For years, major brands and advertisers have relied on Start.io to provide them with high-quality audience segmentation.
For example, if your next ad campaign is targeting high-income families in the New York City metro area who are interested in vacationing in Europe, we can help, by building an audience based on income, location, and interests.
In total, we offer marketers more than 800 distinct audience segments, which they can activate on virtually every major ad-buying platform.
Start.io uses artificial intelligence to build its audience segments at scale, by studying more than 20 mobile signals that we routinely gather from roughly 2.5 billion mobile devices each month. These privacy-compliant mobile signals include things like the user’s location, their mobile app preferences, device type, mobile keyboard language, and others.
For the past year, Start.io’s R&D team has been training our artificial intelligence models to analyze additional signals embedded in the data, to achieve even greater accuracy.
We’ve experienced significant breakthroughs in this work, resulting in some of our most popular audience segments doubling or tripling in size. Here’s how:
Our original AI models built audience segments based on deterministic data—defined as information that is concrete and observable, such as a user’s current location, device type, or the mobile app they’re using.
Start.io’s first-party data is robust and contains enough contextual information for our original AI models to build audience segments with a high degree of confidence.
Our new AI models find additional patterns in the data using neural networks, a subset of machine learning that processes information like a human brain. We’ve initially focused on improving predictions around a consumer’s age and gender.
We train our neural networks using a multi-stage approach, involving several layers of processing to refine a prediction’s accuracy.
In the first stage, raw data from different mobile apps and data sources is collected and pre-processed to extract relevant demographic attributes. Once the data is pre-processed, it is fed into a neural network that helps us convert any mobile app into a vector of numbers. That information is then fed into machine learning models that have been trained to classify the user’s age and gender.
After a year of R&D, training, and testing, Start.io is confident in its ability to predict age and gender at greater scale. This breakthrough is immediately valuable for our customers because we’re now able to double, and in some cases triple, the size of our age- and gender-related audience segments in the market today.
Neural networks continuously learn and adapt to evolving trends in user behavior. Neural networks are trained to incorporate new datasets into existing models, ensuring that demographic predictions always remain fresh and relevant.
Unlock new levels of precision in audience segmentation with Start.io. Contact us to learn more.
