Lessons Learned In A Decade Of Distilling Video

By Roland Sars, CEO & Co-Founder and Geert Vos, CTO & Founder, Media Distillery

Media Distillery recently celebrated its 10th anniversary. In this blog, our CEO & Co-Founder, Roland Sars, and CTO & Founder, Geert Vos, have taken the opportunity to look back on our first decade of distilling video and share some of the lessons we’ve learned along the way.

“The idea for Media Distillery stemmed from a research project originally conducted at the Netherlands Organisation for Applied Scientific Research (TNO) by my co-founder Joost de Wit,” Geert Vos explains. “In 2014 we formed the company with the goal of applying the latest advancements in speech recognition, image recognition and AI technology to enable video search capabilities similar to what Google had done for the web.” By leveraging these technologies, Media Distillery aimed to revolutionize the process of discovering and searching within video content.

“It’s incredible to imagine from the vantage point of today, where we analyze over a thousand channels 24/7, that we originally began with just three,” says Roland Sars, who joined the team a short time later. “In those early days, our focus was mainly on brand recognition. Using state of the art technology to detect which brands or people are visible or mentioned within a video.” 

Geert recalls the frustrations of the technology available even just ten years ago: “Although Artificial Intelligence had been around since the 90s, the hardware available to run it on was still not powerful or fast enough to do the kind of computations we’re all used to in 2024. We were still in what’s eloquently known as the ‘AI Winter’. So, we had to build specific models that would enable our technology to recognize each of the 200-300 logos and images we were looking for within each video analyzed. It wasn’t until 2016 that we began experimenting with AI, training deep learning models to look for all logos and faces at once, rather than just the ones we’d programmed in.” 

The customer base was also quite different. “Although we were working with streaming video from day one, our earliest customers weren’t in the streaming business at all,” says Roland. “Unlike our current market of Pay TV operators at OTT service providers, our early adopters were governments, press agencies and media monitoring companies who wanted our technology to reduce the manual labor of finding out when key people, topics and their brands featured in TV coverage, for reputation management and media value determination.” 

Just the four of us: The first day in our new office back in August 2016

Finding our feet in Pay TV

It was 2017 when the still-small Media Distillery team realized the potential for this technology to solve some of the major issues faced by the operators of Pay TV services who were facing new and serious competition from streaming companies like Netflix. In 2017 they met with Telenet in Belgium, now part of Liberty Global, to demonstrate how our technology could identify topics being mentioned in a live TV show. “This was the prototype for what would later become our award-winning Topic Distillery solution,” says Geert. “The team at Telenet were interested, but they had a bigger problem they needed to solve first and wanted our help to see if AI was the answer. Their challenge was with the start and end times of TV programmes within the channels supplied to them as an operator. The actual timings of transmission often differed significantly from the Electronic Programme Guide (EPG) which was used to create the replay clips. This meant users often faced a frustrating process of fast-forwarding through a long preamble before they got the show they actually wanted to watch. Or the replay was missing the very start or very end of the show. All together it was a poor user experience for an audience getting increasingly used to the instant starts of a streaming VOD service like Netflix.”

Roland remembers how quickly it was possible to adjust the existing Media Distillery solutions to address this need: “Pretty soon, we were running our first Pay TV proof of concept. We started small. Just two or three channels were analyzed using software running on a laptop. But the results were a runaway success and soon we had our first Pay TV customer. Telenet began using our EPG Correction Distillery solution to analyze 50 channels and improve their replay user experience. It’s something we still do for them to this day.”

Geert and Roland celebrate our first awards win: The Connies 2018

Focusing on what we do best

EPG Correction rapidly became the most sought-after part of the Media Distillery product offering, winning a series of innovation awards and picking up clients from Portugal to Chile. The company itself was changing too! “We had grown from a start-up with just five employees to an established presence in the streaming industry with more than 20 staff, says Geert: “In 2020, we made the decision to sell-off our media monitoring business and focus solely on the work we’re doing with Pay TV and OTT service providers. Since then, we’ve grown even further and managed to be ahead of the curve that now sees almost every company scrambling to add AI technology to their platforms and products. It’s an exciting and ongoing journey.”

Roland is adamant that innovation will always be at the heart of the business: “We’ve continued to prove the value of EPG Correction, making the most of developments in both AI and cloud computing to deliver increasing speed and simplicity to our customers. At the same time, we’ve launched new, award-winning solutions like Ad Break Distillery and Topic Distillery that really lean-in to the possibilities offered by Generative AI and semantic search.”

If you’d like to know more about what we’ve learned about AI over the past 10 years, you can download our e-guide, Navigating the OTT’s AI Hype.

So what have we learned in this exciting ten-years? Here are three key lessons we’ve taken from our first decade of distilling video:

Lesson One: Find your niche and super-serve it

“From day one, we’ve offered a complete, turnkey solution to our clients that is solely designed for the needs of our target market, rather than off-the-shelf building blocks that have to be heavily configured or tuned by the customer to their specific use cases,” says Roland. “It was a deliberate decision and it’s been validated by the success of our solutions. The big technology providers offer their own AI tools that can do image recognition and speech-to-text, but the reason Media Distillery’s solutions are so valued by our customers is that we tailor them specifically to Pay TV and OTT use cases. We know this business and we understand how the technologies we work with can be used to tackle real problems with very real impacts on subscriber retention. An AI tool can help you find a car or a duck in a video, but you need a tailored solution to help you find a video segment about cars (rather than just a clip that features that car in the background) and create an appealing chapter marker that will encourage viewers to watch it.”

Lesson Two: Don’t give up on a good idea

“For a long time, our big idea has been this: How can we help decrease the time it takes for the end consumer to get to the content they love,” Geert explains. “EPG Correction is a big part of that, and it solved a pain that our customers were hearing about loud and clear in their NPS scores and user satisfaction surveys. But we’ve also been talking about the concepts behind Topic Distillery (search and discovery, chaptering etc,) and Ad Break Distillery since 2017. Back then, there wasn’t the demand. Video service providers just didn’t see that content discovery was a big problem. Today, that’s changed. They’re all struggling to help consumers find and engage with growing content libraries, and as the market matures we’re definitely seeing a big rise in interest for AI-powered solutions to solve this challenge. We always knew there was a need for these smart solutions to content discovery. Now our customers see it too. It’s all too easy to give up on an idea if you don’t see instant interest, but it’s important to stick to your convictions and keep working away at the ideas you believe in.”

Lesson Three: Innovation is a team sport

“In 2024, Media Distillery has grown to a team of more than 40 people and we were delighted to celebrate our anniversary with them, alongside many former Distillers who returned to help us mark the occasion having gone on to exciting new opportunities after their time with us,” says Roland, recalling the recent 10th Birthday party in Amsterdam. “We would absolutely not have reached our 10th anniversary without the support of our customers, partners and our amazing team of Distillers. Our customers have helped lead the direction of our product development. We’ve worked together with them to find and refine the best solutions for their real-world problems. In the process, they’ve made our solutions stronger and helped our business grow. They’ve also gone on to be our advocates. We could not be more grateful, and we could not be more excited about where we’ll go in the next ten years of collaborating with customers to tackle challenges and delight subscribers.”

A major milestone: Geert and Roland at our silver and gold themed 10th anniversary party

What’s next?

If you’d like to know more about where Media Distillery is going next, or how our AI-powered solutions could help your video service to delight and retain subscribers, we’d love to chat.

May 7, 2024