engagement. I would have liked to have more metrics such as how the number of followers changed over time and the location of followers to see where their digital audience is coming from. It might have also been nice to compare their Twitter account with the audience and posts made on their Facebook to see the differences in engagement and use of the platform.

Link to Tableau Public

https://public.tableau.com/profile/dani.frank#!/vizhome/LIS658_Rubin_Twitter_dashboard/TwitterAnalytics?publish=yes