Some Background

As an employee of a major tech company, there are very specific guidelines in place around social media posting. Yet, in a world driven by social media, I wonder how salient is the sentiment of the average citizen on social media or influencers is on a company’s stock value? We can look at Elon Musk’s tweet on Bitcoin to see how their value quickly declined when he commented on its negative impact on the environment. This analysis will take a look at the 3 million unique tweets posted over the past five years and the daily market share of the top tech companies over the past ten years to assess the potential impact of a social presence on their daily market share.

To break this down further, I wanted to create a dashboard based on several related datasets, identified in kaggle, in order to assess the impact of social media posts based on relationships between social media influencers, followers, and companies (or rather topics) in respect to tweets. By assessing the relevant tweets in a given day against the daily market share value of the company respectively, I could determine if a correlation exists between the market value of a company in respect to their social media presence.

Tools & Process

Process

To analyze the impact, I created a dashboard in Tableau to showcase the daily market value performance of the top tech companies over the past 10 years. I included a reference line to indicate when Twitter began allowing company ads. This was intended to provide additional context for a user to compare the introduction of twitter against the impact that unfolded within the next few years. From there, I used a second visual to assess the tweet volume’s salience on the given company’s daily market closing value’s growth or decline respectively.

Before designing the visuals, I realized that the date formatting did not match. Unfortunately, the dataset was too large to modify in OpenRefine. However, the combination of this resource for and this forum dates helped me convert the format “seconds since epoch” into a consumable date for analysis.

To design the data visuals, I aimed for a minimal look: less titles and wording to ensure that the visual would not come across cluttered by only having the essential text needed for the user to understand what they were looking at. I used colors that aligned with the company logos that were associated with the stock values. I decided on the line graph for the stock values because it does the best job of analyzing data over a period of time. Unfortunately, I couldn’t use the combination bar and line graph overlay to put the twitter data on the same graph because it would aggregate the count of tweets and the daily stock market values for each year; creating a view that could be easily misinterpreted. I used the small rings for the twitter count view to show the volume of the trending topics more visually. The retweets, as I found from looking through the tweet table set, consisted of spam or trending news that was retweeted from sources like the Wall Street Journal.

From there, I incorporated a legend aligned with the line order in the first visual for easy association. I put the legend in between the two visuals for quick reference from either point of the dashboard. I then aligned the dates such that the starting point of the starting date of the tweet count began at the same point of reference as the stock value visual for an easier comparison.

I then enabled filtering for the visuals so when one value is clicked on a visual, the associated value on the other visual is filtered to help the user focus on a specific point in time and company. I initially had a third visual which filtered on the specific tweet associated to the given point in time for the respective company. When I conducted user research through observation and interview, I found that this was not going to be functional. As a result, I removed the table of twitter tweets.

User Testing

I conducted two moderated user interviews and overall usability to gauge the user experience on the following questions.

1. What is the visual trying to communicate?
2. What can you conclude?
3. Is the dashboard overall good or bad?
4. Is any information that you’d like to see missing

These questions helped validate the usability of the information available and confirmed that it was communicating the right thing to the user. The findings are indicated below within the context of three key groupings: response time, drawing conclusions, and where to learn more.

Finding 1: Response time

Due to the size of the data set (3+ million tweets), it slowed the response time down to the point that the users were both unable to interact with the data. They were immediately frustrated and thought there was something wrong with the page. Once the table was removed, they were very happy and interested in exploring the data to learn more.

Recommendation: When using large sets of data, it is best to avoid incorporating a data table in a visual unless you have the proper load balancing for your server to handle the workload. If you’re visual takes too long to load, you will almost immediately lose your audience. They may also think that there’s something wrong with the webpage and leave.

Finding 2: Drawing Conclusions

Users successfully gathered that the number of tweets positively correlate with the stock value, regardless of sentiment; concluding that “if my company is trending, I can raise my stock value”.

Finding 3: Where to learn more

Some of the information users were interested in learning more about was the reason why certain companies were not performing as well, like Microsoft. They were also curious why certain tech companies were not included in the data set altogether, like IBM.

Recommendation: One could include further links to more information if possible to provide further context on the stock trends.

Conclusions

Overall, I’m happy with how this visual worked out. It incorporates the right level of minimalism while creating a compelling narrative to showcase the impact of social media on a company’s stock value.

References

1. Allebach, N. (2019, June 24). Brand Twitter Grows Up. Vulture. https://www.vulture.com/2019/06/brand-twitter-jokes-history.html
2. Dates Display in Unix Time Instead of Date or Datetime | Tableau Software. (n.d.). Retrieved July 27, 2021, from https://kb.tableau.com/articles/issue/dates-display-unix-epoch-instead-of-date-datetime
3. Replacement for unix timestamp in calculations. (n.d.). Retrieved July 27, 2021, from https://community.tableau.com/s/question/0D54T00000C5qr7/replacement-for-unix-timestamp-in-calculations
4. Twitter | History, Description, & Uses. (n.d.). Encyclopedia Britannica. Retrieved July 27, 2021, from https://www.britannica.com/topic/Twitter