Skip To Content

Trending Toward Negativity: My Analysis of Google’s Year in Search 2019

Reading time: 4 minutes

By Jon MacMillan, Product Manager

For the past 3 years, I have done a little data analysis on Google’s Year in Search. This annual event is always a good reminder of what has transpired over the past year and gives some perspective on how those events ranked. For those unfamiliar, each year Google tells us the top trending searches based on a selection of categories. The number of categories continues to grow, and there are common searches throughout the years: major sporting events, hurricanes and other weather events, and a lot of celebrities.

What has recently dawned on me is the overwhelming popularity of negative news. While Google lists the top recipe searches, or outfit ideas, these pale in comparison to some of the other categories. News, people, and passings are all their own categories that dwarf the competition when it comes to overall search popularity. These categories are overwhelmingly negative. For example, the third most popular news search term in 2019 was “Women’s World Cup”. It peaked in popularity the week of June 23rd:


According the World Atlas, soccer (otherwise known as football) is the most popular sport in the world. One would assume that the World Cup would garner a lot of search interest. However, when compared to an event like the Notre Dame Cathedral fire, we can see how little interest there was in comparison:

Negative Trends

Intrigued, I decided to make “negativity” the focus of my investigation this year. Admittedly, my classification of “negativity” is subjective. For instance, NFL player Antonio Brown was considered a “negative” search term, following his release from the Raiders near the start of the season, before quickly signing and then being released by the Patriots.

Other “negativity” I considered: Indianapolis Colts quarterback Andrew Luck’s retirement from the NFL and 21 Savage’s arrest.

I think I was more than fair, however, by leaving Game of Thrones as a positive search term, despite the more-than-controversial final season that most GoT diehards would consider an abysmal end to the franchise.

Business Insider helps explain why it was was so terrible, but it’s also evident based on the IMDB ratings (shown in the chart below).

Processing the Google Data

I had to pull all of this data down into a number of CSV files from Google Trends, since they only allow for 5 comparisons in a single view. Using Construct, I can quickly combine this data and analyze it.

I soon encountered the first hurdle: that Google Trends adds two header rows to all of the CSV files. This is fine when viewing the files, but when trying to analyze the data we would have additional rows to strip out and would need to determine what the actual column headers are.

I could easily go through and delete the first 2 rows in each CSV file. But I ended up with 32 files and I don’t have the time for that. Thankfully, while Construct is code-free, it is also code-friendly. So in this case, I was able to apply a simple PowerShell Script right in my workflow. This removed the first 2 rows in all CSV files automatically. First hurdle: cleared!

 

Then one of the handiest of features in Construct, the Stack Node, allowed me to collate all of the files automatically. I had this broken out into two separate sets of files; interest rating over time and interest by region. The Stack nodes that you see above append all of the files together in a single step.

Regression Modeling

Once I had the data in the correct format, I still had a little work to do. Because the interest rating can vary widely based on what you are comparing against I had to normalize the data. For this, I used Predict to build some regression models in order to estimate a comparative interest rating for all search terms analyzed. I used this normalized value to estimate the actual search volume. Since a term may be represented in multiple categories, I also had to make some determinations on categories for different search terms. I prioritized them based on the order they are presented on Google Trends. For instance, Antonio Brown is in the top five for Searches, People, and Athletes. I categorized him under Searches, since that was the first list presented.

Key findings:

 

Once I analyzed all of the data I was able to output it directly from Construct to Tableau. I created a dashboard that you can explore for yourself below!

Stay up to date
Subscribe to our blog

Subscribe
Notify of
guest
0 Comments
Inline Feedbacks
View all comments