Problem Statement

In this report, we explore the following:

• What type of content should we invest in based on region?

• Which content should be promoted to increase trending views?

• What type of audience-engagement predicts video views?

• Where do patterns overlap or diverge across types of audiences (e.g., regional differences)?

With these questions in mind, our proposed solutions include:

• Content recommendations optimizing trending-viewership

• Region-specific recommendations for content promotion

• Metrics to gauge promotion success tailored to content audience

Content Insights

What content is more likely to trend?

While 67% of trending videos fall under music, entertainment, blogs, how-to’s and comedy categories,

57% of total viewership for trending videos are categorized under music and entertainment.

Bar plot depicting the total viewership by category of video.

Which types of videos trend between the United States and Great Britain?

Regional differences between types of videos and viewership exist.

Certain categories, such as Science & Technology, trend in the US over Great Britain and vice versa. Should we promote content under different categorizations, it is optimal that we ensure the video content is trendy for that specific region.

Tree map depicting a breakdown by region in the total viewership for each category of videos.

The following types of content trended most overall in the US alone:

• Music (216 Million)

• Entertainment (88 Million)

• Comedy (63 Million)

• People & Blogs (48 Million)

• Science & Technology (44 Million)

The following types of content trended most overall in Great Britain alone:

• Music (304 Million)

• Entertainment (63 Million)

• People & Blogs (64 Million)

• Comedy (50 Million)

• How-to & Style (43 Million)

Categories of videos in bold pertain to region-specific top 5.

Among these groups, which specific types of videos provides the greatest viewerships?

Movie trailers, pop music, and Rhett Link trend strongly among others.

Using word-frequency count on video tags, we identified trending patterns within each category of video. Should we promote content that falls under these trending tags, we should expect greater viewership. Other hot-topic trends include humorous content, makeup tutorials, and more.

Word-frequency plots have also revealed short-term trends in Youtube videos.

Although videos under the Science & Technology and News & Politics categories trended strongly in the US, we expect viewership to be short-lived.

According to the tags under these trending videos, most views arose from period-specific events such as Apple’s reveal of the iPhone X and Hurricane Irma. Thus, we recommend endorsing other videos that share more longer-lasting trending power such as those under Sports, People & Blogs, and Comedy.

Which channels produced the most trending videos during this 17-day period?

The grand majority of channels only had a single video trend for the 17-day period. Very few channels had more than 3 videos trending.

Bar plot depicting the number of unique videos trending per channel across the entire 17-day period.

We found that the top-producing channels for viewership content included The Tonight Show with Jimmy Fallon, Jimmy Kimmel Live, as well as Vox media and ESPN sports.

Following trends mentioned above, it would be in the best interest to endorse content from these channels.

Viewer-engagement Insights

What predicts greater number of views in trendy videos?

Number of likes strongly predicts trending viewership above other metrics (r = 0.86).

Scatter plot depicting relationship between number of likes and number of views per video. Values have not been transformed for this analysis.

Dislikes and number of comments were poor predictors of viewership totals across all data. Likes and no other metrics are also found to predict viewership similarly across categories and regions.

These findings may suggest that content-likes may be a good measure for performance of content promotions.

How do audiences from different regions engage with trendy content?

British viewers are more likely to comment and “like” trending videos they watch. In contrast, American viewers are more likely to “dislike” video content.

While we found that liking videos were a strong predictor for viewership, we identified differences in how viewers engage with content across different regions. We suggest taking careful consideration of KPIs utilized in measuring promotion successes when comparing regional viewership.

How do audiences from different types of videos engage with content?

While the majority of viewers behave similarly with how they engage with trending videos by category, top-trending categories, as well as, in Film & Animation depict greater returns in “liking” content per view.

Boxplot depicting the proportion of likes to views as a percentage for each category of video.

These findings may be due to differences in audience polarity of video content. For example, News & Politics may elicit greater “action” among viewers due to the nature of content presented. Another, explanation may allude to greater engagement through fan-based interactions. Fans with stronger attachment to video content may be more likely to “like” content over casual viewers. However, further analysis is necessary to understand the mechanisms for these trends.

How long do videos trend for?

52% of videos trend for up to 3 days between the US and Great Britain.

In general, we found that videos trend for a maximum of 9 days, with the majority of videos trending only for 3 days or less. The number of videos peaks on a 2-day trend duration and drops until day 9.

Interestingly, this pattern is most present in the US-based trends. Great Britain, by comparison, shares more stable trending videos from a duration between 1-4 days. We do not see a drop-off until day 5 of trending. Additionally, Great Britain has longer-lasting trends overall where the US maxes out on 7-day trend duration.

This suggests that Great Britain viewership may be more stable than the US and longer-lasting. These findings should be taken into consideration should we consider setting a KPI for trend-duration.

Bar plot depicting the number of videos trending per number of days broken down by region. Number of days trending per video were calculated by taking the number of days a video was present for in the trending-video dataset.

Some videos at the start and tail-end of the 17-day period may have already been trending for a few days/will be trending for longer, but we cannot capture their trending timelines accurately unless we select a larger timeframe.

Among top trending categories, a greater portion of Music, Entertainment, and People & Blog videos trend longer than 3-days.

Tree map depicting breakdown of trend-duration by category of video. Longer-lasting videos are categorized as trending video that lasted beyond 3 days of trending (exclusive). Short videos only trended for 3 days or fewer.

By comparison, we found that videos under News & Politics, as well as Gaming trend for very short periods. These patterns may be connected to the greater volatility of US viewership, in addition to short-term content bursts from political events occurring during this time-period.

Which days do trending videos receive the most views?

30% of average viewership in Great Britain occurs on Tuesdays and Fridays.

Trending videos receive more views between Tuesday and Wednesday in Great Britain and the US respectively.

Viewership drops between on Saturday, picking up on Sunday once more.

Line plot depicting average number of views by day of the week.