For technical teams

Schema, objects & architecture

OSMM is an open-source schema and interoperability layer for the canonical marketing objects and the relationships between them — including the decision artifacts that today live in documents.

The thesis

Structured context is the path to scalable AI

Decision work has never moved into the MarTech stack. It lives in unstructured artifacts: briefs in Word, personas in PDFs, audiences in Excel, journeys in Miro, keyword strategies as CSV exports disconnected from the content they serve. There's no standard for what these artifacts contain — no schema, no required fields, no relational links.

Free-text briefs and PDF decks don't compose. They consume tokens inefficiently, fragment context across systems, and force every workflow to be rebuilt bespoke. The path to efficient, scalable AI context runs through standard data structures — objects, metadata, taxonomies. That is what OSMM is.

Design principles

Built AI-native, governed for stability

AI-native first

Every object is machine-readable and structured to support agent prompts, orchestration, and automation — while staying human-understandable.

Modular & extensible

Compose and extend without breaking compatibility.

Technology-agnostic

Operates across Salesforce, Adobe, Braze, HubSpot, CDPs, DAMs, and workflow systems.

Backward-compatible

Stable objects evolve under strict semver, with deprecation periods and migration guidance.

Closed-loop by design

Learning objects feed insights from measurement back into the durable context layer.

Object scope

Five categories, ~34 object types

The model organizes around five object categories, each with a distinct lifecycle and read/write profile. A representative inventory spanning the seven workflow phases runs roughly 34 object types; the full registry lives in the schema repo.

Context — durable, reusable business intelligence (Business Context, Brand Context, Audience, Persona, Keyword). High-read, low-write.
Work products — structured outputs of decisions (Marketing Strategy, Campaign Strategy, Journey Strategy, Messaging Framework, Creative Brief, Content Strategy).
Configuration — operational logic for orchestration (Journey Configuration, Personalization Configuration).
Measurement — performance data structured for analysis (Campaign Measurement, Performance Measurement). Append-only.
Learning — durable learning that updates context (Customer Insight, Offer / Creative / Journey Performance, Optimization Recommendation).

Relationships are the model

The objects matter less than the links

The canonical lineage is an invariant: every downstream action is traceable upstream to intent and investment, and every measured outcome is linkable to the objective that funded it.

Strategy → Initiative → Campaign → Audience → Offer →
Creative → Channel Execution → Measurement → ROI

A second axis is the feedback loop: learning objects derived from measurement update context objects (Persona, Audience, Brand Context, Keyword). The graph is a cycle, not a pipeline — the longer it runs, the richer the context layer becomes.

Example objects

Typed pointers, not prose

Every reference is a typed pointer to another object. An agent invoking a campaign can resolve the audience, fetch the budget, retrieve active creative, and execute against a coherent structure — without bespoke integration code per system.

{
  "object_type": "campaign",
  "campaign_id": "CMP-1001",
  "objective_id": "OBJ-200",
  "audience_id": "AUD-101",
  "offers": ["OFR-301"],
  "creative_modules": ["CRT-401", "CRT-402"],
  "channels": ["CH-email", "CH-paidsocial"],
  "measurement_framework_id": "M-101",
  "status": "active"
}

A Customer Insight — a learning object generated from post-campaign measurement — proposes targeted, evidence-backed updates to a Persona, carrying a confidence score for human or automated review.

{
  "object_type": "customer_insight",
  "target_object_id": "PER-lapsed_enterprise_buyer",
  "proposed_updates": {
    "pain_points": { "add": ["integration_complexity_anxiety"] }
  },
  "confidence": 0.82,
  "status": "pending_review"
}

Open-source architecture

GitHub as source of truth

Semantic versioning (MAJOR.MINOR.PATCH): major versions introduce breaking changes; minor versions add backward-compatible objects and fields; patches are documentation and corrections. Objects move through four lifecycle states — Draft, Proposed, Stable, Deprecated — and stable objects evolve under formal deprecation periods, not silent breaks.

osmm/
  schemas/
  examples/
  brand/
  CONVENTION.md
  GOVERNANCE.md
  CHANGELOG.md
  README.md

Governance

Maintainer-led, not consensus-paralyzed

Marketing standards fragment quickly when consensus is the goal. The community submits pull requests, change requests, and proposed objects; a small, opinionated maintainer core — operators over theorists, drawn from CRM, MarTech, analytics, creative ops, and data engineering — approves changes. Credible vendor neutrality is non-negotiable.

View the repository Get in touch