What does a bloody third-person shooter game about robbing banks at gunpoint have in common with a hyper-casual game that involves flipping a strip of cartoon bacon from a skillet to try and make it land on random objects?
The answer: Not much, but both are classified as “action” games, according to Apple and Google in their respective app stores.
(“Bacon: The Game” is a real thing, by the way, and it’s oddly satisfying.)
Privacy changes on iOS and Android, namely Apple’s AppTrackingTransparency framework and the SDK Runtime API coming soon to Android 13 (part of the Android Privacy Sandbox), mean that contextual targeting is becoming more important within the app stores.
But contextual can’t be all that effective for targeting or monetization if publisher apps are categorized too broadly, said Puneet Gupta, co-founder and chief product officer of Kayzen, a Berlin-based startup that helps developers take their mobile programmatic buying in-house.
Broad strokes
Most game advertisers are also developers, and the purpose of advertising in other games is to drive downloads for their own. User acquisition managers are deeply analytical and data-driven individuals who always have performance on their mind.
But the categories on offer within Apple’s App Store and the Google Play Store haven’t changed much since both stores launched nearly 15 years ago.
And the players of “Armed Heist,” the third-person shooter game mentioned above, are very unlikely to be the same people playing that very casual bacon game. Yet these audiences are lumped together because the games they play are being broadly categorized.
To be fair, the app stores do also offer some additional tagging based on subgenre, but this isn’t consistent across iOS and Android.
“They just don’t go deep enough,” Gupta said. “They’re limited, and the real context isn’t there.”
AdExchanger Daily
Get our editors’ roundup delivered to your inbox every weekday.
Daily Roundup
Ironing out the details
Earlier this year, Kayzen launched a new tool that uses a mix of manual tagging and machine learning to categorize apps more accurately and with a greater level of nuance.
The feature, called smart app categorization, classifies publisher apps based on a hierarchy of three levels: category, genre and subgenre.
Take “Armed Heist.” Rather than opting for “action” as the primary category and calling it a day, Kayzen might tag it as “action” or “core.” (Core games are games that generally require some level of skill and practice and often monetize through in-app purchases.)
In this case, the genre might be “shooter,” and the subgenre could be “third-person shooter,” in order to differentiate it from first-person shooter games.
By comparison, a possible hierarchical categorization for the bacon game could be “hypercasual” as the category (because, really, bacon flipping isn’t an action game, however you want to slice it), “arcade” for the genre and perhaps “dexterity” for the subgenre.
Having manually tagged a large handful of apps itself, Kayzen has been using them [specify what them is here] as seed data in combination with data from third-party partners, including Sensor Tower, to feed a machine learning model as a way to tag other apps automatically.
The apps Kayzen has been able to categorize represent somewhere between 70 billion and 75 billion programmatic ad requests per day. Kayzen is also working on adding to the number of categories, genres and subgenres it has on offer.
On a roll
JustDice, a German developer of rewarded mobile game apps, has been working with Kayzen for just over two years.
Since starting to target campaigns based on smart app categories this year, justDice has seen a 2.5x increase in its install volume and a 6x uptick in the number of apps installed per thousand impressions.
All of justDice’s apps are rewards-based, which means that discount hunters is one of its most active and lucrative audiences. But to find that particular type of user in the past meant scraping through the app stores to find titles similar to its own and creating whitelists, a process that was not automated at all.
Now, justDice can target apps that have “rewarded offers” listed as a subgenre.
This “really provides substantially better context in targeting,” said Miika Kenttämies, managing director at justDice.
Although programmatic isn’t its largest media investment, justDice handles all of what it does run in-house, which is why it started working with Kayzen in the first place as “an extension of our in-house team,” Kenttämies said.
“With DSPs, we require 100% transparency as opposed to a managed black box,” he said. “We see running it in-house as a significant value add to our other media buys due to the transparency and ability for hyper-targeted campaigns across exchanges.”
That’s one way to bring home the bacon.
