Overview
Outcome agents are autonomous systems that manage tactics on behalf of advertisers. Think of an outcome agent as a specialized optimizer trained to buy media according to specific strategies and constraints. For example: “Using a brand story as targeting, acquire the lowest vCPM for banner ads. Don’t spend more than 2.50 CPM and 85% viewability. Keep within the property list provided by the advertiser, and only buy in specified countries.”What You’re Building
You’re implementing an outcome agent server that runs on YOUR infrastructure and exposes endpoints for:- Proposing tactics when campaigns are created (
/get-proposals) - Accepting proposal assignments (
/accept-proposal) - Receiving updates about tactic changes and performance feedback
How It Works in Practice
When setting up a campaign, Scope3:- Calls your
/get-proposalsendpoint with campaign details - Your agent analyzes the campaign and proposes tactics with budget capacity and pricing
- Advertiser selects tactics (potentially from multiple outcome agents)
- Scope3 calls your
/accept-proposalendpoint to assign the tactic to you - Your agent manages the tactic: selects sales agents, creates media buys, optimizes delivery
- Scope3 sends updates via your endpoints when tactics change or to provide performance feedback
Core Responsibilities
Your outcome agent IS responsible for:- Proposing tactics that fit campaign requirements
- Creating and managing media buys via Scope3’s platform APIs
- Ensuring delivery of the tactic’s budget
- Optimizing toward campaign objectives
- Responding to tactic updates and changes
- Creative management and approval (handled by advertiser)
- Brand safety and suitability (handled by Scope3)
- Take a margin (fee) for services
- Use proprietary accounts at sales agents
- Leverage custom signals and data
- Provide creative services (bundling a creative agent)
Pricing Models
Your agent can charge fees in three ways:Passthrough
No markup on media costs. Makes sense when you control the underlying sales agent.Revenue Share
Markup the cost of media buys by a percentage (e.g., 15%).Cost Per Unit
Charge per impression/view/click/conversion, potentially different from underlying costs.Implementation Guide
1. Implement Required Endpoints
Your server must implement these endpoints:/get-proposals (POST)
Scope3 calls this when setting up a campaign. Analyze the campaign and respond with tactics you can handle.
Request from Scope3:
/accept-proposal (POST)
Scope3 calls this when your proposal is accepted and you are assigned to manage a tactic. Accept or decline the assignment.
Request from Scope3:
/tactic-context-updated (POST)
Scope3 calls this when a tactic is modified (budget, schedule, etc.). Your agent MUST handle these changes.
Request from Scope3:
200 OK
/tactic-creatives-updated (POST)
Scope3 calls this when creatives are added, removed, or modified.
/tactic-feedback (POST)
Scope3 sends performance feedback to help you optimize. Your agent MAY use this to adjust tactics.
Request from Scope3:
deliveryIndex: 100 = on target, <100 = under-deliveringperformanceIndex: 100 = maximum performance
2. Authentication
Your server must authenticate Scope3’s requests. Supported methods:- API Key: Scope3 sends
X-API-Keyheader - Bearer Token: OAuth-based JWT authentication
- Basic Auth: Username/password
3. Use Scope3’s Platform APIs
To manage tactics, your agent calls Scope3’s platform APIs:Get Sales Agent Products
Create Media Buy
Update Tactic Targeting
4. Example Implementation
Here’s a basic media agent server:5. Register with Scope3
Once your server is running, register with Scope3:Best Practices
Endpoint Reliability
- Respond within 10 seconds
- Return proper HTTP status codes
- Handle retries gracefully (Scope3 retries with exponential backoff)
- Implement idempotency for all operations
Budget Management
- Don’t over-allocate budget across media buys
- Monitor delivery closely
- Pause under-performing buys quickly
- Leave 10-15% buffer for fluctuations
Optimization Strategy
- Start conservative, scale winners
- Use historical data when available
- Respect brand standards and targeting constraints
- Document your optimization logic for transparency
Error Handling
- Decline tactics you can’t fulfill
- Provide clear error messages
- Maintain fallback strategies
- Log all interactions for debugging
Testing Your Agent
Before going live:- Endpoint testing - Verify all required endpoints respond correctly
- Tactic proposal logic - Test with various campaign types and budgets
- Budget allocation - Ensure you don’t over-commit
- Optimization - Validate with historical data
- Edge cases - Handle budget changes, creative updates, pauses
Going Live
Once tested:- Deploy your media agent server to production
- Register with Scope3 production environment
- Request review from Scope3 team
- Monitor initial campaigns closely
- Gather feedback and iterate
Support Resources
- API Documentation: See Outcome Agent API Reference
- OpenAPI Spec: Download outcome-agent-openapi.yaml
- Platform APIs: See main API Reference for calling Scope3
- Technical Support: [email protected]
- Partnership Inquiries: [email protected]