Advancements in Big Data & AI for Media
I’ve just come back from a trip to London to MC the Big Data & AI in Media awards and conference. I’m impressed with the incremental developments that have been made since last year. The overall theme seemed to focus on media organizations attracting and retaining core users/subscribers. Several presenters used either a sales & marketing funnel or pyramid to show loyalty pointing to the ‘tip’ of both as the core focus before monetizing other layers of the pyramid. Below it is shown at the Premium Subscriber.
There are a number of ways that publishers are working towards attracting those premium subscribers. This was well summed up by Gadi Lahav, Head of Product at the FT:
Upend the Enders pyramid and we have a standard sales and marketing funnel. Effectively the FT are looking at moving people down the funnel (or up the pyramid), recognizing that to do that users have to be engaged. Regularly. By analyzing data and recognizing cohorts of users (also referred to as clusters or tribes) it makes it easier to predict and therefore encourage or change behaviors.
Grzegorz Piechota, a research for Oxford and Harvard, ran a study where he subscribed to a number of news outlets, read regularly and then ‘snoozed’ to see what would happen. Only one of the outlets proactively contacted him with a series of messages to re-engage. This was the FT.
In discussions many noted that retention of subscribers was very much the next step. Reader engagement is key to this and there is a lot of efforts outside the publishers ecosystem to bring them in, focused around newsletters and push alerts. In fact engagement and retention are two of the metrics used widely within startups. One company, Bibblio, works with publishers as an internal content recommendation suite to keep readers engaged within the ecosystem.
Last year when I spoke to a senior representative at the New York Times they told me that they were wary of changing anything that may disrupt their loyal subscribers. Since then they have launched paid products for their crossword and cooking verticals, both of which appear to be doing well.
Once engagement and retention of those premium subscribers has been nailed, there will be more of a focus on monetizing the middle of the pyramid. The New York Times just announced that they are looking at a dynamic meter model. Scroll is looking at an alternative revenue to banner ads by asking readers to pay not to see ads. And there are a number of companies working on micropayments in various forms including LaterPay, SwissPay, Blendle and blockchain journalism pioneer Civil.
A few other trends that came through:
There is a build rather than buy mentality for Data & AI. usually at this conference a few tech focused companies get shout outs for their products and services that help media, but this year almost every company presenting had chosen to bulked their own layer of data analyses to meet their own specific requirements. In fact our hosts, Microsoft, use Bing News as it's testbed for AI products.
Data must be comparable, usable and relevant to all departments. As FT’s Gadi put it, publishers need to find their North Star metric. For then there is a (secret) combination of recency frequency and volume. However everyone measures different things so it’s up to each publisher to find their own super metric.
Whilst it was noted that display ads are not the future, there is a lot more focus around contextual ads by topic, placement & integration. This was also recognized in the BIGGIES awards.
Robots may be taking our jobs - well data scientists — as content selection gets more sophisticated but a human input is always needed and at least one editor for oversight. Automated journalism as presented by United Robots is becoming more popular. Further reading onglobally this topic can be found here.
In today’s divided political climate there is a lot of discussion around showing different perspectives of the story. Bing news, which often serves as a test bed for Microsoft’s AI services, is trialling this in beta and planning to rollout globally in coming months. It’s too early to tell if this actually broadens perspectives but I look foward to finding out.
There was a lot of discussion around content recommendations and now we’e starting to find it across additional mediums, most notably image & voice.
A last thing that resonated was from Mads Holmen of Bibblio who admires the use of tribes and clusters to look at behavior but also cautioned around consumption patterns. He told us that his favorite food is lasagne, but he doesn’t want to see it at breakfast time!
— —
If you’re interested in all things around Data, AI and Media the next European event is in Amsterdam in December, and the next global event is in Silicon Valley next May. I’ll update links on this page as and when there is more info. or you can follow me on Twitter or LinkedIn.
