EX.COthe publisher video platform that powers successful video strategies for the world’s leading media groups, today (August 19, 2024) announced an advanced contextual video content recommendation engine for digital publishers. The unique machine learning-based engine uses large language models (LLM) to deliver the most relevant videos from a publisher’s video content bank to audiences in real time. This enables more scalable video integration across an entire site, without the need to create specific content for each article or manually match articles to content.
Autovia, the UK’s leading automotive content and commerce company with trusted brands such as Auto Express, Carbuyer and evo, is one of the first of many publishers currently using EX.CO’s enhanced contextual recommendation engine for their network of websites.
“Context and relevance are critical elements of our business as we strive to provide the best possible user experience,” said Ciaran Scarry, Head of Advertising at Autovia. “EX.CO’s contextual recommendation engine improves our user experience by allowing us to tailor video recommendations to fit the specific content of each page. This is critical for our target audience as they research their next car purchase. This seemingly small adjustment is a huge game changer for us as it provides highly relevant content that engages our readers and piques their interest.”
The LLM-based engine vectorizes text, calculates the similarities between articles and available video content, and then ranks the results to deliver the fastest, highest-quality recommendations. For publishers who need additional video content for their pages, the engine can also access EX.CO’s massive content marketplace, which offers thousands of high-quality videos across various industries from premium sources.
“Today, audiences are only interested in content that is truly relevant to them. However, it is challenging for publishers to produce and match such large volumes of video content,” said Tom Pachys, co-founder and CEO of EX.CO. “After in-depth discussions with publishers and our own data research, we realized that the old ‘tags/taxonomy’ based approaches were not enough. By integrating LLM capabilities with ML optimization models, we developed a new generation recommendation engine. We were surprised by the immediate results, which outperformed our previously sophisticated models, which are still considered best practice in this field.”
The video recommendations are delivered “at the speed of news,” resulting in greater audience engagement and retention, which in turn has the potential to increase other key KPIs such as revenue, brand loyalty and subscriber growth.
EX.CO’s improved contextual recommendation engine is now deployed to EX.CO partners. Top publishers achieve 80% relevance match rate, resulting in 4x higher engagement rate with the video player than the industry average. In addition, average negative engagement with the video player has decreased by 30-40%.
The technology currently optimizes video recommendations based on criteria such as media category, title, recency, sentiment, keywords, and length. EX․CO plans to soon expand this offering to include ChatGPT-like functionality that allows publishers to refine video recommendations using prompts. This trains the engine to provide more relevant recommendations for specific sites, sections, and articles.
