Building on top of Facebook can lead to big things…
Yesterday, I noticed some unusual activity in Talk&Comment’s Analytics reports; basically, there were no active users.
Talk&Comment, a product I built as a side project a few months ago, has got lately a huge success: more than 50.000 daily active users, with 1.000 new ones each day.
It’s basically, a browser add-on (on Chrome, for now) that adds a layer of voice messaging and commenting superpowers on top of the Facebook web platform. And it has been `very` well welcomed by users across 215 countries.
That’s all countries Google Analytics can track, and more countries -195- than officially in the World, FYI. (and I still have no single clue why it’s being used massively in Iraq.)
Except that everything is sent as a http://talkandcomment.com/* link (that then gets rendered as an inline audio player).
And Facebook blocked it, yesterday. Hence the no active users.
Facebook scans all links from non-authority websites (YouTube, Soundcloud, and the likes) and the “safety system” (part of the spam prevention mechanism) tags them Unsafe, if they show the following signals, ordered by severity (probably more?):
- Piracy site (torrents, warez, …, DMCAed sites).
- Very short domain/perceived as URL shortner.
- Same links get shared in a `spammy` manner.
- Different links get shared exponentially by different users. (the current scenario, but also the behaviour of a software worm.)
And if you’re blacklisted, the only way out to get “whitelisted again” is through users letting Facebook know that it was a mistake.
I do understand that Facebook tries to prevent stealing user data and protecting information, and maintaining Facebook a trusted environment. There’s a note by Facebook that explains how Facebook’s spam prevention system works and how it’s prone to errors. That dates back to 2010.
So Facebook might not be (or do) evil. But the anti-spam infrastructure probably needs a lot more development to be stable enough to bet a whole product on it.
UPDATE (Sept. 5, 2014): Facebook fixed the issue now. Hopefully it won’t happen again… anytime soon!