For years, the advertising model was contextual. If an insurance provider wanted to sell its product, it would prioritize advertising on insurance-related sites.
The thinking was that you had to go where the signals were strongest. Only those actively visiting a site that offers information about what insurance packages to buy could be guaranteed to be potential customers. And, with cookies, you could also retarget those individuals as they left to go to other sites.
But the thinking has evolved significantly. Today, not every individual looking to buy insurance visits those sites. Plus, cookies are increasingly less appealing due to growing legislation and deprioritization.
A new playbook is required.
Deterministic identity, the foundation the entire buy-side stack was built on, is no longer reliable. But if you cannot depend on identity, what can you depend on?
The new advertising stack is based on probability
Programmatic is being redefined from a marketplace of transactions into an ecosystem of outcome-driven systems. Effective platforms will behave less like brokers and more like operating systems: integrated, predictive and optimized for business results rather than media metrics.
If someone is likely to want insurance, a probabilistic system can find them across hundreds of thousands of apps based on behavioral patterns that correlate with intent and conversion.
Somebody playing Candy Crush might want insurance, but too few advertisers think to target those users because it’s “not an insurance game.”
To many advertisers, this might feel unintuitive. Why would I find an insurance customer inside a puzzle game? Because people do not live in content silos. Someone who needs insurance isn’t spending every waking moment thinking about insurance and consuming only insurance content. Further, the cost to advertise in a game is much lower than on a financial news site, though the profile is exactly the same.
Prediction is becoming the product
The answer is prediction: the ability to infer intent, likelihood and outcomes using the signals that remain available.
And, just as importantly, doing this in a way that is compliant, privacy-safe and repeatable at scale. The key is SDKs embedded directly in apps, infrastructure that delivers compliance, privacy and scale without relying on identity signals.
The real superpower is being able to predict where the audiences you want to influence will be and how and when to reach them.
The platforms that can consistently predict outcomes (installs, purchases, subscriptions, retention, return on advertising spend) will be the ones that compound value over time. The platforms that cannot will keep losing ground to the ones that can.
Good prediction requires two things most legacy platforms struggle to combine: direct access to supply and a learning system that improves with usage.
This is where SDK-based integration separates the field. Platforms with direct integrations across a large app footprint can provide additional intelligence beyond buying inventory.
An SDK can help marketers observe performance signals at the source, understand placement dynamics, render creative more effectively and close the feedback loop between delivery and outcome in near real time. The SDK can ingest any type of advertiser with any type of outcome and can make it work based on prediction and machine learning.
We’re living in a world of fewer signals. Those who are accruing value are the ones who can work with less but still deliver outcomes.
Platforms that rely on buying supply through intermediaries pay a structural tax: less control, weaker signal quality, slower learning cycles. By owning the SDK, the platform removes the “Middleman Margin” and reduces latency. In ML-driven bidding, seeing the data 50ms faster than a competitor via an SSP bridge is a massive competitive moat.
There are so many players in the space that are just reselling media. Not having direct supply is akin to a 20% or 30% handicap.
Even if two systems are equally sophisticated on paper, the one that controls more of the end-to-end environment learns faster and executes more efficiently.
Ad networks are often merely a supply source and a place where ads run. But platforms that advertisers plug into solve for an outcome, regardless of how complex the path to that outcome is.
The programmatic execution layer becomes more automated, more predictive and more responsive than a human-operated web of settings can realistically match. These platforms are outcome machines. They ingest an objective, translate it into thousands of decisions per second and continuously optimize through learning.
A platform with direct SDK access across its footprint will outperform one that aggregates third-party supply, even if the latter technically reaches more inventory. It’s no longer about how much data one has. Supply integration matters more than supply volume.
Where we go from here
The expansion is already underway. Direct-to-consumer brands, broader ecommerce, financial services, performance-minded marketers who want efficient growth without requiring perfect identity resolution have predictive systems as the strongest toolkit available. There are DTC advertisers spending six figures a day on mobile advertising, a figure unheard of more than a year ago.
If predictive systems can deliver performance outside “obvious” contextual environments, the addressable market for these platforms stretches far beyond mobile gaming.
They can finally combine scale with efficiency to reach high-value audiences at the right time, wherever they are.
“Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.
Follow Mintegral and AdExchanger on LinkedIn.
For more articles featuring Jeff Sue, click here.
