We Read The Articles We Share And We Share The Articles We Read

    At least we do on BuzzFeed.

    In March, I did a quick analysis to see whether sharing on social networks increased as people spent more time reading.

    The result was #OneCoolChart, but I felt that the question merited some further analysis.

    On http://t.co/dhiT5eU9oa, sharing increases as people spend more time reading #OneCoolChart #BuzzFeedData

    So how are sharing and reading correlated?

    This can be interpreted in (at least) two ways:

    * Do we read the articles we share? Specifically, prior to sharing, do we read the articles?

    * Do we share the articles we read? Specifically, if we read more of an article, are we more likely to share it?

    Do we read articles we share?

    Assuming that time on page is a reasonable proxy for reading (or attention minutes, if you prefer), at a basic level we can check if users that share are spending more time on the article.

    Since time on page can vary wildly by length and type of content (such as video), let's compare the ratio of time on page for sharers to non-sharers by article. We can track sharing through our on-page share buttons and observe how long users spend on each article.

    We restrict to Facebook sharing only, as sharing on Twitter and Pinterest have different characteristics. (Not all sharing on Facebook happens through our share buttons, of course. We are not looking at likes, comments, or sharing directly on Facebook without using our buttons.)

    Looking at the 6,916 articles that were viewed more 10,000 times in August, we found that:

    On average, users that share spend 68% more time on page than users that do not share. For 75% of articles studied, sharers spend at least 45% more time reading than non-sharers.

    The effect does vary depending on the article - anywhere from sharers spending no additional time to as much as more than 300% more time on the page than non-sharers.

    There was a very small number of articles where sharers spent less time than non-sharers. Out of 6,916 articles examined, this was true for only 153 of them.

    Mobile users tend to stay on pages even longer if they share, averaging 80% more time on page.

    The same holds true for our native advertising. Users who share sponsored content from brands advertising on BuzzFeed spend 80% more time on page than those who don't share.

    If we read more of an article, are we more likely to share it?

    This leads to another question: are users more likely to share an article the longer they stay on a page? We know that users that share spend more time on the page compared to users that do not share, but what about the converse? Do users that spend more time reading also share more?

    To examine this on a site-wide level, we have to be careful to scale our data points appropriately. Since articles can vary in length and time on page is heavily affected by device (mobile users tend to spend less time), I rescaled time on page to be a percentage of the median time on page for each article and device combination. Likewise, since share rates are obviously affected by the content of the articles, share rates have been rescaled to be a percentage of the overall share for each article and device combination.

    We restrict this analysis to Facebook sharing as well.

    Up until around 300% of the median time on page, share rates continually increase. The initial spike at 20% of median time may come from our share buttons at the top of the article - once users move past those buttons they're focused on reading.

    We see the same pattern for mobile users.

    Even for our native advertising this trend broadly holds true – the longer users spend on a post, the more likely they are to share.

    It appears that we've stated the obvious.

    1. If you share a BuzzFeed story on Facebook, you have probably spent almost twice as much time reading it than someone who doesn't share it.

    2. The longer you spend on a BuzzFeed story, the more likely you are to share it on Facebook.


    3. The two points above are true for sponsored content.

    The percentage of sharers compared to all readers is quite small – in the single digits. But this small population is critical to the distribution of our content not just to more readers, but to the readers who are most interested in our content. When you listen to readers and serve them, they are curious, engaged, and responsible. There might be a small number of people who share without reading, perhaps for superficial reasons, but the vast majority of people do not.

    What are some other questions we can try to answer?

    * Does the distribution change if we segment by referrer? i.e., will readers spend more or less time before sharing an article if it was recommended to them by a friend?

    * How are Twitter and Pinterest sharing patterns (as related to time on page) different?

    * For some posts, would it be possible to identify where non-sharers dropped off?

    If you have ideas or comments, please email us at datablog@buzzfeed.com