> ## Documentation Index
> Fetch the complete documentation index at: https://agenticadvertisingorg-feature-feedback.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.
# L2: Your data, brand, and governance as the control surface
> AdCP Decision-Makers module L2: the brand-side data you control, generation-time brand safety, and why your IAS / DV / Nielsen measurement stack persists. Reasoning only — no code.
# L2: Your data, brand, and governance as the control surface
**\~15 min** | Prerequisite: L1 | Free
If the platform generates the ad (L1), where is your control? It is in **what you put in**. This module is about the inputs a decision-maker's organization provides, and how those inputs — not a campaign UI — are your levers. They're a **dial, not a checklist**: the protocol needs very little to *run* (an account, a budget, what you're buying), and the more — and better — you put in, the more the result reflects your intent.
Three ideas do most of the work:
* **Your inputs are the dial.** The platform generates the ad from what you push in, so your inputs *are* your control. **Brand identity** (voice, guidelines, positioning) shapes how it sounds. A **catalog** drives product recommendations when you're advertising products — there, input quality drives ad quality: a thin catalog (titles and prices, nothing more) produces thin, generic recommendations. A brand-awareness campaign needs no catalog at all — brand identity and a brief carry it. Turn the dial up as you want more control and better results.
* **Brand safety happens at generation time.** You push suitability rules the platform enforces *while the ad is generated*, before anything is shown. That gives you control over how the AI represents your brand that no post-hoc check provides — and where content is generated on the fly and never leaves the platform, it is the only workable mechanism. It is a different problem from adjacency: controlling how the AI talks about your brand, not just avoiding bad placements. Legal and regulatory compliance (COPPA, GDPR, HFSS) is separate and automatic — governance agents enforce shared [Policy Registry](/docs/governance/policy-registry) rules regardless of what you push. **Content standards are the *brand-specific* control you add on top, and they're optional** — reach for them when you want a tighter say over how the AI represents you.
* **Your measurement stack carries over.** You keep your IAS, DV, Nielsen, and Comscore contracts and accreditations and evaluate this channel with the same frameworks — MMM, multi-touch attribution, incrementality. Sponsored Intelligence is a new channel in your plan, not a new measurement paradigm. Two adaptations: pushing conversion events lets platforms optimize toward real outcomes, and on ephemeral AI-generated surfaces the classic *send the page to IAS/DV* adjacency check shifts to the content-standards calibration model — the contract persists, the mechanism adapts.
**Agency?** Your ownership answer is orchestration: you maintain your clients' catalogs and `brand.json` on their behalf and make sure those inputs are rich before a campaign runs.
**Solo or SMB?** Your existing Shopify or commerce feed already *is* the catalog, and a partner handles the plumbing. "Measurement carries over" means your ROAS / CPA dashboards and conversion tracking still work — you don't need IAS / DV / Nielsen contracts to start.
## Reading list
Why your product feed is the main ingredient — titles, descriptions, prices, images become the creative input.
The machine-readable brand identity — voice, visual guidelines, positioning — AI platforms read so they sound like you.
How suitability rules are enforced at generation time rather than verified after the fact.
The model for pushing content standards into platforms and getting an audit trail back.
The short answer: no — you keep your IAS / DV / Nielsen contracts and accreditations.
AdCP is not an MRC-accredited measurement standard; it carries the data your existing tools consume.
## Key concepts
* **The levers you own** — brand identity (`brand.json`) always shapes the output; a catalog and conversion events come in for product and outcome-optimized campaigns; content standards are an optional control on top
* **Input quality drives ad quality** — rich, accurate inputs (a detailed catalog, a clear brand voice) produce strong ads; thin inputs produce generic ones
* **Generation-time enforcement** — content standards applied during generation, not as a blocklist or a third-party bolt-on
* **Compliance is handled for you** — governance agents enforce shared Policy Registry rules (COPPA, GDPR, HFSS) automatically; content standards are your brand-specific control, not your legal backstop
* **Measurement continuity** — you keep your measurement contracts, accreditations, and evaluation frameworks; conversion-event optimization and (on AI-generated surfaces) calibration-based suitability are what adapt
## How good do your inputs need to be?
Better inputs mean more control and better results — this is the cheapest lever you have. Nothing here is a hard prerequisite; it's a **quality dial**, and what matters most depends on what you're advertising.
| Input | Thin | Ready | Strong |
| ------------------------------------------------ | -------------------- | ------------------------------------ | --------------------------------------------------------------------- |
| **Brand identity (`brand.json`)** | logo + name | + voice and tone guidance | + positioning, do/don't language, visual guidelines |
| **Product catalog** *(product campaigns)* | titles + prices only | + descriptions, images, availability | + structured attributes (size, color, category), kept current |
| **Content standards** *(optional brand control)* | none | topics to avoid | + approved claims and suitability rules enforced at generation time |
| **Conversion events** *(optimizing to outcomes)* | none | one core event (purchase / lead) | + the events that define real outcomes, mapped to your success metric |
**Where to spend your prep:** quality compounds in **brand identity** — and a **catalog** if you sell products — so strengthen those before you spend; a sparse feed or a missing brand voice is the thing to **enrich** first. **Content standards** and **conversion events** are controls you turn up when you want tighter brand suitability or outcome optimization — useful, not required. And set **a goal** — what you optimize toward (a target CPA, ROAS, cost-per-engagement, or reach for awareness) — so the platform knows what success means.
## What you'll demonstrate
Sage verifies three demonstrations through conversation — the same for every learner:
1. **Identify the inputs you own as control levers** — brand identity always, a catalog and conversion events for product campaigns — and explain that input quality drives ad quality. `l2_ex1_sc_data_ownership`
2. **Explain generation-time brand safety as control** — control over how the AI represents your brand, distinct from a blocklist and from post-hoc adjacency verification. `l2_ex1_sc_generation_time_brand_safety`
3. **State how your measurement stack carries over** — you keep your IAS / DV / Nielsen contracts and the same MMM / attribution frameworks — and name what adapts (conversion-event optimization; calibration-based suitability on AI-generated surfaces). `l2_ex1_sc_measurement_persists`
## Assessment rubric
| Dimension | Weight | What Sage evaluates |
| ---------------------- | ------ | ------------------------------------------------------ |
| Data ownership clarity | 35% | Identifies the inputs the org owns as control levers |
| Governance model | 30% | Understands generation-time brand safety as control |
| Measurement continuity | 25% | Knows measurement contracts persist and what adapts |
| Org application | 10% | Connects the control surface to their own organization |
Passing threshold: 70%. Scores are internal — you experience the module as a conversation that continues until you have demonstrated mastery.
"I'd like to start certification module L2."