November 9
In finance, there is a metric called alpha that measures excess return over the market that an investment manager generates from skill. If the stock market is up 10% over the year, and my portfolio is up 15%, I have generated alpha of 5%. While all investors start with the same publicly available information with access to the same trading platforms and tools, they add alpha by applying deeply researched insights and proprietary data.
These same concepts apply to advertising. But the advertising ecosystem, including agencies and ad tech, should aim higher than the average, commoditized returns of an index fund — and walled gardens are the ultimate advertising index funds. All advertisers have the same data and algorithms, and the platforms optimize campaigns across advertisers, leaving buyers with little control over how their budgets are allocated. Buyers seeking differentiated performance and more control will find it in the open web.
As with investors, though, the most skilled media buyers generate advertising alpha by bringing data, tools and techniques that drive more conversions, lower media costs or ideally both to their campaigns. To do so, they should focus on building campaigns around the best available signals of true intent, applying predictive technologies that maximize conversions per dollar of budget. Brands should judge us based on our ability to generate the best results for their campaigns. That is advertising alpha.
There is little alpha in legacy approaches
Many campaigns are set up using “people-based” audiences that are typically built around deterministically identifiable users. The industry’s preoccupation with deterministic identity is driven by the pursuit of true one-to-one targeting. On a standalone basis, this approach has severe limitations. Reach is constrained to only the minority of impressions with resolvable identity, leaving out a massive amount of potential conversions. Because people-based data sets are commoditized, many media buyers are bidding on the same users across different brands and campaigns, resulting in a premium for those users based on scarcity, not their propensity to convert.
Another common strategy is to leverage third-party audience data selected from a marketplace or curated by the buyer. Segments chosen this way are typically added to a campaign based on the semantic similarity of their name to fuzzily described personas in the campaign brief. At the last mile, this undermines all of the careful audience strategy that went into campaign planning, resulting in buys overweighted with impressions that are unlikely to convert.
These approaches do not deliver advertising alpha. Instead, broadly available signals are used to buy the same audiences and inventory as everyone else with questionable relevance to the brand or its campaign objectives. Campaigns are managed this way not to maximize performance, but to reduce risk and enhance explainability. But it does not drive optimal performance.
Predictive technology unlocks alpha
Predictive audiences, rather than people-based deterministic ones, activate both levers for better campaign performance by decisioning into impressions based on their propensity to convert for each individual campaign. While they have long been available, the tsunami of AI innovation unleashed by LLMs in the last few years has made predictive methods more accessible, scalable and reliable.
The most reliable way to generate alpha is to seed a predictive model with brand first-party data, which provides deep, high-resolution signals of dynamic intent. It is a signal-rich, proprietary data asset, encoding the brand’s behavioral DNA into its campaign. It is the highest-octane fuel for campaigns to drive both more conversions and lower costs, activating both levers for alpha at the same time. While the models are available to anyone, the data is not.
First-party data can take many forms. Most programmatic professionals think immediately of first-party data as placing a pixel or passing a CRM file, which often comes with latency, operational complexity and/or privacy concerns. In the multimodal AI world we live in, there are other intent-rich, first-party signals to provide the foundation for advertising alpha. Campaign conversion reports, search terms and shopping behaviors are all privacy-friendly signals that can unlock both more conversions and more efficient media buys.
The next wave in programmatic technology is custom bidding algorithms. Using those same predictive models to value individual impressions in real time, custom bidding algorithms allow buyers in the open web to pay only up to exactly what an impression is worth, based on their campaign KPIs. With this approach, prebuilt audience segments, demographic proxies and narrow whitelists only constrain scale and performance. As the industry shifts toward valuing impressions based on the propensity to convert, these static approaches will become obsolete.
The true promise of programmatic
The true promise of programmatic advertising is not one-to-one deterministic targeting. It is to bring together all available customer insights and intent signals to maximize campaign performance. It is time for the industry to let go of its preoccupations with deterministic identity, poor audience proxies and whitelists. Scrap the security blanket of explainability.
This op-ed represents the views and opinions of the author and not of The Current, a division of The Trade Desk, or The Trade Desk. The appearance of the op-ed on The Current does not constitute an endorsement by The Current or The Trade Desk.