After months of pouring over data and great interviews with newspaper and magazine publishers, Pugpig has published its annual flagship report that looks at the state of media apps. It was my final big project there before I joined WAN-IFRA.
As I worked on these reports for Pugpig over the last three years, we noticed a few things:
Performance by various measures varied widely across the hundreds of apps Pugpig develops, and this is consistent with other reports such as ones from Airship and RevenueCat. We had reported these ranges, but this year, we started using box and whisker graphs, as RevenueCat did in their report about subscription reports.
In some ways, no two apps are the same, in terms of features but also in terms of the revenue and engagement strategies that publishers used.
Based on the report we did earlier this year on push notifications, I found the most valuable insights were found in identifying the positive outliers and speaking with them to find out what tactics drove their success.
In the push report we conducted with the support of prominent push providers including Airship, PushPushGo and Pushly, we found that those publishers who were most successful in terms of open rates but also other in engaging audiences after they opened the push notification used tactics to increase the relevance of those notifications.
ACCA Student Accountant surveyed their audience to find out which users were taking exams so would be most interested in push notifications about exam content.
City AM has their newsletter editor manage their push notifications and uses the links that newsletter readers click on most to inform the stories they push to their app users. They had some of the highest open rates of any publisher.
The Boston Globe added a preference centre that allowed users to choose the push notifications they were most interested in. It increased open rates up to three times, and they had some of the highest performance in terms of the number of stories users read after opening a push notification.
We used this model to identify publishers who had outsized success with various app features such as audio, video and puzzles. Foreign Affairs has driven incredible performance with in-app audio. The Baltimore Banner added a vertical video carousel that improved the in-app video engagement. And Stylist demonstrated how adding puzzles to their app could dramatically improve performance.
Foreign Affairs has a brilliant product team. Tasia Fischer, who manages product operations at Foreign Affairs, has surveyed FA’s audience, and both surveys and data show that their audience cannot get enough of audio. The report outlines how through smart use of technology and simple UX tweaks they increased in-app audio listening by 72% year-over-year in January.
The Baltimore Banner leveraged the work of two video producers on their 55-member editorial team for the vertical video carousel in their app. “The other thing that is encouraging is that (vertical video) engagement numbers are going up every month. People are starting to build habits around video in the app, said Eric Ulken, the vice president of product at The Banner.
Stylist shows how puzzles increase engagement not only with puzzles but with their content. Puzzle users read 31% more articles per session and 69% more articles a week. This higher engagement translates into puzzle users having 85% higher 30-day retention rate versus the average users, according to Emma Peagam, the product lead at Hearst UK.
That’s just some of the insights in the report, and it’s well worth downloading.
What I’m reading: FT uses AI to engage commenters, ‘Mythbusting LLMs’ and preserving pre-AI content
Hannah Sarney, the FT’s editorial product director and executive editor, talked about how they were using AI to encourage readers to comment on the articles at WAN-IFRA’s Congress in April. They have found that commenters are four times more engaged than other readers, but commenters also make up a small part of the audience. To encourage more people to comment, they developed a tool that leverages generative AI to create three questions that editors can choose and embed midway through the article. The results are modest in terms of increased comments, but the real results seems to be better comments. It’s a great application of genAI.
LION Pubs has leaned into data and a focus on sustainability under Chris Krewson’s leadership, and they keep building on that success by increasing the amount of data they are using to support their members assess their businesses.
This is an incredible resource by Joseph Lochlann Smith, the “tech lead of the Guardian’s fledgling Newsroom AI team”. Parts of it are accessible to a lay person, but he doesn’t shy away from the complex probabilistic calculations involved. Mythbusting is needed right now with LLMs, and he lays out really good arguments that bust common misconceptions about LLMs and generative AI.
This was an interesting move. Google isn’t being forced to offload Chrome yet as part of an anti-competition/antitrust ruling, but it might be (although I can’t see this happening under a Trump Department of Justice). Grab some popcorn though because all of this will get very interesting.
As others have said, Zuck is hungry for success after his metaverse play didn’t catch fire as he hoped it would. With all of the content being posted on Meta’s social networks, they certainly have a lot of content to train their models, but it still stands to be seen if they can win in the AI battle against newcomers and old stalwarts.
A former Cloudflare executive is looking to create an archive of content before AI. The idea is to have a source of content that is guaranteed to have been created by human beings. The article not only explains the rationale behind the initiative but explains concerns about model collapse, a fear that AI models deteriorate if they were trained purely on synthetic content.
