Jeff Sue discusses how the rise of vertically integrated advertising powered by machine learning is redefining performance through the integration of predictive advertising and direct serving.
For many years, advertising models have been situational. When insurance companies want to sell their products, they prioritize advertising on insurance-related sites.
It was believed that you must go to the place where the signal is strongest. Only those who actively visit the site that provides information about what insurance package they should buy are guaranteed to be potential customers. Cookies can also be used to retarget that user for navigation to other sites.
However, this idea has evolved significantly over the past few years. First of all, not everyone looking to buy insurance visits these sites. Second, cookie options continue to decline due to tightening legislation, aggressive decisions by tech giants to minimize cookies, and widespread privacy concerns.
We need a new playbook.
The deterministic identity that is the foundation on which the entire buy-side stack is built can no longer be trusted. But advertisers still need results to get the most out of their spend. It’s just that the data they have to work with is much less obvious. This raises the difficult question: If we cannot rely on identity, then what can we rely on?
The new ad stack is based on probability
Programmatic is being redefined from a marketplace of transactions to an ecosystem of results-driven systems. Winning platforms in the next phase will operate more like operating systems than brokers, integrating, predicting, and optimizing for business outcomes rather than media metrics.
If someone might want insurance, our probabilistic system can find them among hundreds of thousands of apps based on behavioral patterns that correlate with intent and conversion.
People who play Candy Crush may want insurance, but too few advertisers want to target them because it’s “not an insurance game.”
For many advertisers, this may be counterintuitive. Why find insurance customers in a puzzle game? Because people don’t live in content silos. People who need insurance don’t spend every waking moment thinking about it or consuming only insurance content. Additionally, although the profile is exactly the same, the cost of advertising within the game is much lower than on a financial news site.
It can be probabilistic because there are predictions
The answer is prediction. This is not a prediction in the vague sense of a buzzword. Predictions like Calci and Polimarket do not predict Oscar winners or the next president. We’re talking about prediction as a measurable feature. This is the ability to use the signals that remain available to infer intent, likelihood, and outcome.
And just as importantly, do this in a compliant, privacy-secure, and reproducible way at scale. The key is an SDK embedded directly into your app and an infrastructure that enables compliance, privacy, and scale without relying on identity signals.
Prediction is no longer a feature. It is becoming a commodity
Your real superpower is being able to predict where the audience you want to influence is and when and how to reach them.
A platform that can consistently predict outcomes (installs, purchases, subscriptions, retention rates, return on ad spend) will increase its value over time. Platforms that cannot do this will continue to lose out to those that can.
This is where SDK-based integration separates the disciplines. Platforms that integrate directly across large app footprints can provide additional intelligence beyond purchasing inventory.
In fact, the SDK helps marketers observe performance signals at the source, understand placement dynamics, render creatives more effectively, and close the feedback loop between delivery and results in near real-time. The SDK can bring in any type of advertiser with any type of outcome and make it work based on predictions and machine learning.
We live in a world with few signals. The people who capture all the value are the ones who can work with less effort and still get results.
Platforms that rely on purchasing supply through intermediaries pay structural taxes of less control, lower signal quality, and slower learning cycles. By owning the SDK, the platform eliminates the “middle margin” and, more importantly, reduces latency. For ML-driven bidding, seeing data 50ms faster than your competitors through the SSP bridge is a huge competitive moat.
That’s because there are so many players in this field who are reselling things. It’s not a real “technique”. It’s just a media buy, or just a deal. Not being able to supply directly is the same as having a 20% or 30% handicap.
Even if two systems are equally sophisticated on paper, the one that has more control over the end-to-end environment will learn faster and run more efficiently.
Ad networks are often just a source and a place to advertise. But a platform that advertisers connect to to solve results, no matter how complex the path to those results.
Programmatic execution layers are more automated, more predictable, and more responsive than what the web in a human-operated setting can realistically accommodate. These platforms are achievement machines. They take a goal, translate it into thousands of decisions per second, and continually optimize through learning.
Platforms that have direct access to SDKs across their footprint will perform better than those that aggregate third-party supplies. Even if the latter technically reaches more stock. It’s no longer about how much data you have. Supply integration is more important than supply quantity.
where do we go from here
Expansion has already begun. Direct-to-consumer brands, broad e-commerce, financial services, and performance-focused marketers who want efficient growth without the need for perfect identity resolution have predictive systems as the most powerful toolkit available. DTC advertisers are spending six figures per day on mobile advertising, a number that was unheard of more than a year ago.
If predictive systems can deliver performance outside of “obvious” contextual environments, the addressable market for these platforms goes far beyond mobile games.
Ultimately, you’ll be able to combine scale and efficiency to reach high-value audiences at the right time, wherever they are.
For more specialized articles and industry updates, check out: martech news
jeff sue General Manager, Americas, Mintegral
Jeff Sue is the General Manager of Mintegral, a mobile advertising platform that powers the world’s largest developers with global user acquisition, ad monetization, and creative services.
