Håvard Kvalheim, head of big data at Norway's third largest media group Polaris Media, explains how homepage personalization increased time spent on site by 20% and the number of unique users by 25%.
At Polaris Media, we’ve created a personalization strategy for many of our brands and introduced fully-automated personalization to the iTromsø mobile news site.
If our news sites fail to become a destination for web users, they will become marginalized so we have to keep up the direct traffic to our homepages. Personalization gives us the opportunity to offer the users a better experience and something that is more relevant.
We started out personalization on parts of our pages, with tailored recommendations below articles or at the side and we saw that this strategy increased time spent reading on the sites by 20%.
Personalizing the entire homepage
We built on this and went on to personalize the entire mobile homepage of Norwegian news site iTromsø, which was the first in Europe where everything was personalized. Traffic grew by 13% and unique users grew 25% in three months while average time spent on the front page went up from 22 seconds to 28 seconds.
An option to turn off personalization was prominently displayed on the site, but less than one per cent of visitors chose to do so.
These days, people expect a personalized experience. They are used to it on Facebook and Google search and it is a trend we wanted to experiment with as well. The best thing is that as soon as we started testing it we had an uplift in readership.
Building user profiles
We build an individual profile for each user, these are anonymous user profiles of what they have done and consumed on our pages. And then we also create a content profile for each article.
These are matched and then we set the criteria as to what we want to recommend. What we experience is that below an article what typically works best is contextual recommendations that relate to the article the user has just read. For other widgets, we use a different set of rules such as what is trending right now or what other people with similar interests as you read.
A personalization engine to drive engagement
For full personalization, we use a multitude of rules. We also include a minimum of curated content from each section so we don’t get a filter bubble where people become estranged from content that doesn’t match their interests.
It is important to give the user some variety and we also include something to make sure that the most important content is always available.
We use a personalization engine based on the Cxense data platform, and make use of two or three different tools. One is the insight and analytics tools that give us information about what users are doing on our pages.
The power of a DMP
We also use a data management platform to create different segments which have a different set of rules or recommendations, then an option which combines the data from the users and the content profiles. We recognise visitors’ cookies so we can recognise their device.
We post information on our pages about how this technology works and we also have an opt-out option so they can choose not to get recommendations. It is based on anonymous information which is cookie-based and users can opt out if they want.
Homepage personalization is the future
The technology has generated a lot of traffic across our brands, so it has created more value in terms of potential ad space than the cost of the technology, in terms of investment. We haven’t invested that heavily in development, we have one person who is working full time doing personalization, working with the editorial team to set up the rules then implementing it on the pages.
I don’t think the old model where the entire front page is curated manually is the future. It is too time-consuming and generally only the top third of stories are important enough to warrant manual curation. We need to find a balance. But homepage personalization is certainly the future.