Media Case Study

Tier One Media Company

One of the world’s largest media companies leveraged Cortex to keep website visitors engaged by 1:1 personalization of which videos to watch. A key goal for this media company is to engage users onsite and in-app in order to drive incremental ad revenues. In this case study, we worked with the media company to compare various algorithms for increasing onsite engagement including “most popular” and “most recommended” and found that Cortex recommendations far outperformed “most popular”. Furthermore, we found that increasing the amount of content available for recommendations resulted in additional increases in performance.

Challenges

One of the main revenue streams for the media company is advertising, and video ads in particular drive a premium CPM. Increasing video ad inventory can have a major impact on ad revenues. They chose to use Cortex to optimize 5 slots on their site with personalized videos in order to increase CPM revenue from video advertising.

Some of the questions the media organization asked included –

• How much would “recommendations” outperform “most popular”?
• Which specific videos should be recommended to each user?
• How would increasing the amount of content impact CTR?

The media organization was interested in optimizing the content shown to each user in order to drive higher high-value at CPMs.

Using Cortex

The media team turned to Cortex to help prescribe the best videos to recommend to users. “Most popular”, in which the most popular videos over a 24hr period were recommended, was used as a control group and compared to various iterations of “recommendations” containing different amounts of available content for the recommendations algorithm to choose from. The recommendations algorithms chose from ~50 and ~300 pieces of content respectively by increasing the time-period over which the algorithm could select content.

To summarize the 3 algorithms used –

Results

Results were tabulated over a 3 month period and visualized using Cortex Insights. The results can be summarized as follows –

1. Cortex recommendations outperformed “most popular”
2. Increasing the amount of content available for recommendations to choose from increased CTR performance. I.e. by giving Cortex a bigger library of videos to pick from, users have a higher chance to see something that interests them. 

Both the Vidora recommendations generated a higher number of clicks than the control group of most popular videos. ‘Vidora 2’ generated more clicks than both the control group and ‘Vidora 1’. 

The impact on available advertising revenue was also significant for the media company. The available video ad inventory from this module increased by $1.5m per year. In summary, the media organization was able to not only demonstrate the efficacy of using Cortex recommendations to increase engagement, but also impact top-line revenue opportunities.