Managing innovation as a startup with Roland Sars, CEO & co-founder of Media Distillery

Through AI object recognition, footage can be viewed and analysed to produce next-generation metadata, which in turn can be utilised to boost engagement within video platforms. 

“Media Distillery is a software vendor and we basically apply artificial intelligence to video. Our machine learning algorithms can analyse video in real-time and we can extract, or ‘distill’ as we like to call it, what is inside.

“So these are different algorithms, such as speech, face, logos, topics, subtitles, all those kinds of things. Whenever we process a video in our data centre, all kinds of tags and metadata come out.

“Video used to be a black box. Where a small child can see a video and understand what they’re looking at, whether it’s a cartoon, commercial, or talk show, a computer only knew what the title of the programme was and maybe a short description - but the video itself was always just a black box. 

“We’re consuming more video than ever before and not just because of the pandemic. The trend was always going steeply up, so it’s more and more important to understand what is inside video, because you can do so much more with it once you know.

“We apply this technology to consumer-facing video platforms, the likes of Sky, Virgin and BT. Big players in the TV, broadcast or video industry right now, they all have platforms which bundle traditional linear channels together with VOD catalogues.

“Why do we apply that to this industry? First and foremost there are huge amounts of content being created and distributed. At the moment we are analysing over 20,000 hours of video on a daily basis, completely automatically. We create these tags and metadata, then send it back to these companies so they can increase the discoverability and the engagement of content within their platforms. 

To expand on exactly how Media Distillery’s object recognition AI worked, Roland highlighted a few specific examples of the practical benefits.

“The most obvious one is discoverability. The trend in the industry right now, especially in the younger generations, is that they don’t watch TV in a traditional, linear way. The trend is more towards consumption on demand.

“Another trend, is consumers don’t really know which piece of content is in which TV programme, or on what channel and what time of the day it airs. But they know that they like to know everything about upcoming elections, formula one, a specific football club, a specific person or topic.

“I always explain this with an example here in the Netherlands; one of our biggest sports heroes these days is Max Verstappen, the Formula one driver. If I go to my video platform and search for Max Verstappen I don’t find anything. The reason for that is his name doesn’t appear in the title of a TV programme, despite being talked about every day in sports programmes and documentaries. So if we search for his face or his name in the subtitles, or on subjects such as Red Bull racing or Formula One, we get all kinds of video that can be recommended to the viewer, creating much higher engagement.

“Another is we can identify when exact events are - and by event I mean the exact start or end of a programme, or the exact time of a commercial break. This helps because right now, at least in many countries, if you want to replay a programme that was aired last night at 8pm, the player starts playing at 8pm. In reality the programme might have started at 8:02pm meaning you need to watch the end of the first programme, or the end of the commercial break. 

“The last is that we can extract stills from the video, screenshots to represent content within. Say you have a daily talk show, you might see seven episodes of the same show all with the same stock image of the presenter, which doesn’t tell you anything about the subject or the guests of that particular episode. We can create an episodic image, where our algorithms can automatically detect who appears. This can lead to unique images that look more compelling and are representative of the content that you’re about to watch.”

One of Roland’s biggest frustrations came via the red tape that often has to be crossed when dealing with large businesses.

“Initially, it’s tricky because we’re a small company, like 30/35 people at the moment. But we’re acting in an industry full of dinosaurs, stock-listed huge companies and we’re trying to sell software to them. 

“Obviously we try to act as fast as possible, but there are so many stakeholders in a company, there're so many toll gates you need to move. So the challenge for us is to really keep pace and keep our aggressiveness in this industry while dancing with the rhythm of our potential customers. 

“In the early years that was tricky because we were hungry, we were storming out of the gates and trying to sell our software and these companies said ‘oh that’s okay, sounds very nice, maybe you can come back in two months for a demo’ and I was like ‘why not this afternoon?’ 

“We’ve got used to that, we found a rhythm internally, and once you have enough customers or leads there’s enough to do every day.” 

