"Marketing technology is a product manager whose specific role is the system or platform. It brings together processes and systems from product, growth, engineering, and marketing."
Not just third-party tools — it's first-party solutions too
People + Process + Platform
Critical at 100-150 people; before that, distributed
Role Definition
What Does a MarTech Person Actually Do?
Tool Architecture: Design and manage the stack of first and third-party tools
Build + Buy: Buy tools to get 90% there, then build on top for the remaining 10%
Contract Negotiation: Manage SaaS deals and liability exposure as you scale
Data Flow: Ensure data passes correctly between tools and systems
Platform Operations: Build homegrown solutions to power advanced capabilities
100–150
critical headcount to hire MarTech
30–40
where it's still distributed/village model
The unsexy, high-leverage workContract negotiation. Most startups sign any terms early. At scale, you optimize deals and avoid liability — can save or cost millions as you grow.
Org Design
Where Should MarTech Report?
B2C (Centralized): MarTech reports to Head of Growth or into Product Ops. Growth team is the primary customer. Simple funnel: acquire → product retention.
B2B (Messy): Often lives in Rev Ops because you need to serve both top-of-funnel acquisition AND bottom-of-funnel CRM/Salesforce. The data architecture becomes complex.
B2B2C (Most Complex): You have user acquisition at the top AND company/business objects. Notion, Ramp: managing both HubSpot AND Salesforce is a nightmare with no perfect solution.
Key principle
Follow the pain. Hire MarTech when tools become too complex for one person to manage and systems become a blocker to business plans.
Interview signal
A great MarTech hire is a strong technical architect who can represent the team to engineering and product orgs.
The Change
The Death of Deterministic Data
2010–2020: You knew exactly who installed your app, had their IDFA, could track with precision
Now: Deterministic matching is gone. Apple's ATT, privacy laws, cookie death
The New World: Probabilistic modeling. Build cohort models that work for 30% of users, extrapolate to 100%
Austin on the shift"Ad networks are becoming more sophisticated at the same time it's harder for marketers to understand how money is spent. We're now building models from partial data."
Framework
Attribution Setup: Build for MTA From Day 1
✗Start with first-touch onlyINSTEAD →✓ Design your data collection for multi-touch attribution from day one. Store first and last UTMs on every user.
✗Collect only UTM parametersINSTEAD →✓ Collect referrer, URL, UTMs, AND ad network IDs (Google Click ID, FPID, TikTok ID, etc.). They're free data.
✗Assume you'll add attribution laterINSTEAD →✓ You can't retroactively collect data. Set up the infrastructure now; you'll thank yourself in 2 years.
✗Skip it because it seems complexINSTEAD →✓ It's actually simple: track first UTM, track last UTM on every user event. Fire events with both. That's it.