How artificial intelligence could revolutionise publishing

How artificial intelligence could revolutionise publishing

A personalised web experience for every internet user, based on their behaviour, is not far off, writes Ohad Tzur. Here is what publishers need to know.

While a Westworld-ish future still appears to be some years down the road, the artificial Intelligence era has already begun. Artificial intelligence is perhaps an overused buzzword, but is becoming a key trend in various industries. New A.I. technologies are being developed that will soon impact the online publishing world, becoming an integral part of the digital ecosystem.

Publishers can proactively prepare for the upcoming paradigm shift by better understanding the role of this technology in the future.

First, let’s review the terminology. Is a complex set of algorithms considered to be A.I.? No, it isn’t. In a nutshell, A.I. is a term used to describe how machines use computational capabilities to ‘think’ like humans. A sub-branch of this broad definition is machine learning, referring to the machine’s ability to learn on its own without being instructed specifically what actions to take.

Amongst several categories within the machine learning realm, one of the most common techniques is called deep learning. These algorithms are structured to function similar to the human brain, using neural networks.

You’re probably asking yourself why is this at all relevant or important. Well, artificial intelligence is set to become the core that connects automation and big data analysis.

Despite the increasingly complex ecosystem of signals and levers, publishers will be able to solve problems which were previously impossible to overcome, for instance, applying machine learning to data to enrich experiences through personalisation.

New era, different mindset

Bringing the excitement of artificial intelligence to any business will likely require integrating new technologies into the stack. If you’re not yet ready for this technology to take over, you can start by developing better understanding how it will actually work:

Your goal: Better user experience, higher user value
This isn’t surprising news – Data science shows that user experience and ad revenue are undoubtedly tied. Publishers need to apply a methodology to balance the two sets of metrics, hence the approach should be increasing user value (and overall earnings) in a sustainable way while monitoring and improving UX metrics.

One way to go about this is to run a set of A/B tests to generate insights. Since each variable of the test may (and will) affect other variables, transitioning to multivariate testing is likely to yield deeper conclusions.

The data presented in the chart above represents indexed revenue performance as a function of users’ time on a given page. If a publisher’s monetisation strategy relies on ad revenue, it would be critical to find the optimal layout to keep users engaged with the site’s content.

Session Value instead of Impression Value

While CPM is a widely used metric in the digital world, enabling buy and sell side transactions, perhaps it’s not an optimal metric for publishers.

Why? Because CPM does not factor in user experience metrics (more specifically, session length or page views per visit) and their overall impact on total revenue.

Publishers may have a better option – optimising towards EPMV (earnings per thousand visits). This allows publishers to assess the entire session value instead of the value of an ad slot or a single page.

EPMV as a goal also accounts for the value that subsequent page views have on the total value of each visit. If you’re already connecting this to what I’ve relayed above, session length is tied to the number of impressions on each page, so the ad combination for each user should be optimised on the session level.

If you’d like to better understand this, I’d recommend reading this article on EPMV, written by Tyler Bishop.

Transition from “the average user” to “the user”

Consider the following scenario – a publisher runs a layout experiment to test two different navigation bar settings on her mobile site. Layout A, which displays the navigation bar at the top of the page, is the preferred option with 70% of users demonstrating better engagement and higher ad performance value.

However, Layout B, which displays the navigation bar at the side of the page is the preferred option for the other 30% of users. Most publishers would make the decision to show Layout A to 100% of their users, because it’s favoured by the majority of users. Is this the optimal configuration?

Absolutely not. The better option would be to show Layout A to the 70% of users who prefer it, and show Layout B to the 30% of users who prefer it. The holy grail here is to create a personalised experience for each one of the users.

The example described above examines just one variable; imagine what would happen if we were to tailor every element to the individual reader. I’ll refer to it as the upcoming revolution of the online world: Better engagement and higher value to both publishers and advertisers.

As we all follow – and help create – the continuously changing future of the digital landscape, there are opportunities for us to rethink our approaches and methodologies.

What have we done, what have we learned, and what should we change moving forward? A personalised web experience for every internet user, based on their behaviour, is not far off. Ultimately, artificial intelligence will be able to deliver this and much more…

Ohad Tzur is VP of global partnerships at Ezoic.

Media Jobs