As holding companies and agencies rush to build their own versions of AI-driven platforms, Horizon Media is pitching its version of Blu as a kind of black-box consultant that helps clients not only build campaigns but also find groups of customers they might have missed.
Overseen by Horizon Media President Bob Lord and run by Horizon Chief Product and Data Officer Domenic Venuto, Blu is essentially a content marketing platform that uses a variety of LLMs to help independent agency clients determine broader business goals through the prism of media (more on creative input later).
Blu was first launched in 2020 as an advanced data platform that employs various machine learning elements. But it has evolved into a soup-to-nuts consulting format for the clients who use it. These customers include retailers like Bob's Discount Furniture, grocery chains like Wegmans, and telecommunications companies like Spectrum. Venuto described it as “AI-native,” meaning it's built from the ground up rather than on top of legacy systems.
This is aimed at using TransUnion data as a data spine while taking into account the update rates of different main LLMs and exploiting them. Lord said one essential element to speeding up processes without sacrificing data quality is organizing data in a flat grid structure. The old dataset was set in table rows, which made it difficult for the AI to read. “In the new AI environment, actually designing the data so that LLMs can easily access it is actually the magic when it comes to speed,” Lord said.
Lord said the system is meant to work within each client's data wall, which he believes is a key point of difference from competitors. “Client A owns the data. It's on its own cloud. As the platform learns, it learns Client A's ways. So those models stay within the client, giving you a competitive advantage. … My competitors don't do that. They reuse models and pool models on behalf of their clients.”
Blu is built with transparency in mind, listing all sources from which LLM obtains information along with prompt results. “If you don't have transparency about the data source, where it comes from, what the data is, how it's trained, what the model looks like, it's not going to be well-received because people won't trust the system,” Lord said.
And the main focus is on building an audience, either to better understand existing customers or finding future customers, explained Zach Kreis, vice president of data solutions at Horizon. He calls himself a “client architect” and sits between the business solutions and product teams.
For each client, Blu is also available as a business consultancy, analyzing the client's business, competitors, and both existing and potential customers. For example, a supermarket client might use Blu to decide where to open a new store based on customer potential and create a media plan for when that store opens. “When we think about media strategy, we recommend the best plan that not only achieves front-end media metrics, but actually moves the business forward,” said Claisse.
“This is where we move from being just an agency of record to becoming more of a growing agency, where we have the tools and capabilities to answer business questions that go beyond just media, the traditional media questions that we were responsible for,” Venuto added.
At least one client using Blu attests to the appeal of the wide range of consulting the platform offers. “You can build your own audiences just for analytical purposes. You can build dashboards and develop modeling to answer questions beyond just targeting this particular ad,” said Lauren Pearson, CRM director at Bob'sdiscount Furniture. “Instead, you can answer questions that meet your internal business needs, such as what these customers are interested in besides furniture, where else they shop, and what their browsing behavior is. This broadens your view to be more like a business consultant than just an agent.”
Venuto said the platform will eventually include creative input as well, and is piloting partnerships to access and optimize creative by plugging in via API, MCP, and agent options. “We see an opportunity to use technology to creatively assemble and disassemble, score and generate, and feed those signals into media and vice versa,” Venuto said.
