We recently published our game on the Playstore and as we all know, unless you have already a big following it’s simply impossible to gain traction and visibility without advertising your app.
Since we are a small team of three people with no online presence at all, we decided we should experiment with a small amount of money with a few ad networks, and depending on the results we might try spending a little more.
To better understand our strategy and our problems(and also, because the game is great, really!), you should download and play a few minutes of our game Hyper Cell Defense :
HYPER CELL DEFENSE
I created an account and considered a few ad networks, but with many of them I didn’t even manage to create a campaign for various reasons.
That left me with the three providers that I actually managed to create and test a campaign: AppBrain, Applovin and Chartboost.
Before delving into the details, it is important to quickly remind you a few of the important terms that I will be using:
CPM: Cost per Mille — How much is spent so that your ad is shown 1000 times.
CPI: Cost per Install — How much will cost, or has costed to acquire an install.
CPC: Cost per Click
Most of the ad networks work with CPM/CPC model, meaning you pay a fixed amount for 1000 impressions or for a single click. This however, does not guarantee you will get any number of installs if you in some way screwed up your campaign.
Since we are complete newbies in the subject, we chose to go for CPI. This way we would know exactly the number of installs we will be getting for our short budget.
For the ad formats, Appbrain displays a banner that looks like this:
Appbrain Banner
Chartboost and Applovin allows you to upload a set of creatives. We used four different videos, and this is the one that performed better:
We split the test in two parts: first 100$ budget targeting only Brazil, then after analyzing the results spending the other 100$ targeting US. We went for Brazil because the CPI was not that high and also because we are Brazilian, so we believe we understand the market (whatever that means).
So I started the campaigns for each of the platforms, trying out different CPIs and measuring how it affected the speed in which it acquired users. These are the results:
Result for Brazil campaigns
Although I said earlier I would do CPI, I initially set Applovin CPI to 0.3, then 0.35, then 0.4 and did not get any impressions. So I just went against my principle and picked some other type of campaign, which they try to optimize the spending with dynamic CPM (whatever that means).
FIRST EXPERIMENT CONCLUSIONS:
Looking just at the numbers, it is clear that Chartboost was the best choice for this experiment. Not only they had the lowest average CPI, they also delivered it fast compared to Appbrain which has a higher CPI.
Applovin was incredibly fast, but also incredibly expensive.
However, we should not just look at the numbers. The quality of those users is also very important. To make sure we were getting quality users, we had this version of the game to always record and submit a “replay” of the first stage to our backend web server. The replay is just a big string containing all the user touches and enemy spawns with timestamps, which I can put it into our replay viewer.
Analyzing the user replays I had a very bad surprise — A lot of the users were not even trying to play the game! Sometimes it looked like humans, but they were definitely not interested in playing the game. Other times it was clearly a robot. Here are some examples of both cases:
Weird Human
Robot most likely
For all of the providers, at least 20% of the acquired users were like that. At first I was very shocked and angry, so I wrote an email to their support with my findings.
Appbrain was the one I found more users like that. However their support was very concerned with my findings as well. They downloaded the game and tested it, and I even sent them a few of the replays so they could watch and share their opinion. At the end, I just accepted there would be bad users when you acquire them in a non organic fashion. In fact, even a few of the organic users presented similar weird patterns — probably crawlers that were downloading and testing the apk to some other alternative playstores.
I also attribute Appbrain bad performance in this aspect due to the ad format: they only display banners, while the other two have 30 seconds video showing the gameplay. This way the player can get a glimpse of what the game looks like and the likelihood of him/her keeping the interest after downloading is greater.
In any case, this is something that you should keep in mind. 20 to 30% of your users will be bad users. And I’m not counting the users that didn’t even submit a replay — meaning they never actually played the game. Bottom line, paid users will never be as good as organic users.
Another interesting finding is that, although we initially just wanted to test the quality of the players, we found a problem with our tutorial: a few players were trying to fire cells touching the dotted line, not to the left of it. This resulted in those players failing the level because they couldn’t fire properly, and most of the times they just abandoned the game.
For our second part of the experiment, we used the same creatives and strategies, only this time we targeted US, which has considerably higher CPI.
These are the results:
US campaigns
Yes, the last line is correct. Applovin burned 300$ (that’s what I had in balance at that time) with 8 installs at a whopping cost of $40.36 per install. That’s what I call a premium user! Not only that, the Click Rate is ridiculously high — for the exact same creatives I was using in Chartboost, it got almost 6 times more clicks in Applovin. Way to go, “dynamic CPM optimizer”!
I immediately wrote their support asking what the hell, and after 3 days I still didn’t get an answer. I will update this article if they ever reply.
Edit: I exchanged a few emails with Applovin support and the conclusion is that I’m too stupid to understand their system:
Onscreen instructions
Apparently I misinterpreted their onscreen instructions saying that they will ensure I’m hitting the inputted CPI (which was 0.5$, not 40.36$).
CONCLUSIONS:
Appbrain is good if you don’t have the resources to build creatives for a customized campaign since they only work with banners. You can always increase your CPI bid to get faster results. Their support is top notch.
Chartboost is our pick to spend a little more money. They deliver results fast and for the lowest price of the three. Additionally, they are specialized in gaming advertising so you can even pick game genres and subgenres when defining your target.
Stay away from Applovin, unless you want to pay 40$ for a single user (which has 20–30% chance of being a bad user).
Finally, always, always go for CPI unless you have a big budget and great creatives which you’re sure will perform well.
Good luck with your campaigns, I hope this article helps you with the first steps in promoting your game.
Portuguese 
English
Leave A Comment