A/B test Facebook posts with Spark

I thought this was in reference to Apache Spark.

Apart from the name debacle I think what they are doing is fascinating. I imagine this would be immensely useful for marketing. Does anyone have an idea how they would be doing this? I suppose they would need data from previously posted facebook stories.

The way this is described is kind of confusing, especially since "dark posting" isn't really a thing anymore (i.e. unpublished posts are simply just ads). Basically how the manual process works:

1. Post several ads (which will show up in people's newsfeed's along with their friend's updates, Page post, etc) that mimics the exact phrase and imagery that you would otherwise put in your Page post.

2. Create like 5-10 of them with different creative, verbiage, etc.

3. Whichever ad gets the most engagement, you copy that ad to a Facebook post which does show up in a Page feed (and if you're lucky in people's newsfeeds). Hope it goes viral and you'll end up with clicks to your content.

Unless you have viral headlining and content ("30 things these 6 girls never thought was possible until now") this won't be that useful, as the cost/benefit of it probably isn't worth it.

I wonder if FB will buy them or build this in themselves. Given the shift to charging for any sort of significant organic reach, it seems natural for anyone marketing on FB to want to do this since the posts they are putting out there are essentially ads this point, and should be tested as such to the extent possible with the tools available.

Having a native tool to do this lets FB assure people that they are getting the most bang for their buck now that they need to pay to play.

They could have an an amazing project here but fail solely because their name is already used by much more well-known projects.

30 seconds of googling could have prevented this.

As if it wasn't confusing enough with Apache Spark, there's Spark Framework (http://sparkjava.com) which also has an extremely similar-looking logo to this Spark.

good luck to them on trying to rank for "spark", i have a hard enough time sifting through spark notes when looking for apache spark stuff.

Why not just write something that's worth a damn and isn't just filler?

I probably won't use an analytics product made by a company who immediately runs into a trademark suit because they're not familiar with one of the preeminent analytics solutions on the market.