6 Years and Trillions of Notifications Later, Here Are New Learnings

In my previous piece on learnings from meeting 300 global publishers, I wrote about the constantly changing rules of the game. That got me thinking about the way this will truly disrupt the digital newsrooms.

To set context, iZooto delivers 10 billion notifications every single day for over 1000+ publishers across the globe. An overwhelming 98 percent of these notifications are sent by digital newsrooms. Editors across large newsrooms, media houses, and small offices push out stories that matter to their users. Each professional and publication has a different recipe for user engagement. Some prefer getting all their work pushed to their users. Others believe in a minimalist yet targeted approach.

Spray and pray notifications don’t accomplish much

The fundamental need for a marketing channel such as push notifications focuses on building a new audience. Prevalent platforms such as search and social media offer anywhere between 25-75 percent of total visitors depending on the nature of consumption. While that seems like a good way to ensure steady growth, in reality, it’s not. The problem with a search and social source for user traffic is that over a period of time, the publisher is no longer in control of who their users are. 

As long as the system works, the publication is fine. But as is the rule of the game, the moment either the search engine or the social media platform changes its rules, there is a greater instability in traffic patterns. 

Going back to the trillions of notifications that we processed so far, here are some trends that stand out:

Crashing CTRs

Publishers - large and small - share notifications with their users. If you take away focus on relevance, CTRs drops from 3% to 0.30% especially for large publishers.

Leaky buckets

The fundamental problem plaguing publishers is that although they’re gaining new subscribers at a steady rate, churn is also a major concern.

Plummeting Performance of Notifications

This is where publishers are hitting a wall. Overall clicks and user traffic figures are not growing anymore.

In case you’ve been a fencesitter, data reveals that pushing all the content you create to everyone doesn’t yield results. This also opens up new learning that algorithms can make better decisions. 

The rise of algorithms

Google Discover powers anywhere between 30-70% of publisher traffic. The click-through rate observed on the platform is in excess of 5 percent. It’s not a coincidence that Facebook has decided to go the TikTok route by populating the Facebook Newsfeed using an algorithmic recommendation engine.

While making the announcement, Meta CEO Mark Zuckerberg said, "One of the most requested features for Facebook is to make sure people don't miss friends' posts. So today we're launching a Feeds tab where you can see posts from your friends, groups, Pages, and more separately in chronological order. The app will still open to a personalized feed on the Home tab, where our discovery engine will recommend the content we think you'll care most about. But the Feeds tab will give you a way to customize and control your experience further."

Our own experience with content recommendation has been phenomenal. Compared to standard push notifications, our content recommendation engine powered by AI results in 6X to 10X higher click-through rates when compared to an editorial push.

Greater efficiency with AI 

The numbers seen on Google Discover, the swift move by Facebook to go the TikTok route, and our own observation with AI-led recommendation engines indicate a clear future for the media industry. The role of human intervention will diminish in determining the reach of content. 

Algorithm-led decision-making will enter newsrooms and change the way news is created, distributed, and consumed. 

There are several tools that promise AI-driven content creation. But they’re just vain promises. In reality, there’s a huge gap to cover before they even get worth considering. Google and Facebook both deprioritize machine-generated content, and if things continue this way, then AI can at most help enhance relevant content, but not replace human editors.

Clearly, AI can take over content distribution which is a different ball game altogether. In the future AI will decide what, how, when, and where you'll consume content, not humans. This is already happening in parts but I expect this to mature over the next few years.

Also Read: Here's How Brands Will Compete In The Future Of Digital Publishing

At the pace with which the media industry is evolving, algorithms will not only influence but actively shape how publishers create and distribute content. In fact, algorithms and automation will be so popular that it could mean the difference between being on the winning or losing side as a publisher. In terms of scale, the larger publishers with a reach of millions of users are expected to have a distinct advantage over the relatively smaller ones.

The existing Inside Out process for Content Recommendation 

For years, publishers have considered recommendations as an inside-out process. Typically, they would display relevant stories on the home page that visitors would discover during each site visit. Based on user browsing patterns, the website would define relevance using algorithms that consider the past behavior of the user on the website. 

The data gathered then displays popular content on a widget on the sidebar of the website. For most publishers, this has yielded mixed results with limited success.

Opt for real-time tracking over past data

Although looking at past browsing patterns by users does seem to make sense, the reality is different when you look at data. Most publishers see anywhere between 50 to 80 percent of new users every month. These are users without any recent history on the website and that means there’s no data for algorithms to drive content recommendations. 

Since this is the dominant user profile for most news websites, looking at past browsing behavior hasn’t yielded any results. Even if you consider additional metrics, publishers see user retention rates of anywhere between 1 to 5 percent over a 12-week period. Once again, that’s not enough data for AI to kick in.

News publishers work with relatively higher publishing volumes. It is common to have over 300 stories filed every day for active publishers and based on size, there could be over 100 content writers feeding stories to the newsroom. Each one of them is creating content on a topic that is relevant to a specific audience profile. The decision to approve these topics depends on human intervention. Since this is still a subjective process, it is still influenced by user behavior as seen for the same publication instead of the larger online environment that includes competing websites, Google Discover, social media, and other platforms. This kind of inside-out process is obsolete.

Also Read: Customer Data Platform: News And Media Publishers Have An Opportunity Before Them

The answer to this problem is to take an outside-in approach to content creation and distribution. As users spend countless hours every day consuming content on the internet, the gold mine of data is sitting outside the publisher ecosystem in walled gardens and with other publishers. Farming this data to bring relevant insight and enable decision making (either manual or automated) would be the differentiator in the future

Personalize experiences for the audience

In the data age, users expect personalized website experiences. Content recommendation is no longer limited to small widgets. In fact, they are getting increasingly prominent and are taking up more real estate on web pages and apps.

There is a huge opportunity for publishers to study user behavior and consumption patterns, and offer them tailored recommendations. I hope to see massive changes in algorithmic platforms to help keep digital media newsrooms relevant and ready for the future.