Restoring data signal to audience-based analysis using contextual classification - without relying on PII
Data cleansing is an unavoidable, cumbersome first step and constant consideration throughout any AI or machine learning endeavor. Across the data science field, experts estimate that up to 50%-80% of the time invested in development can be spent simply on obtaining, migrating, merging, and cleaning data used in ML or data science projects. Thoroughly reviewing data to ensure that keywords “water skiing” (no hyphen) with “water-skiing” (hyphen) or “water ski” are grouped together is essential but a low-value and often time-consuming exercise for teams that is, in most cases, better suited for machines.
In today’s soon-to-be cookie-less environment, marketing breakthroughs to establish robust audiences are among the most pressing demands facing data scientists and ML teams working with media owners or brands. Speed is critical as the shift approaches. With Netra’s technology, data science teams have immediate access to a consistent taxonomy categorized with IAB standards through Netra’s automated processing. Data scientists can immediately start work with clean datasets on the complete set of content assets (text, imagery, and video). For data science teams, Netra’s technology eliminates the enormous hurdle of categorizing content and transforms once un-penetrable video content into usable data.
Within the industry today data science and product teams at CTV, online video, social media and measurement vendors are already using Netra’s comprehension platform to build a common data schemas to power better products and data science outcomes.
By providing robust and clean data classification about content, Netra opens the door for data science teams to restore the signal lost after removing cookies and PII-based audience targeting. Using Netra’s technology to implement precise contextual segmentation unlocks useful signal information on audience intentions or characteristics that is not currently available from existing third-party providers.
For example, consider a data science team that wants to create a pool of users interested in Caribbean-location vacations. The solution is to create a segment of users that is high-propensity for targeting, particularly within video content viewed. This content can include tourism videos, Caribbean travel-related content, and even more nuanced themes such as keywords of “Grand Cayman” or “Barbados,” or even related images. Netra’s platform with Active Learning can be quickly trained to identify video content that is associated with the Caribbean Islands. Once these elements are identified and trained, Netra’s AI creates a high-propensity Caribbean segment. As short-form video content is scanned through our API with CV, any video that contains related Caribbean content through our frame-by-frame analysis is identified and tagged. Other elements around the scene and context are also identified and relayed back through the API.
Further, brands and publishers that have first-party data can match these audiences via tech like clean rooms, then extend the overlap to build look-alike audiences based on contextual similarities of the matched audience. Traditionally, most independent publishers lack scale with their first-party data, deterring the appeal of this technique to brands. However, with a better comprehension of the content, publishers can help brands combine the PII safe attributes of first-party data with the scale, signal and privacy-first nature of contextual targeting.
These segments become powerful resources for data science teams, as they can create other highly relevant segments and profiles of audiences through a total comprehension of all assets, including text, imagery, and video. These segments lead to a virtuous cycle in that the high quality of analytics unlocked through segmentation can inform future targeting and segmentation revision to improve performance.
We aim to provide an infrastructure for all, through which every company can build their own applications and products to start unlocking the value of their video assets.
To learn how data science, product and audience teams are using Netra’s APIs to better comprehend and tag their content and creative to deliver better audiences at scale, please contact us.
With AI-based technology, classifying creative creates richer audience targeting and better measurement
Global Alliance for Responsible Media (GARM) is an industry initiative that has adopted a unified language to describe harmful content.
Total Content Comprehension (TCC) is the complete understanding of all the content on a page including text, audio, image, and video formats.
Get the latest on the application of Computer Vision and Artificial Intelligence and its use across all types of visual content.
