5 things I’ve learnt about the news industry while running a journalism startup
A strong press is a vital part of a healthy society, but the news industry is struggling. Only a few news organisations have worked out how to offer content for free, pay journalists and make a profit.
Although that’s scary, it’s also exciting. The future of news is up for grabs — we just have to find the right models to make it sustainable.
I’m a journalist, and I set up The News Hub because I believe in the future of news. We launched over a year ago, and in that time I’ve learnt a lot about the industry.
1) There are massive categories of content that are under-served by traditional news organisations
To understand this first point, I need to explain how The News Hub functions. Whereas a publication focuses on a particular type of content for a certain audience (think of The Times), our approach is closer to Medium’s — our users can read and write what they want across news, life & sport. Unlike a publication, which chooses a specific editorial direction (such as UK-centric, right-leaning current affairs), The News Hub is an open platform for news and opinion. We host a much broader content landscape than traditional publishers — and we can therefore observe where demand for a particular type of content is not met by supply.
For example, when a contributor started writing about animé, the Japanese style of animation, we thought it would be pretty niche. We were wrong. Soon after the contributor began posting, he was driving significant levels of traffic, chiefly through Google.
It shouldn’t have surprised us. Animé is loved around the world, but there are few, if any, English-language news organisations covering it. As a result, when fans search for their favourite characters or shows, they often click through to The News Hub because we are indexed on Google News and offer animé content when traditional news organisations don’t. We’ve learnt that the same principle applies to other categories, such as video games. Both are hugely popular subjects (the video game industry is bigger than Hollywood), yet they are under-served by the mainstream press.
2) Quality content still wins out
For years, industry commentators have suggested that online content is a “race to the bottom” — publishers focus on low-brow content because it’s more popular than high-brow content, delivering the most views and therefore the most advertising dollars. But that’s not what we’ve found.
Being an open platform, we have a much wider range of quality than you would find on a traditional news publication. We can see how articles of different levels of quality perform, and contrary to industry headlines, our most popular articles have been fairly high quality. Seriously.
3) Native advertising networks may underpin business models of the future, but not the present
In this context, native advertising refers to sponsored content — brands or agencies paying to get press coverage.
First, let me stress that I believe that native advertising is the most likely saviour of the news industry. It’s how I think news organisations will deliver advertising that readers don’t think is annoying. I believe that in-house native is how publishers will get around adblockers in the long-term. It’s also better for brands which can use native to tell more powerful stories than ever before. If you’re interested, I expand on these points in the video below.
However, our experience with native advertising networks has been a huge disappointment. For anyone unfamiliar, a native advertising network injects sponsored content into a particular placement, fitting naturally within its environment. You can see an example from our Trending feed below.
When we launched The News Hub, we were excited by native advertising networks. They’d told us to expect CPMs (cost per mille, or cost per thousand views) of around £5 per placement and those placements to be filled 30% — 60% of the time (known as fill rate). We thought that if we had a few placements, and if we established a waterfall to serve ads from different networks on each placement to maximise fill rate, we could monetise relatively effectively through native advertising networks alone.
We were wrong. We have generally encountered poor customer service, immature technology (which has made integration a headache), and disappointing fill rates. For the moment, we are better off monetising through display networks rather than native networks. And that’s a big shame.
4) Readers want curation, not personalisation
A number of news organisations including the BBC and the Guardian have put personalisation at the centre of their applications. We’ve taken a similar approach at The News Hub — allowing our users to shape their news experience by following their favourite journalists and selecting their favourite categories, such as US Politics but not Golf. However, our data suggests that readers still favour curation over personalisation.
Fewer than 15% of our users have personalised their reading experience, preferring instead to be served every type of content and then pick the stories they want to click on. In other words, personalisation is still secondary to intelligent curation.
5) You *could* make a profitable journalism business from Taboola alone
I want to emphasise that this is just a theory but I think it’s at least plausible that you could “buy” traffic from Taboola at a lower rate than you could monetise that traffic. You would therefore earn a margin between the costs of the traffic and the revenues generated off the back of it.
For anyone who doesn’t know, Taboola is a content marketing service that drives traffic to a particular web page, for a fee. You’ve probably seen it in action, or Outbrain, its main competitor — I’ve attached a screenshot of the latter’s widget on The Telegraph, below. The host of the widget, in this case the Telegraph, receives a fee whenever a reader clicks through to one of the articles being advertised. Taboola takes a commission on each click, while the “traffic-buyer” pays the total.
Taboola charges roughly £0.10 — £0.20 per click, which means traffic buyers pay £0.10 — £0.20 whenever they generate a view directly from Taboola. You would therefore expect 1,000 views to cost £100 — £200, but our Taboola campaigns have cost us as little as £1.36 for every 1,000 views (although not strictly true, let’s call this CPM, or cost per mille, for simplicity). This is where it gets interesting.
When you buy a visitor from Taboola, that visitor may ultimately generate significantly more traffic than the one view you’ve paid for by either (i) clicking on more articles, and/or (ii) sharing articles with their social media followers which attracts more visitors, who may also share the articles themselves. You pay for a certain number of visitors directly from Taboola, but those visitors can kick-start a chain reaction which drives considerably more views than you’ve paid for.
If it costs £1.36 to generate 1,000 views and you make more than £1.36 in revenue per 1,000 views (otherwise known as RPM, or revenue per mille) then you’re clearly making a profit on that campaign. But the problem is that Taboola traffic is extremely volatile.
Although we’ve been able to buy traffic from Taboola for as little as £1.36 for 1,000 views, our average CPM from Taboola across our experiments is £10.31 — and that includes campaigns with CPMs as low as £1.36. Taboola’s performance is variable because you don’t control the chain reaction which occurs when Taboola visitors start sharing articles.
But let’s imagine you optimise your Taboola campaigns, reducing the average CPM to £5 — £7. And let’s imagine that all of your advertising placements combined deliver an RPM greater than your marketing costs (£5 — £7), which is plausible. You would have built a profitable news business through Taboola alone.
Instead of optimising your content for social media, you would optimise your content for Taboola (or a similar service), market it solely through Taboola, and make a profit off the back of Taboola — profits which could support all kinds of journalistic endeavours.
I’m clearly not the first person to think of this so I’m interested to hear what others are doing in this space. Is anyone making this model work? Or, is it simply not possible?