Online targeting has been focused on features, such as geography, age group, etc., as a way to define optimal users. What these features have in common is that they remain constant over time. Your interests, for example, stay the same for months, at the very least.
In contrast, your willingness to engage in a specific action changes over time, often during the course of a day. For instance, you might be a pizza lover, but you wouldn’t eat one for breakfast. You might need a sofa, but you’re less likely to buy one made of leather if, yesterday, you downloaded a vegan-recipes app. You might be an avid gamer, but you’re less likely to buy Fortnite skins if you didn’t purchase any during the last three months.
Even so, ads for these and other products will appear online based mostly on users’ fixed features.
It’s a 2D targeting strategy in a 3D world
Over the past three years, Bigabid has developed a revolutionary in-app data platform that looks at the dynamic conditions that make up the state of the user just before he/she performs a desired action.
Once an action is made by the user (e.g., download, purchase), we register the complex composition of the moment that preceded it, focusing on the user’s:
- Long-term history (as far as regulation allows, often over three months): For example, the user time spent on hyper-casual games has increased over time.
- Immediate history (past 48 hours): For example, the user downloaded two apps yesterday, a news app and a language-learning app.
- Present actions: For example, time of day, device used, etc.
- User patterns and behavior: Using a proprietary tool we developed called “Deep Categories,” we seek to understand the users through a high-resolution understating of the apps they user engaged with. For example, user is crazy about music and slot machines and mildly enjoys puzzles, etc.
- Fixed features: age group, geography, etc.
Based on these consensual data points, we can craft an understanding of this ideal pre-action state. We then scour billions of users and millions of moments in search of temporal variables and recurring patterns, utilizing the full potential of machine learning.
Our AI algorithm is now able to predict what makes up the ideal state when a particular message is relevant for a certain user.
We call this ‘Precondition Targeting’
This robust and proprietary mechanism serves millions of ideally timed in-app ads per day, which translates into bouts of growth and increasing ROAS (by up to 42 percent!) for both our UA and retargeting campaigns.