TONIK+ Video Intelligence: a data-driven approach to the art of remixing content
Status Quo
Are you getting the most out of your content? Chances are you aren’t. Most companies reuse portions of their long-form content for shorter spots. For example, a movie studio may cut a 30-second spot from content in their 3-minute trailer, or a fitness app may cut a 60-second ad from a combination of multiple short films. How do you decide which scenes are used for those shorter pieces? More often than not, the decision is a subjective one, based on what the creative whims of a handful of employees in a conference room. This is an extremely small sample size and one that is not indicative of the overall population of humans who were served your content. The best case for the new spots is that those employees were able to pick the top scenes with 100% accuracy and stitch them into a cohesive narrative. The common case is that the resulting ad will be middling and won’t move the needle either way, with a strong chance of a spot that ends up underperforming. TONIK+ doesn’t take chances on subjective opinion when we can empirically determine the best scenes to use for future content, and neither should you.
TVI
Analysis of content by TONIK+ Video Intelligence (TVI) is informed by two sources: retention data and video intelligence via machine learning. Video retention determines how many users are still watching a video at a particular point, and serves as a reflection of quality. A video with a high completion rate is objectively better at keeping viewer attention than an equal-length video with a low completion rate, as it indicates that fewer users left before finishing. Now extend this line of reasoning to quartile retention data, i.e. Video Watches to 25% of the videos’ length, Video Watches to 50%, Video Watches to 75%, and Video Watches to 100%. If two equal length videos A and B have first quartile Completion Rates of 80% and 70%, then the first quarter of video A was higher-quality than the first quarter of video B. This can continue to be extrapolated for even more-granular retention data. We process retention data on a scene-by-scene basis, giving us precision insights for every piece of content. This requires more granular data than the typical quartiles since a video is rarely going to consist of only four equal-length scenes. Once this data is obtained, we can transform it into a scene retention score through proprietary statistical transformation.
Video intelligence forms the second major input for our content analysis. Once scenes have been identified, the video intelligence algorithm identifies characteristics contained within that scene, shot-by-shot. Our video intelligence software currently identifies aspects of the shot (“outdoor”, “taxi”, “smiling woman”, “sunny” and such) as well as “proper nouns”, allowing TVI to identify celebrities, cities, and more. These characteristics are coupled with the scene retention scores and transformed into a TVI Score, a 0–100 rank score of each scene within a piece of content. The top scene will score 100, the bottom will score 0, and the distribution of the rest of the scenes will correspond with the overall quality of the content itself. A piece of content that sees scenes scoring consistently in the 90–100 range will be of a higher quality than a piece of content that sees scoring mostly in the 60–80 range. From there, identification of the top-performing scenes, the narrative throughputs across good and/or poor scenes, and recommendations for remixes and future content are generated and passed along.
Scale
While our TVI analysis is groundbreaking and effective, we aren’t limited to video-by-video analysis. Multiple pieces of content can be daisy-chained together to analyze performance across all content, providing the best-of-the-best scenes, an understanding of which piece of content is scoring better than the others, and the opportunity to craft remixes that capitalize on the top scenes across your entire portfolio of work.
TONIK+ can also create a virtuous cycle of TVI analysis and remixing. TVI analysis of a 2:30 trailer can be used to create a 60-second spot. If this 60-second performs well, it too can be run through TVI analysis to create a 30-second spot. That 30-second spot can be used for a :15, the :15 can be used for a :06 or :10. This process of analysis and distillation of content can keep producing high-performing remixes from the original piece as the length of spots required decreases.
Finally, TVI is also able to increase content efficacy by analyzing the performance of scenes on an audience-by-audience basis. Typically, retention data is stored as an amalgamation of all viewers that feed into the video: organic and every audience served by paid ads all fall into the same overall retention data bucket. Through the use of some slight tweaks to a campaign ahead of time, we are able to fully capture isolated paid audience retention data, providing a clear picture of what a particular cross-section of users value in your content. This can be stacked against other audiences and even a wide-release version of the content, providing not only analysis between paid audiences, but how those audiences prefer scenes relative to the overall organic population.
These unique audience “fingerprints” can then be further used to craft remixes specific to that particular audience. If Males 18–34 favor scenes A, B, and C and Females 18–34 favor scenes D, E, and F, remixes can be constructed that only contain A/B/C for use with the M18–34 audience and D/E/F for use with the F18–34 audience. This granular focus on audience boosts performance even more than the standard analysis due to the comprehensive drill-down into a particular audience’s preferences.
A Better Way
TONIK+ Video Intelligence provides a data-driven approach to the art of remixing content for mass distribution. By pairing statistical analysis of retention data with video intelligence via machine learning, our TVI Score can effectively determine which scenes are resonating the most with your user base, cut that into effective content, and massively improve your results. We can scale this to be platform and audience-specific, giving you limitless options for tailoring your content and ensuring your message resonates.
