As we begin 2025, Artificial Intelligence will rank high across the plethora of trend reports, ones to watch article content and year forecasts that are published at this time of year. Within the advertising space, AI is a frighteningly homogenous, catch all, term. There is a blurring of the lines between Generative AI, regression models and large language models (LLM), under the umbrella of artificial intelligence.
In the digital advertising ecosystem, we are mostly talking about LLM and regression models and their utility in forecasting outcomes from vast data sets. The most common of these are the ad serving algorithms used across the digital channels.
Dentsu’s spend report outlined that more than 59% of global ad spend was algorithmically enabled media in 2024, and this is set to grow to more than 79% by 2027. The barrier to entry for harnessing algorithmic ad serving is low, the availability is high and so this trend should not be a surprise to many.
The Effects of Algorithmic Convergence
The widespread adoption of advanced technology has led to a universal reliance on algorithmic optimization, yet the fundamental nature and implications of these algorithms remain underreported and underexamined. As these systems strive for a single point of maximum efficiency, large digital ecosystems rely on standardized optimization methods, diminishing opportunities for human-driven adjustments and reducing the unique differences among competitors. Within a given industry, organizations often employ the same, platform-owned optimization tools, making spending levels the primary source of competitive advantage—and leaving those who control these platforms in a strong position.
Additionally, a large portion of advertising spend relies on publicly available data sets, leading many advertisers to reach similar audiences based on comparable data points. Although the move toward cookieless approaches initially prompted exploration of alternative data sources, the momentum has slowed with the discontinuation of certain initiatives. Meanwhile, algorithms that also shape creative outputs compound the challenge: campaigns end up optimized in much the same way as competitors, often using overlapping audiences and producing similar creative elements.
The Solution – The Future of LLMs Lies in Developing Unique Data Sets
A competitive advantage is established once a unique, or even better, proprietary, data set or combination of data sets is created, acquired or earned. This is the key win for brands looking for a competitive advantage in 2025. The use of algorithms is only going to grow, the competitive advantage for businesses lies in a unique data set to feed that algorithm.
At dentsu, we have been tackling this for our clients with three main approaches:
Developing zero party data (ZPD) collection through customer centric strategies - ZPD is obtained when customers willingly share data such as preferences, purchase intentions and behaviors. This is collected through a value exchange, commonly through push notifications, in-app messaging or email. This gives our clients a proprietary data set, based on their actual customers and what they choose to share with the brand.
Utilizing our own dentsu collected data - Using our proprietary data from our industry unique CCS panel, we can take motivations and behaviors and map these into addressable audiences, across programmatically served media.
Combining data sets to create unique solutions – With the Merkury data suite at dentsu, we are able to take a combination of audiences such as first party data (1PD), retail media (2PD) and CCS data to build custom audiences which can be activated against programmatic buying campaigns.
Source: dentsu X
