Booking's Homepage Carousel Experiment Fails As Predicted
I was glad to detect this carousel experiment that Booking recently ran on their homepage because we've already seen similar experiments fail in the past. Instead of showing 5 location tiles, they tested a version that only showed 3 tiles at a time with an ability to slide for more - a carousel. It wasn't the automatic slider type that would unleash the wrath of Karl Gilis, but it was a user-invoked slider nevertheless. So how did we know that this a/b test's probability to fail was greater than to succeed? The two evidence-based patterns that hinted at the probable negative outcome of this were More Results and Static Image Options.
Highlighted UI Changes From This Leak
Fewer Location Tiles
Instead of showing 5 location tiles at the same time (A), the variation introduced a carousel that would only show 3 tiles with the option to slide for more (B). The carousel variation (B) also had approximately 10 or 12 tiles in total that would start to repeat after reaching the end.
Interestingly, it should also be noted that both A and B also contained a small "X" in the top right of each tile. Clicking on this "X" would replace a given tile with a different one - an interesting little interaction (I wonder if they tested that as well in the past).
Overall, the two patterns listed below were already suggesting based on past experiments that showing more options (as in the control) might have been better. And in the end, it seems like Booking discovered a similar and consistent result.