We’ve nearly tracked one billion shares at Kaleida. What is a ‘share’ anyway?
Since we started tracking the news last year we’ve collected nearly 1 billion Facebook shares at Kaleida. But what does that mean? In upgrading to Facebook’s latest API we found some inconsistencies. The gaps create confusion and might be open to abuse.
In April Facebook released an upgrade to their API and exposed more data about what a ‘share’ is. They broke the overall figure down into its component parts, showing comments and reactions in addition to URL shares — posting it to a page or your timeline. The new API is version 2.9.
The change clears up some confusion about what a ‘share’ is but exposes other curious things about the way the numbers are reported.
The v2.8 API documentation stated that the share count was an aggregate figure but that it was “approximate”. In earlier versions of the same endpoint the share count only included URL shares but didn’t include likes or comments, or so the documentation said:
We performed a few tests and discovered some interesting characteristics about the data Facebook can provide about a particular URL in version 2.9 of their API. A URL share, a reaction and a comment are all treated as unique items, as you might expect. And the
share_count reported by version 2.8 is the aggregation of those.
But this led to another interesting discovery. Elsewhere in the API, you can ask for a
social_sentence: a readable statement that often gets included on Facebook buttons and apps and many other places. That still is based on the aggregate of shares, reactions and comments, which makes it misleading — in fact, untrue.
In this example, the
social_sentence says “270K people like this”. The article we’re looking at here is Breitbart’s famous and also rather misleading article claiming President Obama awarded himself a medal.
That 270K number in the sentence is in fact a duplicated figure and inaccurate. It’s not 270K people because one person can contribute multiple shares and comments.
Here is the breakdown of social activity for that article:
Those are big numbers, and most of those actions influenced who saw the article in some way. However, in a sample of comments we found 6% were duplicate IDs. And likely many people who commented also reacted or shared.
We also found that the number appears to include private activity in addition to public activity. This may not be new, but we weren’t aware of it until now.
Several things are still unclear. For example, is a like on a comment on a post with a URL considered part of the rolled up engagement count for that URL? The share number was ‘approximate’ in the past, but they don’t say that anymore. Is it no longer approximate?
All this could be considered pedantic, of course, but given Facebook’s influence in the world, particularly on publishers, it seems precision and clarity would be of critical importance. A ‘share’ is, after all, the Facebook currency.
It’s easy to imagine, for example, that someone familiar with the cracks in the system would use it to inflate numbers. Perhaps commenting privately will jack up the total share count to a point where an article starts bubbling up in various social trackers even though real people have yet to discover it.
With a small number of fake accounts someone could inflate all the figures and force something to the surface which then gets the organic viral activity in response to the large fake numbers.
With each release of the API Facebook is exposing more data and becoming more useful to publishers as a result. Unfortunately, the latest release doesn’t go far enough. The confusion over shares — a key currency of the social web — will persist.
The next 1 billion shares Kaleida captures will be more precise than the first billion. Hopefully by the time we hit 10 billion all this will be cleared up.