Bibblio’s CEO Mads Holmen on Netflix’s earnings through content recommendation.

From content to digital experiences

Our CEO Mads Holmen spoke at the London Book Fair 2017

All the new ways in which content can be treated are found in the Tech area at London Book Fair. Our co-founder and CEO Mads spoke at The Tech Theatre, located on the show floor in this snazzy space.

Below you’ll find a transcript of his talk, From content to digital experiences, discussing the attention economy, the changing role of publishers and what experiences they need to offer to engage their audiences.

Kicking off with a bit on Bibblio

I’m Mads and I work for Bibblio. We are a content recommendation engine. You most likely will never see Bibblio because we live on other people’s websites. And if we do well, you’ll never notice that it was us that helped you find the right content. Imagine you’re reading an article and when you get to the bottom of it, you get some related, recommended content. That might just come via us. It’s still the website’s own content, but the calculations and computations will be done by Bibblio.

Time well spent

I promised I would talk about innovation in education, but I’m going to start a bit wider than that. As a civilisation we spend over 10 billion hours a week on Netflix, YouTube and Facebook. That’s a lot. Most of what we do on those platforms is decided by algorithms.

It’s all a recommendation

When you load up Facebook and see the News Feed— that’s a recommendation engine. When you log in to Netflix and you see your home screen — that’s a recommendation engine. When you see Discover Weekly on Spotify — that’s a recommendation engine. In fact, if you go across these platforms, over 80% of the time spent on those platforms is generated by algorithmic recommendations. The vast majority of what people do on those platforms now is decided by algorithms, and not by humans.

Advertising-driven environments

This has some pretty important implications. One of them [fake news] we are beginning to talk about now. I’m quite pleased with that because this has been our world for a number of years but now the rest of world is starting to ask questions about it too.
This is especially important in an advertising-driven environment. I like to remind people that as long as they are not paying for anything — and that site is relying on advertising — they should remember that they are in fact the product. It is really Facebook selling us to advertisers.
They want us to stay for as long as possible to consume as much as possible. They don’t really care what we consume. If you’re happy with cat videos all day, to use a cliché example, you’ll get cat videos all day.

Goals and dreams

Imagine you’re twelve years old, you’re in school and you learn about the carbon cycle. You think, “Wow, I’ve never heard about that before”. When you get home, you go to YouTube and might type in ‘carbon cycle’, and if you’re lucky you’ll get a good search result and click on it. But what YouTube will now do on the right sidebar is start recommending the things that you normally do: gaming, sports, et cetera. Because to YouTube my goals and dreams just became statistics. For them it’s just about time. They just need people to watch more videos.
I think there’s a very important process here where we as consumers gently get persuaded back to our Freudian self. — Mads Holmen
It’s the lowest barrier to participation that these platforms are going for. But if we are trying to learn and expand our mind, we want the opposite. That’s where Bibblio comes in.

Subscription-based environments

The second you are in a subscription-based environment like Netflix it gets a little better. Now as a user you are paying for the service and the worst thing that could happen for that site or service is for you to stop subscribing. So the second we are in that environment — in our case education, learning and knowledge platforms — we really want people to be happy and satisfied with the service so that they keep subscribing. And that’s the core of what Bibblio is trying to achieve.

We are trying to make sure that users who come to platforms find the best possible content so they leave being satisfied and come back more often. And when they have to pull out the credit card again, they don’t cancel but carry on with that subscription.

Because we are worth it

Sometimes it is worth paying. The second you are in a paid environment the user is the customer again.
When we look across most of these advertising platforms…
Do you know what we are worth a year? In advertising revenue? For all the advertisers over the course of a whole year on the internet that target us with all sorts of crap? About twelve pounds. — Mads Holmen
Twelve pounds is the revenue all those sites make out of us in a year. If there was a model where you paid twelve pounds and never saw an ad online again, I think most of us would find that an interesting proposition. And what Bibblio is really trying to do — some would call it education — is make sure that people all across the internet find content that is better for them, and not just better for the advertisers.
Bibblio wants to help us be a little bit more our aspirational self and a little bit less our Freudian self. — Mads Holmen

Visitors bypass the homepage

One of the problems is that these days most visitors don’t come to the front page anymore. For a lot of the sites we work with, 80 to 90 % of their traffic lands directly on their content pages. People click through from social straight through to a content page, then read that article and leave again.

