Professor Linus Wilson discusses his new paper “Clickbait Works! The secret to getting views with the YouTube algorithm on episode 11 of the Finance Professor Podcast.” There is a lot of contradictory advice about what metrics the largest video sharing site in the world and the second largest social network promotes. Using a new data set available to YouTube creators starting in 2018, Dr. Wilson finds that click-through rates are by far the most important predictor of a new video getting views from YouTube’s black-box recommendation system.
The link is https://ssrn.com/abstract=3369353
Click the download button for a free download. Next, if you don’t want to register or login with SSRN, click “Download without registration” under the gray text under the picture of Greg Gordon.
“Clickbait Works! The secret to getting views with the YouTube algorithm”
By Dr. Linus Wilson
In 2018, YouTube began releasing click-through rates (CTR) data to its video creators. Since 2012, YouTube has emphasized how it favors watch time over clicks in its recommendations to viewers. This is the first academic study employing that data to test what matters more for views on YouTube. Is watch time or CTR more important to getting views on YouTube? This paper finds no to limited evidence that higher percent audience retention or and total average watch time per view are associated with more views on YouTube. Instead, videos with higher CTR got significantly more views as did videos on trending or newsworthy topics. The marginal benefit in terms of views scaled by subscribers of increased CTR is between 71 and 318 times larger than the marginal benefits of increased watch time per view.
Journal of Economic Literature Codes: D12, D22, D26, D83, D85, L15, L21, L82, L86, M15
Keywords: YouTube, algorithm, search, discovery, video, CTR, click-through rates, clickbait, watch time, audience retention, neural networks, recommendation systems
Wilson, Linus, Clickbait Works! The secret to getting views with the YouTube algorithm (April 9, 2019). Available at SSRN: https://ssrn.com/abstract=3369353
For all of Linus Wilson’s research go to