The potential of personalisation

How much could personalisation increase engagement?

Here is my best guess at engagement for 100 stories on a typical news homepage. (I estimate that relevant stories have a CTR of 15% as opposed to 10%).

  • 30 stories relevant tag/category. 4.5 interesting
  • 70 stories not relevant. 7 interesting
  • TOTAL: 11.5 interesting

With a mix of personalisation and curated:

  • 60 stories relevant tag/category. 9 interesting
  • 40 stories not relevant. 4 interesting
  • TOTAL: 13 interesting

That’s an increase of 13%. I’ll update the numbers as i continue to gather data.


Personalization: the secret sauce

In my last post i discussed why personalization is a good thing. In this post i’ll talk about how to implement it.

But first, here are some more reasons for personalization.

  1. To scroll through a day’s worth of Metro headlines on mobile would take 40 pages on an iPhone5.
  2. Metro users rarely use the site navigation on mobile – only 2% of visits use it!


My demo from early 2014 shows how a user’s interests can be deduced when the they are shown one headline at a time (a la Metro10). Read the detail in this post.

In fact, this can be generalized to a stream of headlines (a la on mobile) and even more complicated desktop layouts.

How it could work

My algorithm relies on knowing if the user has clicked on an article or not.

When there is a stream of headlines, “not” is given when the post scrolls out of view (beyond the top of the viewport).

This works for more complicated desktop layouts too, if you ignore the sidebar (assume that users view the posts in the main column). You may also need to ignore some widgets (e.g. the horizontal trending bar on Metro squeezes 5 posts on one line so users aren’t likely to absorb all of them).

What do you think?


How your news brand can reclaim homepage traffic from Facebook

The popularity of Facebook is incredible. A third of the UK population visit the site every day. The average time per visit is 20 minutes.

People don’t visit newspaper homepages any more, they visit Facebook instead (especially on mobile).

Mark Zuckerburg says Facebook will be “the best personalized newspaper in the world”.

What’s even more incredible is that news sites are making very little attempt to fight back. It’s very much business as usual. Want to change it? Here’s how…

1. Personalize

To be a destination, it helps to have the right blend of content. Of course, the “right blend” is different for everyone. Millennials are narcissistic and want to see a reflection of their own interests.

The answer is personalization. There are two tricks:

  • Blend it… 70% personalized 30% “main line” stories
  • Make it simple by avoiding the whole faff around registration/login

2. Aggregate

Facebook (and Drudge Report) aggregate content from other places. Sites that want to be user-centric should do the same.

Users care about the personality of their content, not which publisher it’s from. Aggregate stories that fit your brand and users will love you for it.

For example, Metro might link to an UsVsTh3m game, a MailOnline celebrity gallery, or an i100 infographic.

The bulk of content (say 80%) would remain pure Metro, but if another site does a story better then link to it.

3. Turn consumers into co-creaters

A recent Wired article about how to make products compelling says, “The key is for the user to contribute some element of their own—a tweet, a comment, a video—and for that, in turn, to set in motion a chain of events resulting in the delivery of the next trigger”.

Upworthy have a “Submit link” menu in their site navigation.

4. Let users express themselves

People love to express themselves. When Metro added polls to the site it was massively popular – 5 million votes in one month!

It worked so well because the polls were simple – just two options. It’s simple but more expressive than a Facebook Like.

Bringing it all together

Personalized news that has personality and none of the dross of my Facebook feed… and i get to contribute / give my opinion? Yes please!

All four hypotheses above are verifiable with relatively simple experiments (example). Admittedly polls didn’t have an obvious impact on homepage traffic but the next iteration might have an impact.

UPDATE: For a broader look at how make your homepage a destination, check out my more recent post.

Top categories: a probabilistic approach

Here is a common scenario. A sequence of article headlines is shown to a user. The articles have different categories. The user reads some of the articles. Which categories does he like? Which categories does he dislike? Probability to the rescue…

Consider the example of an app that shows 10 headlines. Each article has a category: news, sports or showbiz.

Consider a user, let’s call him Bob, who reads 3 articles per session (on average). For Bob, each article has a (3 / 10) = 30% chance of being read. This is the “global” average.

We can calculate the % for each category. Let’s say that from the last 10 articles…

3 news articles were suggested, 1 was read. Success rate = (1 / 3) = 33%
3 sport articles were suggested, 2 were read. Success rate = (2 / 3) = 67%
4 showbiz articles were suggested, 0 were read. Success rate = (0 / 4) = 0%

We can infer that Bob likes sport and news but dislikes showbiz. Any category with a success rate greater or equal to the “global” average (30%) is a success. (The threshold for dislike is best determined by experiment).

Obviously, the results get more accurate as the history increases.

We could show Bob more news and sport, and show him less showbiz. Will this skew the results? No, the probabilistic approach is robust with respect to self-fulfilling prophecies. It will self-correct quickly if Bob occasionally gets shown an article from a “cold” category.