Why We're Huge Geeks About Delivering the Best Ads


User Acquisition

April 27, 2015


min read

Imagine a player is on his or her umpteenth failed attempt at Flappy Bird and sees an ad flash across the screen for a ... teen mental health clinic.Huh? It's a true story our Head of International, Pepe Agell, often tells, and this unfortunate mismatch between ads and players happens all the time. Seeing an ad for a completely irrelevant game ? or worse, an unrelated product or service ? diminishes the gaming experience for the player and discredits the advertiser.At Chartboost, we're obsessed with making the mobile game advertising experience valuable for players, advertisers and publishers alike. Every day, we analyze new variables and data points to make sure our servers deliver the most relevant messages to the right audience. (You might even say we're huge geeks about it.)Here's a step-by-step explanation of how our servers make the best possible match between players and mobile gaming ads:

Step 1: Gathering Data

When you install Chartboost's SDK in your mobile game, it sends relevant player data (the app ID, player location, player device, etc.) to the Chartboost ad server, which currently supports more than 150,000 titles and 20 billion monthly gaming sessions.

geeking out images-01

Step 2: Ranking Ads

Next, this rich data is crunched through an algorithm that ranks advertisements by eCPM, a monetization metric that measures the effectiveness of an ad. This algorithm, which is constantly being updated and improved is special because it takes into account both bid and conversion rates for different ads. Most of the time, conversion rates take precedence over bids, but given the choice between a $20 bid that seldom converts users and a $1 bid that converts nearly every time, we?d opt for the latter. Looking at both bid and conversion rates has the added advantage of providing more opportunity to advertisers with smaller pockets. We?re also constantly tweaking our algorithm for ranking ads in the same way devs run A/B tests on their games ? just on a much more massive scale.Say, for example, you've built a Candy Crush Saga-style match-three puzzle game and you're looking to acquire U.S.-based users. The ad server will rank other match-three puzzle games higher than, say, ads for an MMO or casino game. Every player's data generates a new and distinctly ranked list of the ads in our servers. And because we're constantly slicing and dicing our ever-growing data pool in myriad ways, the rankings are fluid?and unique to Chartboost.

geeking out images-02

Step 3: Applying Business Logic

Then, this enormous, optimized list of ads gets further narrowed down in the validation stage?when business logic-based rules filter out ads based on targeting variables. These include game category, device type, model and operating system, where the player lives, which games are already installed on the device and numerous other rules that deliver the most optimal ad for a specific player.So for our Candy Crush Saga-style game example, this second step eliminates all the ads that aren't relevant to your target audience, like ads for games they've already installed, can't access in their home country or aren't supported on their particular device type?leaving your ad as the best match for the player.In other words, delivering ads to your target audience is more complex, and effective, than simply showing the ads that generate the highest revenue. Rather, a massive and constantly growing set of variables impact how your players see the ads they do. And it all happens in a cool 80 milliseconds.Geeking out a little? Welcome to our world.

geeking out images-03