That desire to be ahead of the curve was a factor in a few of the things Roland had found most surprising since starting the business. These realisations also led to the business rethinking and developing their own offering to match the needs of their customers. 

  “I found it surprising how little these big companies actually know about the content they’re creating or distributing. So talking about huge video platforms, or TV operators, if they usually get content from content creators or broadcasters, they distribute it through their platforms or set top boxes to consumers, but they don’t have any clue of what’s inside. 

“Once you want to make meaningful and valuable recommendations, or you want to create valuable search results, you really need to start understanding what the content is. I always say content is king, but the metadata is the emperor, because if your content is not tagged or not enriched with metadata, there’s no way to discover it.

“Another thing is how we came in four years ago explaining this example of Max Verstappen, and potential customers would say ‘this sounds awesome but you’re three or four years too early, we have much more urgent problems’ - like the start times or the image indicators. 

“We learned that we’re always on the edge of what is scientifically or academically possible in terms of machine learning. However we really needed to take two or three steps back in terms of where these platforms were in terms of their own product development and innovation. This is always the case when trying to disrupt an industry, you’re always a few years ahead.”

When it came to assembling his team, Roland revealed it came down to striking a balance between theory and production, while allowing employees to work on something truly tangible. 

“I think our team is a nice mix of academic people that really understand the latest with regards to machine learning. However, what we have done from the beginning is not just stay on the theoretical or academic level, but to think how can we really add value in a production sense? How can we really bring whatever we find out in theory into production? 

“That’s what makes our company really interesting to work for, or at least that’s what I hear from our colleagues. They say often when you’re a machine learning graduate you can start working for insurance companies or banks, working on models that can optimise insurance products, but it’s not really tangible.

“With Media Distillery, they can really touch something which everyone knows and understands. Everybody watches video and has emotions connected to it. Our team can really do something that has an impact on the millions of video viewers that we look after.

We currently reach about 20 million viewers on a daily basis, so that has a huge impact and that’s what they really like.”

In terms of culture, Roland says the values instilled within his business have been built around three key pillars. 

“The first is innovation. What we have tried to define in the culture, and I think we’ve succeeded, is kind of a fail-fast approach. So try things, it doesn't matter if you fail, we won’t blame you or anything like that but you need to learn from it. 

“Try something, if it fails, do it again, grow from it and try something new. 

“The second is collaboration. We’re a very close team, very smart individuals that all have their own expertise and what they’re good at.

“However, everybody’s T-shaped. They have one speciality but also have some interest in a broader subject. So someone might be very knowledgeable in machine learning, but also have an interest in Java technology. So in that way, we shape very high-performing teams that really get the best out of each other and help maintain a high level of well-rounded knowledge.

“The last is energise, l mentioned the industry that we’re in, it takes a lot of energy to get through to large companies and do business with them. We do everything with a lot of energy and we show our power and enthusiasm with whatever we do both internally and externally.”

While every business will differ in its approach to raising funds,(Read about our capital domain) transparency is something that should be present in any effective funding strategy. While it may be difficult to admit faults in your operation, keeping them hidden will only do more damage than good in the long-run. 

“An important element is honesty. Obviously, you’re pitching your company to someone you want money or knowledge from, so you’re pitching your business in the most positive way.

“My experience is that it’s important to be open and honest about things that you need help with, things that challenge you. This openness can also open the eyes of investors to where they can really add value. No business is perfect, every company has challenges, especially in the startup phase. 

“Just so if they do come on board, it doesn’t result in surprises once the deal is done. I compare investors to a marriage, it comes with openness and trust.”

While discussing some of the frequent mistakes entrepreneurs make, Roland explained how it can become easy to stop working on yourself as focus shifts to improving your team. (Read our People and Culture Scaleup Guide) . 

“One thing that comes to mind is a question that I sometimes get asked by colleagues. We have a learning budget and we very much stimulate everybody to go to conferences or educate themselves with courses or whatever.

“Sometimes they ask what I do as a CEO to learn and to educate? I think that’s one common mistake entrepreneurs make, they don’t invest enough in developing themselves.”

