Real Time Data
Real time data is valuable in two ways, mainly. The first is that it crudely allows content providers to recognise spikes in popularity of pieces of content, so they can “double down” and mash on the proverbial “promote this content” button in order to drive engagement and uniques on their sites.
The second way in which real time data is valuable is in the context of historical data. A rich historical data set allows content providers to establish significant correlations between behaviours and outcomes. Websites can then use real time data to identify, segment, and serve consumers in real time, based on how their behaviour correlates to historical data.
Algorithmic Content Curation
There was consensus among the panel that strictly algorithmic content curation lacks the je ne sais quoi of human discretion. Computational editorial methods are necessary in order to deal with the massive scale of the Internet, but ultimately, human intuition and sensibility must supervise the editorial/curatorial process.
Complete Discussion Notes:
Discussion prompt 1: “Only human curation can deliver real time content that is relevant to people.”
Data in consumer space is noisy. There is a need for tools that filter and remix and use data in order to take informed actions.
The sheer scale of the web’s content and user base makes manual curation difficult. One needs computers to understand users and rationalize content strategy/decisions
To know if curation is working, one needs performance metrics. Metrics allow for pattern recognition and machine learning, which allow for automation.
Again, human curation is super difficult when dealing with something as massively scaled as the internet.
Dan Patterson, ABC News Radio: In his experience, computers help curators get really close to the content sweet spot, but there is a need for human discretion to truly optimise. Human/computer-curated radio out performs strictly computer-curated radio.
Machines with humans to override them when necessary – “humans vs. cyborgs”
Jon Steinberg: some editor offices look like trading floors, with six to seven editors looking at many data displays, looking to pick up on lower-level/subtle trends. They utilise algorithmic-based curation methods.
David Teten: Trend line is pointing toward automation.
Steinberg: it takes a computer to notice in an efficient manner when content in the background gets lots of (social) traffic. Example of the frowning flower girl @ the Royal Wedding. First, how do you recognise the genesis of such a trend as it forms? Then how do you best capitalise it, push it further?
You can’t curate until you aggregate.
There is an algorithmic limit to natural language processing. The promises of the technology have been slow to materialise, and some of the challenges are proving to be intractable.
Big data and automation help to define performance goals and match them to actual performance. Curation is an optimization problem – what gets deployed, in which order, in which rank? These are optimization problems.
Andrew Montalenti: It’s not a strictly binary argument. To get the best performance, it’s about how humans can leverage digital tools. Cyborg.
“Filter bubble.” Is all this automation creating personalised content echo chambers? Cf. Eli Pariser.
Sociocast: Predictive, real time data. They process 30m+ user events on a daily basis. All this data allows for discovery of significant correlations. behaviour of humans is redundant. They use their insights to behaviorally target ad serving.
Breadcrumb path of data. Calculate quick decisions to create better user experiences. Ability to experiment and discover what data is good to use, when to use it. All in this space of content provision and digital advertising are trying to optimise to create better experiences for people.
Scripting languages and data stores today did not exist even a few years ago. Technical feasibility of this domain is new and still expanding.
Search data is great for understanding and correlating user actions to user intents. But search doesn’t generate intent. Data collection and user tracking allows websites and 3rd parties to organise and classify users, content, ads, etc. and answer basic questions such as (1)how’d a user arrive at a page?; (2) what did they do while on the page?; (3) where and when did they leave the page/site? Data, metadata, etc.. reveal patterns. Once classifications and trends are understood and modelled, one can use real time data to recognise correlative behaviour and exploit it.
Different kinds of data -> real time, panel data and survey data.
David Brinker, The Daily: Different patterns of news consumption (e.g. some read in the morning, some in the evening). Within apps you can’t get all this data [my note: the issue of devices, particularly in the context of Big Data and behavioural, demographic, and psychographic segmentation, is a really interesting and underexplored topic, deserving of its own panel discussion!] For the Daily’s publication schedule, real time data is not very relevant. Furthermore, their advertising is brand-based, they want to create an intimate experience between themselves, their users, and their advertisers. It’s not direct response.
Dan Patterson, ABC News Radio: branding and relevant content delivery to older, ageing people -> still looking for an intimate connection. Advertisers want people to have intimate experience with ad, brand, product. Dan’s philosophy is that if you take care of your users, advertisers will follow.
Someone commented that most AM radio ads are direct response. Advertisers want consumers to call/act immediately
[My comment regarding David and Dan’s comments: every content provider and their mothers are trying to create these personalised, intimate experiences. iPad apps and radio are interesting in that they try to do so with limited access to data, particularly real time data. Neither panel members offered what kind of data they use to inform editorial decisions.]
Tony Haile, Chartbeat: Glenn Beck monitors two Chartbeat dashboards that display his website’s stats, as he presents his radio show. Dynamically curates his radio content in response to the real time web data.
Andrew, Parsely: almost 50% of news consumers get their news from the web. He believes there is no distinction between print and web audiences anymore [?]
Real time data allows editors to respond to immediate goings-on, not past goings on. Example of Nick Denton, identifying a spike and doubling down on that piece of content.
Feeding winners and starving losers. Throwing gasoline on a fire to turn it into a conflagration. Exploiting ephemeral opportunities to the maximum.
“Feed fatigue”. We live in a digital world of fleeting content relevance – if a publisher misses an opportunity, it could be gone forever. No second chances, can’t revisit an event or meme that was hot three days ago. Need to make sure that deployment happens at the right time.
The key to getting someone to do something is timing and context. Marketing 101. The value of real time, then, is that it enables one to recognise when a consumer is vulnerable to influence, as he is vulnerable to influence. recognise when someone is in your wheelhouse, and knowing what to do to optimally exploit him.
With respect to engagement, there are value tiers for different consumer news products. E.g., Twitter is temporal, noisy, and therefore discounted. Magazines and newspapers of record are highly curated, edited, and thoroughly substantive.
Lawyer: asked about FCC Do Not Track regulations.
It will harm consumers and businesses. The internet allows for co-created value. Consumers give information to companies, who use it to create more enjoyable, personalised products and services for consumers.
Someone else says that the legislation would mainly, negatively affect the collection of cross-session data, which is not very valuable or informative, anyway.
.Standard hot air and buzz words. “Data economy”.”the notion of privacy is changing”.”user opt-out”
.Discussion shifts to the topic of making analytics products actionable for their users:
User experience of data products is hugely important. Key performance indicators and other such visualisation/communication methods that make data accessible and digestible. Need to create affordances as to what to do with data.
Automation is the future. Make more money at less cost.
.Discussion shifts to macro-effects of shift to automation:
David: creative destruction and the rate thereof is dramatically disrupting legacy economies, labour markets, etc.
To close out the discussion, each panel member suggested one technology/trend that will die out soon. Answers included: 3rd party data, niche social networks, most of the tablets slated for release, dumb phones, haphazard social publishing, standard display ads for brands, lots of seed startups, AOL, traditional display networks, and irrelevant TV adds.