Google Play Icon Experiments
Introduction
So, you spend years working on your game, iterating, improving, polishing every single pixel and even adding some screenshake so that the players will love it. But, at the end of the day, what will make a lot of those "blind" players see you is how good (or bad) your game icon looks. This is one of the first marketing tips we, devs, get when it comes to standing out in the stores. The icon must be attractive, striking and eye-catching. But how much can it affect our downloads? Time to find out...
Store Listing Experiments
If you don’t know, store listing experiments is a great feature provided by Google that allow us to test combinations of text and graphics of our game on Google Play and find the ones that bring the most installs. We can change the game icon, screenshots and short descriptions and the Play Store will automatically distribute users across different store listing pages and measure the conversion of our experiments.
And what should we change?!
Google suggests that we test one attribute at a time to get the clearest results. It's recommended to start testing with the game icon, because this is the change that will have the greater impact on whether users pick our game from search results and category lists. Screenshots and descriptions may also be changed, but, according to some experiments I saw from other devs, the results are much less impacted by them. This is where I started, I followed the recommendation and provided some variants for the game icon
Experiment
This is the Tower Crush’s current icon. I think it looks really good, but I wanted to make sure it is not just my biased opinion, after all, I've been looking at this icon for a long time and got used to it. The current icon is used as the control variant.
And here are the variants. One is more focused in the character (A), other in the tower (B) and the last one is also featuring the game character, but with a different take, more 3D.
We can't assign more than 50% for the variants (combined). So, I assigned 15% for each new variant and the control variant got the rest 55%.
Results
Without further ado, here are the results of the icon experiment.
The variant B, the tower, was the best performing variant. I confess that I was a little surprised by the outcome of this experiment. In my opinion, it was the simplest variation. But that's the greatest advantage of store listing experiments, we can take decisions based on statistical data and not on opinions :)
Another huge insight from the experiment is that if I had chosen the third variant (C) as the original icon and never made any experiment, this decision, alone, would have costed almost 50% of my downloads. If we think about how expensive user acquisition costs are today, a bad choice here may cost thousands and thousands of dollars. And it's just an icon...
Conclusion
Store listing experiments are a great feature provided to Android devs. Easy to setup and may have a great impact on downloads. The beauty of it is to stop relying on personal tastes and biased opinions and leave the decision to the “collective wisdom”. If you haven’t used it yet, start now. It’s going to be the cheapest user acquisition campaign you will ever make.