The paid media lab

The episode delivers a handful of genuinely useful practitioner frameworks - headroom within Google before diversifying, the cheapest conversion being least incremental, collapsing the search funnel via conversational AI, and tiered attribution time horizons - but these are interspersed with substantial filler: podcast birthday chat, a digression on UK vs US marketer mindset, and which LLM tools the guest personally prefers. Insight-per-minute is moderate.

“the crazy thing is usually the cheapest conversion is sometimes the least incremental”

“you can take Someone who's having a conversation about something that's very informational, discovery based and then take them on a journey through that conversation to the point where they then buy a product”

The most original idea - that conversational AI could collapse the search funnel by guiding users from informational queries to purchase intent within a single chat session - is genuinely fresh and argued from first principles. Everything else (learning periods, audience over-indexing, MMMs as attribution gold standard) is standard industry thinking that circulates widely in paid media circles.

“if you imagine if they engineered that uh, experience to guide you to a point where they then put something in front of you as a product or service, it's almost like a salesperson”

“you can't say directly comparable because those two studies probably use different methodologies. But let's just say for argument's sake that they were somewhat comparable. The amount of commercial terms that are comparable to Google OpenAI are a lot smaller”

Matt Shenton is a legitimate senior practitioner - Biddable Director at a real agency for eight years, with named large-scale clients including Amazon Prime Video - giving grounded, experience-based answers rather than thought-leader platitudes. He is not a C-suite operator who has owned P&L at scale, and the host is openly a content person rather than a paid media specialist, which limits the depth of dialogue.

“Matt has been a Crowd for the past eight years and moved from the UK to their New York office in 2021. In that role, he's worked across almost all industries and developed paid size strategies for some large multinational brands including Amazon Prime Video”

“at Crab we've developed those tools, we have reporting suites that bring those two data sources together”

Some concrete specifics exist - automated SERP-scraping tools firing every 5-10 minutes for volatile brand bidding, TikTok for beauty as an audience over-index example, B2B SaaS on Reddit - but the most interesting data point (Google vs OpenAI commercial search share) is cited with notable vagueness ('a few weeks ago from a while back') and no client campaign results, dollar figures, or conversion metrics are ever named.

“I think I was looking at a study a few weeks ago from a while back that showed that Google is 14, 15, 16% of searches. Google searches like really high intent, high commercial. And then OpenAI released that um, study a month or so ago and I think they had commercial terms at like 2 or 3%”

“there are automated third party tools that can scrape a syrup every 5, 10 minutes. Is there a competitor? Are you P1 in organic? Put a negative keyword in, pull the ad out, save the money, reassess”

The host introduces a useful follow-up around marketing efficiency ratio and presses specifically on brand search incrementality testing, showing some preparation. However, he opens with extended filler (podcast birthdays, UK vs US mindset), openly admits to being out of his depth multiple times, asks an explicitly 'fairly broad question' on attribution without sharpening it, and never challenges or productively disagrees with any of the guest's claims.

“Um, fairly broad question, just your thoughts on attribution and it is a bit of a mess right now”

“sometimes I get a little bit out of my depth in the conversations because you, you have more hands on experience with paid media than I do”