Here's a chart of Facebook's overall app score according to Applause Analytics over time (since 2011). It is an abstraction on the generic star rating.
What it shows is that there has been a general trend down in their app score - there seems to be a further dip around the time of Messenger but the reviews seem to indicate a lot of frustration around stability.
For apps that have been popular for a long time, I wonder if there is a negative bias in reviews (beyond that already present in all online reviews). A customer who gets an update that breaks an app they use every day is very likely to submit a negative review (even if they have reviewed positively in the past). However, a customer that gets and update that make incremental changes that largely don't affect them has no incentive to go post a positive review of the new version. So, the positive reviews come only from new customers, which are relatively small for apps that have reached saturation.
One interesting sub-trend in the data would support that view - it is that the review star rating, meaning reviews+star rating is, on average, more negative than just star rating. So people post a negative review when they feel pain, but not in the case where they are just rating the app.
I still don't understand why people hate Messenger so much. It works fine for me, indistinguishable from when it was included in the main Facebook app.
Messenger is much more intrusive to the whole phone experience than the Facebook messaging functionality that was included in the app. Though, to be honest, even that went downhill in the same direction as Messenger -- though not as far -- before it was removed and messenger made mandatory.
Many don't trust FB anymore or the invasive data tracking of Messenger. I've overheard comments from friends and family that they won't be forced to download another app, especially one that keeps such close watch on users.
Ideally open-source. Surprised MongoDB 2.2+ is such an issue but thanks for the feedback and suggestions. We'll look into PostgreSQL, Riak and RethinkDB.
But dremel doesn't support incremental updates. Dremel is designed for read-only data. All its columns are indexed. The whole table needs to be re-built after update.
I think it will affect the supply rather than the demand of HTML5 games initially. Firefox OS has a very small market share in comparison to iOS and Android so while there are more folks playing HTML5 games, it isn't a huge amount. However, big name companies backing HTML5 like Mozilla with Firefox OS and Samsung/Intel with Tizen, they're doing their best to get popular games ported over to HTML5, as well as more new content.
Best fake recommendations ever! But yeah we make the assumption the developer and their app (food delivery or otherwise) has some data (users, items, actions) on which to build a model. Otherwise you won't get very far...
Not so familiar with Keen IO. Looks like analytics-as-a-service for collecting and visualizing data. We're focused on helping developers build predictive features like recommendation, discovery, etc.
Sounds like nReduce.com but don't think it worked out. Agree with gaborcselle that bonding would not be the same without meeting in person. Also cohort effects are diminished with smaller rolling cohorts. Starting and finishing something at the same time is a big deal.
Think about your best friends. I would hazard a guess they are probably from school and college. Now compare to friends you've made taking MOOCs (i.e., remote) and at work (i.e., rolling cohorts). Over simplified but you get the idea.
Wow, that was quick! Smart move by Cloudera. If any existing users are looking for an open-source alternative to Myrrix for a scalable recommender system build on Hadoop/Mahout check out http://prediction.io
