NZME

How To Boost Customer Engagement With A Content Recommendation Solution

Objective

To boost audience engagement and re-circulate content to recommend the ‘next best article’ for New Zealand Media and Entertainment (NZME) readers.

Solution

Datalicious created a bespoke content recommendation solution for NZME built on the Google Cloud platform (GCP).

Result

NZME now has a flexible and fast mechanism to recommend the best content to users and increase reader satisfaction and revenue.

the background

Date: November 2020

NZME has an audience of 3.4 million across print, digital, radio and video platforms. Their network of over 50 media brands includes the well-known outlets, The NZ Herald and Newstalk ZB.

NZME wanted to improve their content recommendations on the NZ Herald website by hoping to give its readers a more personalised experience and achieve greater control over the content that was being recommended.

The website previously depended on an automated content recommendation engine (CRE) for audience engagement and recirculation. However, NZME had very limited control over their existing CRE and wanted to increase the relevance of their content recommendations to better target their audience.

They required a bespoke solution that gave them full visibility and control over their content recommendations. NZME knew that suggesting truly personalised and pertinent content would create a better user experience and enhance audience engagement, thus keeping readers on the website for longer periods of time.

the objectives

Gain greater control over content recommendations.

Gain greater control over content recommendations.
Gain greater control over content recommendations.

Gain greater control over content recommendations.

Boost audience engagement by re-circulating content to recommend the “next best article” and enhance user experience.

Boost audience engagement by re-circulating content to recommend the “next best article” and enhance user experience.
Boost audience engagement by re-circulating content to recommend the “next best article” and enhance user experience.

Boost audience engagement by re-circulating content to recommend the “next best article” and enhance user experience.

Grow paid subscriptions and increase direct ad revenue.

Grow paid subscriptions and increase direct ad revenue.
Grow paid subscriptions and increase direct ad revenue.

Grow paid subscriptions and increase direct ad revenue.

the solution

As part of the Google News Initiative (GNI) Data Lab Program, Datalicious was the first technologist to develop a bespoke CRE algorithm that recommends ‘the next best article’ based on content similarity and a user’s historical behaviour for a publisher.

We created a CRE and API system for NZME built on the GCP. It uses machine learning, artificial intelligence (AI) and Natural Language Processing (NLP) to recirculate and recommend relevant articles to NZME users. This enables them to read more content, spend more time on the NZ Herald website and nurture readership loyalty.

In addition to providing suggestions on relevant topics, the algorithms are programmed to recommend articles to readers based on previously consumed content and topics relevant to the reader’s location. The articles are also classified into respective categories like sports, entertainment or politics that are shown to the reader.

Thousands of historical articles were run through Google’s Natural Language Processing API to train the algorithms. Keywords and phrases were extracted that made it possible for the intent of the material to be understood. Google Analytics 360 (GA360) was also used to develop an understanding of the topics that engage individual NZME readers. Over several weeks, this behavioural data is extracted from GA360 and the articles are scored based on reading signals and user behaviour. This data is then continuously refined to keep up to date with changes made to content preferences. 

As NZME also wanted to recommend content based on location, Datalicious included a location reference database as part of the custom solution. Google’s NLP tool helps the algorithms recognise when an article contains a reference about a suburb, town or place, and then ranks these according to how often they are mentioned.

These location references are matched to their local region using a custom mapping table. When an article is specific to a given location, the CRE can classify it according to the part of New Zealand (NZ) it refers to. Now, NZME readers get content recommendations that are also relevant to the area they live in, no matter what NZ Herald article they happen to be reading at the time.

Over time, these AI-fuelled recommendations become increasingly relevant as the algorithms operate on a continual feedback loop, learning more and more about the type of content the reader prefers.

the results

NZME has achieved incredible results with the new CRE built on the Google Cloud Platform, including:

  • Rapid roll-out of several algorithm advancements that were optimised to match NZME’s existing CRE and market leader within just 9 weeks.
  • Content recommendations delivered within 500 milliseconds (ms) from the moment a reader arrives at an article.
  • 3X improvement in recommending content relevant to the reader’s location.
  • Plus, it contributed to an impressive 23% increase in website visitors for a targeted audience.

The new CRE continues to make a difference by providing:

  • A better user experience and audience engagement as NZME readers receive relevant and more personalised content recommendations.
  • Greater control over ad revenue from paid placements.
  • The ongoing ability for NZME to customise content recommendations based on their changing business, brand and audience needs.
  • Optimal engagement and traffic retention on NZME branded websites.
  • Full visibility of content performance across NZME platforms and channels, with easy access to real-time and historical analysis.

NZME now has a flexible and fast mechanism to recommend the best content to its users and increase reader satisfaction and revenue. This achievement was vital to the success of the project and gave NZME the confidence to scale up the new CRE with a road map of further enhancements to come in 2021 and beyond.

Find the solution to your challenge

The NZ Herald is focused on keeping Kiwis in the know and now delivering the “next best” article recommendations with our purpose-built engine, has resulted in a more relevant and engaging experience for our audiences than the previous off-the-shelf product. It was a huge collective effort between the Google News Initiative team, Datalicious and NZME to create the engine which sits at the frontier of technology, by utilising AI, machine learning and natural language processing tools. The results have been exceptional for us, unlocking deep insights on article affinity and increasing recirculation rates for audiences. We are looking forward to working with our partners to build on our success, with future iterations of the recommendation engine.

Andy Wylie - Head of Data & Analytics, NZME - 2020

Get in touch
with us

Enter your details in this form to request a call back from one of our specialists.