There’s a new contextual curation platform in town, and it has the backing of some industry heavy hitters.
The startup, called Classify, aims to evolve contextual and semantic targeting beyond keyword-based taxonomies. It offers buy-side deal curation tools, as well as a sell-side product for publishers to package their inventory into curated deals.
Classify is entering a crowded space of AI-powered contextual curation and sentiment analysis offerings. But the company is already teasing some high-profile integrations thanks to its network of well-connected advisors.
Classify was incorporated in August and exited stealth mode today with the announcement of a partnership with Scope3, a carbon ratings service for online advertising that was co-founded by former AppNexus CEO Brian O’Kelley and which has grown into an ad tech and agentic AI business. With the integration, Classify will be available on Scope3’s Agentic Media Platform, a marketplace of AI-powered custom bidding tools.
Networking on a solution
Classify is the brainchild of Brendan Norman, formerly of Facebook and Unity, and co-founders Rory Partalis, a former product manager for Shazam and Amazon, and Nick Ross, who led the University of Chicago’s AI and machine learning data science clinic.
During his time at Facebook (in the pre-Meta days), Norman was involved in building the platform’s Audience Network, which taught him the ins and outs of contextual targeting, he told AdExchanger. He previously worked on LiveRail’s SSP, which Facebook acquired in 2014. Later he became head of strategic supply for Unity.
After Unity went public, Norman launched a project unrelated to Classify. However, he added, marketing his own project quickly made him disenchanted with the ad targeting solutions available.
“I had a lot of trouble finding the right type of curated inventory to run ads against,” he said. “Contextual data providers would bucket things into the very high-level IAB Content Taxonomy.”
Using a taxonomy-based approach “has been great for the last 15 years or so,” Norman added, but it led him to think there had to be better ways to understand content.
Norman reached out to Partalis, who, since leaving Shazam, had launched a machine learning startup called Partpic that was acquired by Amazon in 2016. So the two decided to work together on building an in-house AI model for contextual ad targeting.
AdExchanger Daily
Get our editors’ roundup delivered to your inbox every weekday.
Daily Roundup
The third co-founder, Ross, was introduced to Norman by a mutual acquaintance and another Classify investor, Mark Robinson, the founder of games analytics company deltaDNA, which was acquired by Unity while Norman worked there.
After connecting, the trio of co-founders brought on Michael Misiewicz, a former data science team leader at AppNexus, as an investor and strategic technology advisor.
From this initial network of friends and connections, Classify proceeded to build an impressive board of advisors and angel investors. Misiewicz’s fellow AppNexus alum, O’Kelley, recently joined as an advisor and investor (not to mention the Scope3 commercial partnership). Classify declined to share details about its funding to date.
In-house AI
Classify’s resources have primarily gone toward the development of its in-house AI model, Norman said. The company has less than 10 employees, with its limited man power concentrated on the engineering side, he said.
Classify initially considered building its contextual and semantic targeting tech on the back of another large language model (LLM), Norman said, and had looked at working with OpenAI’s ChatGPT. But Classify determined doing so would be “really easy, but incredibly expensive,” he said.
Instead, the company worked with Misiewicz to sketch out a plan for building its AI tech from scratch. In doing so, Norman said, Classify avoided some of the outdated approaches baked into legacy programmatic tech. It also tailored the LLM for advertising use cases rather than adapting existing tech for ad tech.
Classify also runs its AI tech on servers that are owned by the company, and is working on having these servers run exclusively on renewable energy, Norman said. Classify believes building sustainable AI solutions is crucial, since the technology’s environmental impact has become almost an afterthought in the AI arms race, he said.
Where does Classify get its training data? The AI solution scrapes publicly available content on the open web and parses it for context and sentiment as a way to find other webpages that map to those same contextual and sentiment signals. The tech’s main focus is on-page content signals, rather than media quality signals about the number of ads on the page or the site’s brand safety, Norman said.
When given a campaign prompt, the solution curates a list of pages that fit the advertiser’s audience parameters. Rather than surfacing similar domains, the solution curates publisher pages down to the URL level.
Although Norman had a lot of criticisms for content taxonomies, he added that Classify is also compatible with taxonomy-based curation, since that is the market default and in high demand.
Building partnerships
Classify collects a CPM-based fee for its curation services, which are layered on top of many DSPs or SSPs that might be in a given supply chain.
Norman declined to name any SSP or DSP partners at this time. However, Index Exchange, Equativ and Media.net are also integrated with Scope3’s agentic AI platform. And Norman highlighted PubMatic as an SSP whose approach to curation is in alignment with Classify’s, though he would not confirm a partnership between the companies.
Classify also has some publisher launch partners, which Norman also declined to name at this time.
For now, Classify is focused on facilitating curation for open web publishers, but it’s open to in-app integrations as well, Norman said. And there is untapped curation potential in the CTV market, he added.
Going forward, Classify will rely on its network of advisors to expand its pool of partnerships, Norman said. The approach has already proven fruitful for solving the initial marketing challenge that led to its founding. Now the company looks to bring that solution to the wider market.
“It’s kind of fun,” Norman said, “solving my own problem through my own personal network and some old clients.”
