Since publishing a piece on the change to the Guardian’s Facebook Platform implementation a couple of weeks ago, I’ve been pointed to a post by Martin Belam, who designed the app in question.

On the decision to host content in an iframe:

In my view every design should be testing a hypothesis, and the hypothesis for displaying Guardian content within Facebook was that 77% of people who visited guardian.co.uk from facebook.com immediately bounced away. We figured that this bounce rate might be reduced by:

1. Keeping people in the Facebook environment, so they were still signed in to chat and still able to see messages and status updates arrive from their friends in the ticker that Facebook were introducing at the same time.

2. Instead of showing them links to content related directly to the story, or to the area of the website which had published it, we would show them links to content that was popular with people like them on Facebook at that moment in time.

We wanted to shift that 77% metric and get more people clicking on a second story.

And something I never questioned:

Nobody outside the Guardian seemed — or still seems — to understand how the Facebook sharing was still driving lots of referrals from facebook.com to the main site, and lots of search traffic began to be related to stories that people had only been exposed to because of a “frictionless share”.

The app was designed to be a leaky bucket, and most of the traffic still ended up on the Guardian website.

And something I’d always suspected:

But as I used to say, “I can show you an analytics graph with the numbers of people visiting. I can’t show you an analytics graph of the number of people being annoyed.”

A real privilege to hear this kind of stuff from the horse’s mouth. To describe the app as a failure would certainly be harsh – it was, after all, changes on the Facebook side which reduced visibility of news content. Nevertheless, the missing ‘graph of annoyed people’ means that we’re still just playing hunch vs. hunch.