Roland says that was actually one of the key inspirations for him joining Rise, Techleap’s own acceleration programme in the Netherlands dedicated to startup growth. 

“We wanted to develop ourselves and learn from peers, because it doesn’t really matter what industry these peers are in. Everybody’s facing similar problems or challenges when it comes to things like HR, funding and scaling internationally. 

“So develop yourself, don’t just think you're the lone CEO or founder and you need to know everything, that’s a common mistake. It’s very natural to be fragile or vulnerable as a CEO, so just open yourself to experiences to learn.”

For obvious reasons, COVID saw demand for Media Distillery’s object recognition software rise higher than ever before. However, they’re not exempt from the common issues hitting most businesses like not being able to meet face-to-face with potential customers and the toll that working from home inevitably takes on staff. 

“Our industry is lucky enough to not be too badly impacted. Actually the amount of video consumption just grew, therefore the importance of video platforms did too.

“What is a challenge, is that we rely on b2b enterprise sales, which are about building relationships and building trust on what you can deliver in terms of value. This is often done by on-site workshops, lunches, conferences being on stage speaking about your technology - all things that were not possible in the past 12 months. 

“When you’re talking with existing clients, it’s not that hard because that’s already a  relationship. However, building new relationships from behind the screen, that’s difficult. Where usually we’d have a three-hour workshop on-site with a lunch in between, now it’s often a 45 minute / one-hour online session.

“This isn’t just a challenge for us, it’s also a challenge for our customers. They also have roadmaps and innovation projects going on, so they have to select and meet vendors. They would usually do this by going to conferences or attending trade shows - but they now need to do this through webinars or by Googling. 

“That made the business a lot less fun because travelling and meeting people is something I like very much.

“Internally our story isn’t that different from others. We’re all working from home with the exception of two or three people. In the beginning, everyone was full of enthusiasm, but now it’s been a year, the lockdown in the Netherlands got stricter, the days got darker, the weather is cold, people are kind of fed up.

“You feel that people are really missing the vibe and relationships with colleagues, the informal coffee chats, and that kind of thing. I think we’re doing a good job, we’re doing employee surveys very often; checking how we’re doing as an employer? How are our staff feeling? What is their mood?”

Upon being asked to recall the best piece of business advice he’d received, Roland shared a suggestion from a podcast (have you checked out our podcast?) that he now utilises when it comes to approaching potential customers. 

“One thing that comes to mind has to do with selling to enterprises. It was in a SaaStr podcast, they basically said try to position yourself much further ahead than you actually are. Present powerpoints and demos as if the product is already there, as if you’re ready to go live tomorrow. Because eventually when such a big enterprise buys a product, it will take months before they’re actually ready to deploy.

“Present yourself larger than life, that’s what we started doing. We presented from day one like we’re ready to deploy tomorrow when actually we’re not - but customers were never ready to deploy tomorrow. 

“I think this gave us the advantage in that if we had gone in and said ‘we have this idea, but we still need to develop it, do you think it’s something that you would be interested in?’, then I don’t think we’d really resonate with the large companies. I think the only way that works is to really present yourself as if you were three years ahead of yourself.”

While remaining on the subject of advice, we also asked Roland to call on his own priceless experiences, and offer his fellow entrepreneurs three standout learnings. 

“First is if you have an idea, try to do it with a team. I think personally, entrepreneuring or starting a business with somebody else is a lot more fun than just doing it on your own. 

“Second, surround yourself with good people. It can be advisors or mentors, someone you really learn from, who you can speak to about challenges and get answers from. Being an entrepreneur or a CEO of a small company can be lonely because it’s hard to share your challenges, so it’s good to have a team around you. 

“Lastly, I think it’s important to really focus on the exact value that you’re delivering, really go deep on that product-market fit. You often see entrepreneurs that just start and they think they have a great idea but they forget to really test it. Then they fail or grow slowly because they forgot to really nail that product market.”