Google Ads, YouTube, agentic AI, and marketing strategy — written to connect dots, not confirm what you already know. The essays here feed the operating manuals in the Frameworks collection and the live experiments in the Lab.
Every marketing KPI in common use measures activity, not decisions. The gap between the two is where margin and competitive advantage quietly leak out of the modern enterprise — and it has been that way since the discipline was instrumented. Agentic AI is the first technology that makes the correction operable, not aspirational.
"Marketing KPIs and business KPIs have been structurally misaligned since the discipline was instrumented.
AI is what makes the correction operable."
Everyone agrees AI agents need a kill switch. Nobody will name the number at which theirs fires. The kill condition is the priced threshold at which a human takes the decision class back — and "leave it alone" is a kill condition too, set to infinity.
On July 1, Google's new Ads terms authorized automation by default and left you the liability. Nothing changed in your account, and that's the problem: a default is a decision made for whoever hasn't priced the alternative.
Every AI agent needs an owner — settled. But you own a decision class, not a bot, and the thing nobody is owning is what the decision costs. An unowned decision is an unpriced decision: governance assigns blame, decision economics assigns a price.
Cannes declared the AI-novelty era over: prove the result, don't just use the tool. But proof is a ratio, and the industry forgot the denominator. The denominator is the decision — here's how to make proof countable.
Google's new metric credits campaigns with conversions that haven't happened yet — and it's headed for Meridian. How QFC works, where the line is, and the pre-QFC baseline to freeze while there's still a "before."
The cart was the last first-party event before revenue. Universal Cart moves it inside Search, YouTube, Gmail, and Gemini — and takes the Retail Decision Stack's ground truth with it. The fix: move it to the order.
Google now predicts your conversions with Gemini and feeds them into the same Meridian MMM meant to verify your spend. The optimizer and the auditor share one brain — and independence just became a setting you have to defend.
AI made the machine faster and left your org as the slow part of the loop. The time between a signal arriving and a budget moving is now measurable — in days and in dollars — and it's the metric nobody is paid to lower.
AI Max for Search, the SQR shift, and the prose-input change ship together this quarter. One prediction layer, three surfaces, one sealed auction — and one decision every marketing team has to make before September.
Consumer FinTechs run on Cost Per Lead. The business runs on funded, retained accounts. The space between is where neobanks burn working capital chasing the wrong applicant — every quarter.
Retail marketing runs on Cost Per Click. The business runs on retained margin per cart. The Retail Decision Stack rewrites the KPI hierarchy for an AI-mediated funnel — for DTC, big-box, and CPG.
Health plans run their marketing on Cost Per Enrollment. The business runs on member-years. That gap is where the budget gets spent on the wrong audiences, the wrong creative, and the wrong channels — every AEP, every cycle.
The five signal layers that determine whether Performance Max compounds or fails — and the audit checklist most teams skip.
Default attribution, modeled conversions, consent-mode gaps, and the specific places your data is over-crediting some channels and burying others. With the audit checklist.
Google's open-source Marketing Mix Model isn't just a measurement tool. It's the confidence infrastructure that lets enterprise teams finally stop defending budgets with broken attribution.
A marketing decision doesn't stop at marketing. It moves through CAC, LTV/CAC, working capital, and the multiple the market pays — the chain most teams never draw end-to-end.
Most organizations are running Google's products as separate channels. The ones pulling ahead are running the whole system.
Beyond prompts and chatbots — AI that acts, decides, and executes across systems. What this means for how marketing gets done.
Most teams are locking their workflows to a single model. That's the wrong bet. Here's how to build infrastructure that survives the next model shift.
In 2016, AlphaGo played a move no human would make. That single moment explains where AI and marketing are heading — and why most teams are still playing the old way.