Create custom scoring models
The Scoring Studio skill in Adobe Marketo Optimizer provides an AI-native lead scoring solution that enables you to create, configure, and publish lead scoring models. The studio combines an agent-driven workflow with a visual UI — you can build scoring models through natural language prompts in the Coworker chat interface or by interacting directly with the UI controls.
- Skill -
scoring-studio - Invocation - Use a slash command to open Scoring Studio. For example: “open Scoring Studio.”
- Reads from / writes to - Marketo Optimizer scoring service; reads Marketo Engage lead fields and activity types
On launch, Coworker automatically fetches relevant context — including activity types, lead fields, person lists, and existing score lists — to ground its suggestions in your data.
Create a scoring model create-model
When you open Scoring Studio, Coworker proposes a relevant example scoring model pre-populated with a static list and a set of scored activities. You can accept this suggested starting point or provide your own prompt to define a custom model.
Preview the model preview-model
After you provide a prompt, Coworker generates a model preview before making any changes. The preview surfaces:
- Scoring dimensions in use
- Attributes and activities being scored
- Static lists or smart lists applied as segments
- A summary of the model goal, target segment, and primary signals
You can review the preview and choose to create the model based on it, or continue refining through the chat before finalizing.
Model structure model-structure
The created model is organized into dimensions and signals. You can configure each signal using the property panel in the UI:
- Signal type — Activity-based or attribute-based
- Activity or attribute — The specific item to score
- Signal parameters — Adjustable settings for the signal
You can build and configure models entirely through Coworker using natural language, or interact directly with the UI controls.
Publish a scoring model publish-model
When your model is finalized, instruct Coworker to publish it. The publish process handles the following automatically:
After publishing, you also have the option to trigger a manual run to process scores immediately.
View scoring results view-results
When a scoring run completes, scores are written back to Marketo Engage via the lead import process. After the import completes, updated scores can be verified directly in Marketo Engage.
After each run, you can view a results summary that shows:
- How many people were scored
- The individual score changes per person
An audit log is available for reviewing additional run details.