The website Edgerank checker released a report this week suggesting that 3rd party publishing tools such as Conversocial, Hootsuite, Tweetdeck etc are punished by Facebook's EdgeRank algorithm. As one of Facebook's preferred developers (PDC), we speak directly to Facebook's platform and development on a regular basis; and our team have been working on the Facebook platform for a combined 15 years. We have a lot of direct data and experience on these kind of questions, and we can confirm 100% (and have had this confirmed by Facebook directly) that third party APIs are NOT punished at all in Facebook's EdgeRank algorithm.
However, there are some reasons why, on a general look at 3rd party published posts vs native Facebook published posts, Facebook native posts would have a higher average engagement, which the study mentions briefly. Below we examine these in a bit more detail, and give some more information to their validity:
Facebook Collapses 3rd party API Updates
When there are a large number of updates from the same app in a user's Facebook newsfeed, Facebook will sometimes collapse those, only showing one by default, with a 'click for more from...' to see the rest. If you are using a widespread free tool, such as Hootsuite or Tweetdeck, which have millions of accounts, it's quite possible that this collapsing could happen on a regular basis. This is far more unlikely for a premium enterprise tool like Conversocial, however it is still possible.
This was initially designed to fight spam. Facebook are aware of this issue for publishing applications, and are working on a way of making this better.
UPDATE: Facebook have announced that this no longer happens for any page publishing tools, so they are treated in the same way as native posts.
High Chance of Being Scheduled or Automated
Third party tools are often used to enable automatic posting, for example via RSS feeds. Although Conversocial allows this, our own internal data from clients show that in general automated posts do get much lower engagement than manually written posts. However, there is no difference between a manually written post that's published natively on Facebook, instantly via Conversocial, or scheduled to be published at a later date, for example in the evening. In fact, some of our clients have reported up to 30% increases in post engagement by utilising insight from Conversocial on when their fans are most active, and scheduling their posts accordingly.
So, as the vast majority of automated posts come from 3rd party tools, and these get much lower engagement, this is likely to skew the results - and says nothing about manually written or scheduled posts via 3rd party tools.
Content not optimised for Facebook
In their study they point out that many third party tools allow cross posting of the same message to multiple social networks - Facebook, Twitter, LinkedIn, Blogs etc. They point out that this makes the content worse for each of those platforms. We agree - Facebook, Twitter et al are very different. People use them differently, consume them differently, and interact with each other in different ways on them. Writing the exact same content across channels is a Bad Idea. That's why in Conversocial, we allow you to post to multiple Facebook pages at once, or multiple Twitter accounts at once, but not to post the same update to both Facebook and Twitter. So, this point is probably true for other 3rd party apps, but not for Conversocial.
I hope that helps clear up any confusion. If you have any specific questions, feel free to email us on firstname.lastname@example.org or tweet us @conversocial.
Follow us on Twitter here.
UPDATE: Facebook announced that they found a bug causing some publishing tools to reduce the number of impressions; this was because it was aggregating the 'quality score' of posts (which control how many impressions the posts get) across all the pages using the publishing tool. They have now fixed this bug and re-confirmed that page posts using third party tools are treated the same as native posts. Read their announcement on the bug fix here.