Hi everyone. This week, I’m going to discuss two topics specific to personalisation, looking in particular at how far it can be taken to apply to more areas of a publisher site. Can a whole homepage be entirely personalised while the notion of news judgment and an institution’s voice still exist?
And … I’d love to hear from you. E-mail me at email@example.com to say hi and tell me how the INMA Smart Data Initiative could be useful to you: from programming to speakers, to topics that are on your mind and you want us to take on.
P.S. Stay tuned for my next newsletter, in which I’ll continue my exploration into the topic of personalisation, this time looking at some of the upsides and downsides of certain methods for passive personalisation.
One of the three tentpole topics of the Smart Data Initiative is personalisation — it’s also one of the modules for our master class next month. So I’ve been doing various interviews with publishers on the topic. A couple of weeks ago, one of them said to me (I paraphrase): “We’re not going to personalise everything because the newsroom knows better than the algorithm.”
This gets at the heart of one of the faultlines for how publishers have or haven’t embraced personalisation.
There are areas where personalisation has unambiguously been present for quite some time: Our ads are usually, de-facto, personalised (née ad targeting). And for many publishers, Taboola or Outbrain are handling their recirculation area’s editorial links with a personalised organic selection — in addition to the widget’s role in presenting commercial content recommendations (Disclaimer: Yours truly used to work for Taboola).
There have also long been some products based on active personalisation. “My FT” from the Financial Times is one such product, where a user builds their own custom feed of content from the FT. The aggregator app Flipboard introduced some personalised feeds features with some well-refined filtering for a user’s niche interests.
But this is all done on an opt-in basis. And there is not a lot of agony over editorial personalisation as long as it’s a supplemental experience. These customised feeds have long been the province of small user groups, which are often very active and loyal. But because of the hurdle of requiring a user to put up with upfront “costs” of configuration, the real uplift of personalisation is in so-called “passive personalisation.” And one where the default experience is personalised — not a secondary experience that only a subset of users would know to use.
So we come to this unease with a by-default experience of passively personalised editorial content selection: Personalisation seems to negate the notion of an editorial voice and editorial judgment. After all, the very notion of a judgment implies the taking of a position. If everyone sees and hears something different from a news organisation, what is its true voice? Or, put another way, in personalisation, where does editorial news judgment go?
Personalisation and editorial judgment aren’t antithetical (an interesting academic study from Digital Journalism in 2019 cross examines these feelings among European news publishers). But there are some aspects of how personalisation is often done that, indeed, set personalisation outcomes to reflect trends, or deep vertical interests, more than editorial notions of importance, diversity, or quality.
To have personalisation algorithms that can reflect such notions of editorial quality, we have to think differently about the type of inputs that will go into these personalisation algorithms.
Goals and rules of personalisation
First, let’s remember that a goal of a personalisation is always going to be one of two things:
An upside of a measurable event — whether that’s the increase of a direct metric or the increase of an audience behavior (active days, likelihood to subscribe at N days, etc).
To limit the scope of a negative measurable event. (It sounds like this is stating the same thing the other way around, but it’s isn’t!)
For a personalisation algorithm to work, it has to have inputs — rules (the algorithm) and an outcome we can measure to see whether personalisation “is working.” That’s the loop.
Where we’re lucky is that choosing a measurable event to service in personalisation can often be a proxy for several of the organisation’s goals. Trying to apply a recommendation algorithm to content and looking at CTR (for example) is useful and valuable because CTR can be:
Directly making money (for example, the thing being clicked is a performance-based ad unit).
Making money because there’s an extra pageview and this page has ads on the page.
Making money because as the user’s session grows longer, she is habituating on our pages and is more likely to turn into a fan and later a subscriber.
So CTR in this respect is a proxy for both user satisfaction and for money. (Anecdotally, a publisher in India told me that their new foray into personalisation is doubling CTR from previously unpersonalised spaces. The upsides of this type of change won’t linearly grow like this forever, but what changes can you make to your product that yields that kind of upside right at the gate?)
It seems therefore that editorial personalisation shouldn’t be so contentious. Journalists too want their readers to engage further with what they are producing. Even if we look to something so crass as CTR, that metric correlates with a number of things that the organisation’s social mission as a news organisation probably wants to support: more engagement with its overall product.
Further afield on the wide, wide Web
Here is one good read from the wider world of data. This week: You’ve made it this far in this newsletter, so you deserve a reward. Here’s a data scientist who took on one of the important questions of our time: Finding the best Wordle opener, this time, using unsupervised learning.
Dates to remember
The Smart Data Initiative’s first master class for 2022, Transforming What We Build Using Data, is next month. See you there!
Meet the community
For each installment of this newsletter, I am hoping to introduce one member of the community in this space. Can we introduce you? Could you answer a few questions to get to know you better?
As our first introduction, meet Guilianna Carranza, chief data officer at Grupo El Comercio. Here is some information on her:
- Suggested follows: @PepeCerezo @davearon @MasonData @searchliaison @ecdata_pe @Retargetly.
- What topics are on your mind these days and what should people connect with you for? First-party data strategy, data compliance, automated content (we’re working on it!).
- Where should people connect with you? LinkedIn and Twitter.
- Are you hiring and want to promote a job posting: Coming soon!
About this newsletter
Today’s newsletter is written by Ariane Bernard, a Paris- and New York-based consultant who focuses on publishing utilities and data products, and is the CEO of a young incubated company, Helio.cloud.