Episode 208: Exploring causal context graphs and machine customers, starting in retail media networks
Feedback loops, causality data, governed commerce audiences, drift detection, and what happens when agents become the storefront.
This is the 404 Martech newsletter. 4 quotes to ponder, 0 chill for this thing, and 4 ideas to try at work, pulled from this week’s Humans of Martech conversation. Bonus: 4+ remote martech jobs posted last week and 4 martech pros open to work.
Happy Tuesday folks!
How did you stumble into the martech galaxy? I feel like everyone has a different road and I could make a sister podcast that exclusively focuses on career journeys.
Not sure how practical that would be for folks though.
I do touch on career paths a little bit in most guest episodes. It’s almost like careers for folks in martech are like a collection of “tools in a toolbox.” This week, I sat down with Anthony Rotio, and his toolbox is one of the most unique I’ve ever seen.
Anthony is a Harvard computer scientist who started his journey building reinforcement learning robots. Naturally, his first job out of school was… joining AB InBev, where he spent seven years leading marketing for global giants like Budweiser and Bud Light. He eventually took over a struggling retail division with a negative EBITDA and turned it into a profitable growth engine.
But it was a late-night phone call in 2022 that changed a bunch of things for him. A mutual friend connected him with Sam Altman (nbd), who was quietly working on “this chat thing” (which we now know as ChatGPT). When Anthony showed Sam his team’s natural language audience building tool, his excitement gave Anthony the “green light” he needed to lean all in on the future of AI and data strategy.
Today, as the Chief Data Strategy Officer at GrowthLoop, Anthony is helping brands navigate the wild shift from manual execution to agentic commerce.
Here is what we dive into in this episode:
Most Marketing Systems Don’t Learn Because They Lack Feedback Loops
Agent Context Graphs for Drift Detection in Marketing Systems
The Evolution of Retail Media Networks
How Agent to Agent Commerce Operates Inside Marketing Funnels
Google Universal Commerce Protocol Explained
Beyond the tech, Anthony is a “girl dad”, a huge sci-fi book nerd and an avid home chef who believes the ultimate goal of tech should be to leave humanity in a better place than we found it.
Recommended martech tools/agencies
We only partner with products and agencies that are chosen and vetted by us. If you’re interested in partnering, hit me up.
🐾 AttributionApp.com: Multi-touch attribution software that gives you auditable, full-funnel clarity across every channel — find out your “true” CAC.
🦸 RevenueHero: B2B scheduling and routing product to instantly connect prospects with the right sales reps to drive qualified meetings.
🎨 Knak: Go from idea to on-brand email and landing pages in minutes, using AI where it actually matters.
📧 MoEngage: Customer engagement platform that executes cross-channel campaigns and automates personalized experiences based on behavior.
4 quotes to ponder
“When you think about compound interest in finance, the early part looks almost linear. People want big bumps now, even if those bumps never build momentum.”
Compounding demands time, and most marketing orgs run on months and quarters. Anthony is naming the tension that drives half the bad decisions in enterprise marketing. Early compounding looks flat on a chart, so leaders chase spikes they can screenshot, over systems that learn. They trade reinforcement for reaction. Everyone nods at the idea of compounding, and then they behave in ways that kill it.
Durable growth comes from short feedback loops that stack over time. That way you can show progress without torching the engine that creates it.
“You need tight definitions of done and tests that actually matter. Agents can wander all night if the system knows how to evaluate the result.”
Anthony is describing evaluation discipline. Without clear experiment definitions of done, an AI system will generate a lot of output that feels productive and still moves nothing. Clear KPIs, logged interventions, and recorded customer state turn marketing into something testable. Causality data, captured at the moment of action, gives the system a memory. Once you’ve collected enough data, you can simulate decisions instead of blasting customers with variations and hoping one sticks.
“If I know I need toilet paper, deodorant, or the same shirt I always buy, that becomes a very predictable, verifiable realm.”
Retail creates an environment where repetition makes outcomes measurable. Anthony explains that when customers restock the same essentials on familiar cycles, the data becomes stable enough to forecast and act on with confidence. In those moments, intent does not need to be inferred. It shows up directly in purchase behavior.
That predictability is what powers retail media networks. Retailers sell access to confirmed buyers, brands measure impact against real transactions, and performance can be validated quickly. When behavior repeats, measurement holds. When measurement holds, media dollars follow.
“You give it parameters, context, and system instructions, and you have to get through that agent if you want to reach me.”
