Extracting Themes from Scope3’s Product Announcements

Brian O’Kelley’s Scope3 pivoted back into ad tech. We dive in to see what lessons we can learn.

Extracting Themes from Scope3’s Product Announcements

Admittedly I’m using my platform to give BOK a lot of airtime. I tweet-stormed his announcement, had him on the main stage at my event, rebroadcasted our conversation on the podcast, and now I’m dedicating my weekly newsletter to his recent product announcements.

I’m not doing this to necessarily predict success or failure of his business initiatives, but because his series of announcements unpack so many interesting themes within the currently moribund ad tech zeitgeist that I believe they will be seen as a watershed moment in this particular era.

Recap

If you haven’t been following, BOK and Scope3 made a bunch of products announcements that I think can be best summarized as:

• A brand safety product using custom-trained AI

• An SPO-in-a-box product that allows you to create your own deal IDs across DSPs (not yet available)

• An “agentic AI” agent that that curates and scores inventory for you at either the SSP or DSP level (also not yet available)

Theme 1: The era of AI for the open web

If there’s one thing the current crop of AI large language models really excels at, its understanding big blocks of text. And what is the open web, if not a very large corpus of blocks of text?

LLMs can evaluate a webpage for targeting, brand suitability, tone, emotional content, and really anything else of value to advertisers at a level and scale that is vastly better than most of the solutions currently in the market. That is a fact. If you were building much of ad tech today from scratch you would certainly use LLMs for the contextual and safety functions, it wouldn’t even be a debate.

The first big theme from the Scope3 announcements was an ushering in of the era of AI to the open web at scale. They join another group of ad tech veterans, the Goodhart brothers at Mobian, in moving beyond keywords and taxonomies, and thus vastly improving both the coverage and accuracy of these types of solutions. Within a couple of years, every solution in the market will be AI-first.

Theme 2: The end of keywords and taxonomies

Taxonomies have been a core part of digital advertising from the start. We’ve got the IAB content taxonomy, along side those offered by private vendors. DMPs used to tout their taxonomies as key selling points to buyers, since the seemingly tidy bucketing of data allowed buyers to better plan media.

The problem is that LLMs don’t need hierarchy or taxonomies to do their jobs, only people do.

A media planner feels comfortable recommending that her client targets “beauty” and “lifestyle” content. An LLM can just target whatever content is most appropriate for the clients’ customers, without the middle step of classification and human readability. BOK wrote on his blog about this, and gave examples of how AI is better than taxonomies for categorizing things of interest to advertisers.

The end of taxonomies will take a while, since humans really do like them. Trusting AI to make certain decisions may also be bundled with systems like PMax that take away other controls, so there will be natural resistance. But you can be confident that taxonomies will getting filed away, perhaps in technologies→antiquated.

Theme 3: Curation is the anti-SPO

I’ve written this before (see Curation is the anti-SPO, plus Audigent interview.)— the best way to understand curation is to see it as pushback against DSP efforts to control supply — it is the anti-SPO.

Quick history lesson:

1. Header bidding becomes widespread ~2014-2017

2. QPS at the DSP level skyrockets due to duplication

3. DSPs start crudely cutting down supply, dubbed “SPO”

4. DSPs start cutting out exchanges to reduce fees and duplication

5. SSPs strike back by “curating” Deal IDs that DSPs can’t throttle

What’s interesting about this little history lesson is that the concept of “SPO” has never really been tied to a specific product a customer could buy. There was no “box” on the LumaScape for SPO, it was just something that existed between systems. As such, SPO and curation had unclear market impact for actual brands and agencies, mostly in the form of performance buys through vendors like Audigent, and in some cases agreements between agencies and exchanges for incentive payments.

Scope3’s agentic agent is the first major product that essentially does SPO on the buyer’s behalf, agnostic to which SSPs or DSPs may be involved. There are some others, like Swym, but if Scope3 can really deliver what they demo-ed at their event it would be the debut of a brand new category of ad tech, and one that would go a long way to shifting the balance away from DSPs and towards exchanges and curation. Never a dull day on the ad tech beat.

Theme 4: Custom algos get portable

The business of creating custom algorithms has been around for 10+ years and was a big part of how my company, Beeswax, went to market. We’ve also got Scibids, Chalice.ai, Cognitiv.ai, and others.

While Scope3 did not technically announce a custom bidding solution, it was pretty obvious that’s where they were headed. What was interesting about their solution was the idea that they could train an agent for a particular brand, and then implement that agent in a mixture of buy-side DSPs, like The Trade Desk, and sell-side curation platforms, like Index Exchange. Model portability can be a major selling point for custom algorithms and maybe could expand the market significantly. (cue angry emails from everyone saying they already do that).

Theme 5: Carbon is not a viable ad tech market

Call it what you want, but Scope3 just pivoted, and hard. The company was the clear leader in the nascent carbon measurement business and decided it was not a viable path forward. My understanding is that competitor Good-Loop is also backtracking from carbon towards more general corporate responsibility.

What caused this retreat:

• Measuring carbon ended up being entirely correlated with measuring quality. The ad ecosystem burns carbon by doing stupid things, like reselling the same ad slot multiple times, duplicating bid requests, and supporting made-for-advertising sites.

• Corporate mandates for reducing carbon in ads were not as widespread as expected.

• Political pressure disappeared with developments like the SEC backtracking on requirements for carbon.

Theme 6: Innovation is still possible in advertising

AI is a massive, disruptive technology wave headed towards the advertising industry at a rapid pace. Disruption means opportunity, and the Scope3 announcements show that — as a recent meme I stole from a random newsletter - You can just do things.

Reading list

• Marketecture Live news: Innovid brand launch (link)

• Marketecture Live news: Perplexity session covered by AdExchanger (link)

• Netflix Ad Suite (NAS) going live April 1, apparently (link)

• Sam Altman non-commital on ads for ChatGPT (link)

• DoubleVerify updates its “News Accelerator” products and policies (link)

• Brian Weiser revises US ad forecast downward (link)

• Roku experimenting w/ autoplay ads that load before the home screen (link)

• DV360 adds household targeting using IP addresses (link)

• Ziff Davis acquires The Skimm (link)

• Wonder acquires Tastemade for $90m (link)