Personalization, Machine-Learning & Amazon: What’s Next in Content Discovery
Given the overwhelming options viewers have when choosing what to watch, it was only a matter of time before tech giants started offering entertainment discovery tools. At TiVo, we’ve been developing our personalized search and recommendations platform for several years, and we believe there’s room in this space for multiple players with diverse options catering to industry needs.
Amazon recently released their personalization toolkit: Amazon Personalize, the AWS (Amazon Web Services) API-activated, eCommerce-focused product, operates via algorithmic data processes. We wonder, as reThink does, do AWS customers want a master personalization algorithm? Will it ultimately improve the viewer experience? Is the product flexible enough to meet their needs?
Google has also been dabbling in cloud-based discovery services with an AI-powered product similar to Amazon Personalize that allows customers to upload data sets for training models. The use of Google data here is of interest, but like Amazon, it’s a black-box solution and doesn’t offer a breadth of flexible capabilities.
Amazon Personalize appears to be focused on the company’s retail expertise, rather than the unique and ever-changing needs of the world of digital entertainment.
Search and recommendations tools are vital for surfacing premium data in entertainment discovery systems. The recent entries into the space we mentioned don’t include robust search functions, rather, they operate through lists of items their services reorder or “rank” based on viewer preference.
For recommendations, these new products appear to support single use cases (e.g., a single recommendations carousel of new products) instead of including many items populated across the UI. TiVo’s Personalized Content Discovery is a superior solution because it supports multiple use cases, which is essential when trying to present sophisticated entertainment carousels for display in a stunning UI.
Amazon’s suggestive recommendations, as an example, can be based on three different model types (called “recipes”): user behavior, a combination of user behavior and metadata, and cold-starting to promote new items in the catalog. These recipes are independently generated and called, and may involve additional costs if multiple models are built. TiVo’s Personalized Content Discovery product includes suggestive recommendations but also powers predictive recommendations, which ingest viewership and engagement signals unique to each customer for surfacing content based on viewer behavior.
The trending/popularity data models of these new products work to surmise what viewers are most likely to interact with. TiVo’s personalized, machine-learning-powered predictions truly learn patterned behavior and help expedite time to content.
In addition to being relevant and accurate, search and recommendations experiences should be aesthetic and easy to navigate. It’s not clear if these new cloud-based services are making significant investments in customizable, dynamic interfaces or simply integrating triggers from first-party data sets into existing UIs. TiVo has been building beautiful UIs for customers across the landscape for years, and our discovery solutions are easily integrated into these experiences for best-in-class content delivery.
While smaller businesses might be tempted to start offering machine learning to connect viewers to entertainment with a black-box solution, implementing content discovery services eventually requires a certain level of expertise and attentive customer service. And, in an industry that needs more support than ever to maintain customer loyalty – given the rapid growth and deepening fragmentation of content catalogs – it’s vital to select a partner with years of proven entertainment experience and expertise, like TiVo.
© 2019 TiVo Corporation. All rights reserved. TiVo and the TiVo logo are registered trademarks of TiVo Corporation and/or its subsidiaries worldwide.