So historically we have been spending a lot of time on designing our front door, but the problem is people are coming through the windows, chimneys and basements. — Mads Holmen
If we don’t create a great experience on each one of those pages, if we don’t present the users with the best possible content for them at that moment, they will just leave again.

The medium is the message

This could also be dangerous for brands. There have been some studies, especially in news, showing that when people click through to a news article from Facebook they are very unlikely to remember what site they were on.
As an analogy, let’s do a quick show of hands. How many of you know that National Geographic is good at photography? [Most hands go up.] Now how many of you can name one National Geographic photographer? [Silence and giggles.] See, the medium is the message.
We all begin to remember Facebook more and more and the brands we visit less and less. I think that’s a real danger when you are a publisher or content platform. It’s dangerous to just rely on social platforms for your traffic. Because people tend to remember the platform, and not the destination.

What’s a good recommendation?

I think there are three factors to good recommendations. There is what we call relevance. This item is related to that other item based on the topic, or possibly it has the same author. We call that an item-to-item recommendation.

Then there’s behaviour. Some content may be more popular than other content.
I jokingly say that Facebook’s algorithm is like a giant contest where cute cats and Kim Kardashian always win. — Mads Holmen
We can look at overall popularity in itself, or we can look and say, “After this particular article this article seems to do really well or this article doesn’t do very well”. And we can begin to adjust the results based on that behaviour. Ultimately we want something like what you see on Amazon, ‘things that other users like you have liked’. This is called aggregated behaviour. Looking at users who have done similar things to you and recommending results that they’ve liked. This is also known as collaborative filtering or matrix factorization.
And third there’s personalization. When you log into Facebook now, Facebook has about 2,500 relevant items to choose from when they decide what is the first thing you should see in the News Feed. They use a lot of your personal information and what you’ve done in the past to personalize those results for you. Netflix does the same.

The value of content recommenders

Across these three buckets someone like Netflix deploys around 260 different algorithms to decide what you should see. They spent around $200m dollars a year on this one problem. So recommendation isn’t an easy thing. It combines how recent the item is, how relevant the item is, what this user would prefer, what they have done recently, and what they did a long time ago — all these factors are combined to make a decision on what should be the number one recommendation.

This is really a reflection of having now entered a world where there is too much content. We could never hope to consume even a fraction of the content that is online.
Publishing is changing from printing and distribution to filtering and curation. — Mads Holmen
That’s a whole new world for all of us. It’s a world where we need machine learning, we need algorithms to get smarter over time, and learn from the data we produce.
Finally, we need to make sure that we can deliver this as an experience to the user. I think my biggest learning from working with our clients has been in the digital environment of the future, we won’t be able to charge for just the content. When we as users subscribe to places like Netflix and Spotify, we subscribe for access to the content, but the reason we are happy to keep paying is because of the experience that it gives us. That’s the ultimate value that I as a user take away. I don’t think that in the future content alone will be able to supply that experience.
Thank you.

Many thanks go to The EdTech Exchange, Europe’s largest community of EdTech founders, for hosting Mads’s talk.


Turn your visitors into regulars

Bibblio is a content recommendation platform that helps content businesses and publishers deliver more relevant and engaging discovery experiences to their users. Visit us on Twitter, LinkedIn and Facebook.

More juicy posts by Bibblio:
How recommender systems make their suggestions
The future comes highly recommended
Why this story is not the most important — it’s the next one
Advertising vs Subscription
60 YouTube Channels that will make you smarter

A single golf clap? Or a long standing ovation?

By clapping more or less, you can signal to us which stories really stand out.