Should you mess with the machine?
This was the question we were asking ourselves. For one account we saw that the Instagram feed clearly outperformed Facebook news feed.
This was happening when we were letting the machine do its thing – and when we were letting the algorithms optimize our campaigns.
Which begged the question -> couldn’t we just double our overall return on ad spend by just targeting Instagram instead of all our placements(Facebook, Instagram, Messenger)?
Tempting.
But with Facebook issuing forth strong recommendations for Automatic Placements – and avoiding manually targeting placements, this was an area that we wanted to tread with care in.
So. We ran a test(for those of you being nitpicky -> we used Facebook’s split test feature). This is what happened.
Boom. Machine pwns human.
It turns out, our manual selection of Instagram as a placement yielded far *worse* results than automatic placements. Also Instagram as a placement targeted by itself performed far worse than it did as a part of an automatic placements mix. Our hypothesis for what happened is this: the algorithm has more data points to pick high-value users from with automatic placements than with manual placements.
In other words, sure, users coming from the Facebook news feed might be worse on average than users coming from the Instagram feed alone- but if you choose automatic placements, you get the best users from the Facebook news feed *and* the Instagram feed, which on the whole tend to be more valuable than just the best users from the Instagram feed alone.
As it were, Facebook’s internal literature essentially explains this exact behavior intuitively(source: https://fbinsights.files.wordpress.com/2016/07/facebookiq_placement-optimization.pdf ).
As is becoming increasingly true: when confronted with a machine, do not mess with it.
ABOUT ROCKETSHIP HQ:
Mobile User Acquisition Show | How Things Grow | LinkedIn | Twitter | YouTube
(Pic credit: @cdr6934 on Unsplash)