“Our data strategy, which has been in the making for two years, involves capabilities such as customer relationship management, preference management and the things that help us move from a mass-scale audience format to an individualised scale,” Ng says.

The way Stuff thinks about its audiences is changing, Ng says, and one of the big projects he and his team are working on is in part preparation for the ‘death of the Cookie’, following announcements from major software players like Apple and Google around their plans to disable third-party cookies that track user habits across the web.

Preparing For The Death of Cookies

“Over the past couple of years, what I’ve been trying to do is streamline our data channels. We are in the middle of creating a customer data platform,” Ng says.

“We are looking into solutions to manage customer data and make sure we use it in accordance with the reason it was collected, as well as give control back to users. The natural way of progressing to that is to have the user provide a persistent ID, like a login.”

Ng references innovations occurring in Europe, where the EU has unveiled plans to roll out a digital ID called a ‘wallet’ for use online. This will allow EU residents to retain control of their own ID rather than continually creating new IDs for websites and sharing those with tech giants.

“Once you have a persistent ID across platforms it changes the way content publishers like us would operate. Stuff.co.nz for example is a mass-audience platform, unlike say, Netflix, which is purely a subscription model,” Ng says.

“The pressure to adopt one-on-one relationships with each audience member, moving from a mass audience perspective to knowing and managing individuals poses a huge challenge.”

The media consumption model has changed considerably in the past decade. Subscription and paywall access to general news offered digitally has become more and more common on major news brands. These require users to log in and become known audience members of a given brand.

Ng says that while Stuff is dedicated to remaining a free news platform, the subscription or login model is likely to grow considerably across industry as privacy regulations change.

“Stuff’s direction is to remain free, but there is huge value in the data that customers will choose to share with us in exchange for content delivery,” he says.

“Our goal is to be the platform that understands New Zealanders the most. We see helping customers through that lens as our aspiration.”

Gaining a more granular understanding of audiences provides publications with a stronger platform to offer advertisers, who could target consumers based on traits and preferences, and expands the horizons for how audiences can be served content.

Understanding Users

Psychographics go beyond traditional demographic classification by studying consumers based on their activities, interests and opinions.

“Traditionally with demographics we can get broad classifications of users when they answer certain questions or declare things about themselves in a profile,” Ng says.

“With psychographics, we can potentially use something like comment behaviour to understand interests and sentiment. For example, some users will comment exclusively on property and real estate posts, discussing the market. Machine learning can be used to pick up on this and build something of a user category.

“An article is one input component of a machine learning algorithm, and the comment is another. You can use a general adversarial network (GAN) to model psychographic traits, and when you produce new articles, you are able to test what a likely response might be.”

Considered Approach

While excited by how new audience modelling may improve the site experience for users, Ng says it needs to be done carefully and ethically, so that editorial decision-making remains respected and that news services do not become feedback loops.

“There’s always those ethical considerations and the question of creating an echo chamber. They are valid concerns that we all have,” Ng says

“This is one of the main reasons we have not gone full blown on this stuff. We’ve made certain algorithms explicit so the editorial team can determine how much of the homepage is generated based on modelling as opposed to manual selection.

“We are also working with some champions in the editorial team, and I’ve asked my team to do research and to learn to think more like journalists so editorial and technical teams can meet each other halfway.

“This has allowed us to progress with introducing new digital content features carefully and in a controlled way that everyone feels comfortable with. We’ve done a lot of work on this over the past two years, and we’re excited to see it progressing.”

CDAO New Zealand  Online will be held from Tuesday 16^th to Wednesday 17^th November. Download the agenda and claim your FREE place today!