Machine customers shift power to the interface that stands between brands and buyers. Anthony argues that assistants now act as gatekeepers, evaluating relevance before a human ever sees an offer. Personalization isn’t enough anymore. The agent filters based on rules, context, and constraints set by the user. Marketing no longer persuades a person first. It must first satisfy the logic of the system representing them.
0 chill for this thing
Anthony has 0 chill for using ‘quick wins’ as an excuse to avoid long term marketing investments and feedback loops.
Anthony argues that modern marketing fails because it lacks real feedback. Most teams operate in permanent ‘campaign mode.’ Revenue dips, a new ‘batch and blast’ campaign launches, a report gets pulled, and the cycle repeats without any system that actually learns.
Short-term pressure drives this behavior. Companies report monthly and quarterly, and CMO tenure keeps shrinking, which pushes leaders to chase quick wins. Instead of investing in long term stuff, they optimize for visible spikes. Compounding, however, looks linear at first. Impatient leaders misread that phase, make reactive decisions when numbers drop, and often rip out systems right before they would have started paying off.
Anthony’s solution at GrowthLoop is to build an agent context graph that snapshots causality over time. That means recording where a customer was, what conditions existed, and what intervention occurred. With that foundation, teams can move into verifiable domains, predict impact more accurately, and let results compound. AI will amplify whoever builds the cleanest feedback loop.
4 things to try at work (and outside) this week
Audit your campaign-to-campaign intelligence
Review your last three marketing launches to see if the data from one actually changed the audience or strategy for the next. If you find you're just "rinsing and repeating," document one specific insight from the results that must be applied to your next brief to break the "campaign mode" cycle.Start a causality snapshot log
Choose one key customer segment and work with your data team to start recording 3 specific data points for every interaction: the customer's state at that moment, the message sent, and the resulting behavior. Explore the idea of building the foundation for an "agent context graph," which eventually allows you to answer "what if" questions and simulate outcomes before you ever hit send.Identify your internal "Maverick"
Find one person in a technical role who is frustrated with slow processes and wants to move fast. Latch onto them as a "change agent" to pilot a new data-to-marketing workflow that bypasses traditional silos and provides immediate "speed to value".Hydrate your customer data with SKU-level history
Ensure your marketing activation layer has access to specific purchase history (SKUs) rather than just aggregate "total spend". In a future of agent-to-agent commerce, providing this specific "data hydration" will be the edge that determines if your brand is surfaced to machine shoppers.
Listen, watch or read the full episode here.
Open to work: 4 badass martech pros currently looking
Charmagne Jacobs. I’m a Global B2B Marketing & GTM Executive specializing in making complex SaaS platforms easier to buy, adopt, and scale in multi-stakeholder markets. Led global marketing across 18 markets, driving 456% platform growth while partnering with the CEO and Board on commercialization and investor strategy supporting $25M+ in capital raises. Seeking a VP/EVP/CMO role.
Ian Tremblay. I’m a full-stack digital marketing and operations professional with expertise in paid media, analytics and RevOps for B2B SaaS. I have experience in startups as well as larger organizations, in-house and agency-side. I like to help GtM teams optimize their processes from acquisition to closed-won, integrate tech stacks, and build reporting that cuts through the noise.
Alexander Huzar. 13-year Ex-Oracle Eloqua veteran with deep IT roots. I build “forensic-level” data accountability into SFDC integrations and GDPR workflows. Seeking an Eloqua Ops role to implement proven governance frameworks using programs and other automations. I fix and automate the plumbing fast so you can focus on strategy, not firefighting.
Nicole Alvarez, Most operations fail because they ignore the human element. I get marketing and sales ops aligned for companies that need to scale without breaking. With my background in Cognitive Science, I design for how your team actually thinks and works. My bread and butter is HubSpot, so let me know how I can help make it work better for you.
Open to work? If you’d like to be featured, you can set up or update your free profile here.
Bonus: 4 remote martech jobs posted last week
Lifecycle Marketing Manager at StartEngine (FinTech)
Associate Principal Analyst - Marketing Analytics at Spreetail (ecomm)
Senior Growth Marketing Manager (B2B) at Hatch (CRM)
Revenue Operations Coordinator at Flow (Real estate)
Bonus:
Sr. Director of Communications at Carrot (Fertility and family care)
Channel Operations Specialist, Revenue Operations at Docker (US and Canada)