Build personas from audience evidence
Use audience evidence to define personas, then connect approved profiles to prompt setup. Decide whose buying decisions to observe before writing questions.
Use this workflow when
“Use our customer interviews and campaign history to decide whose questions we should measure.”
Before you start
Select the workspace, approved brand materials, de-identified customer and campaign records, and their reporting periods. Leave unsupported ages, job roles, and behavior unspecified.
1. Read the measurement channel and existing setup
orc workspace current --json
orc channels list --json
orc personas list --json
orc prompts list --jsonRecord the stable channel id you will use as YOUR_CHANNEL_ID. After connecting a persona to a question, follow this order: channels list → prompts create --platforms <channel> → runs create --model-channel-id <channel> --wait → answers list.
Read remaining pages before creating a duplicate profile. Matching names do not establish that two profiles describe the same audience.
2. Separate observations from assumptions
Have the agent extract needs, alternatives, buying triggers, and constraints. Record the source document, passage, and date for each finding. Keep stakeholder assumptions separate, and retain conflicting evidence.
3. Register the approved profile
Choose profiles using business relevance and available evidence. This example creates a profile; replace the name and description with approved content.
orc personas create --name "YOUR_PERSONA_NAME" --description "YOUR_APPROVED_SUMMARY" --idempotency-key YOUR_UNIQUE_KEY --json
orc personas get YOUR_PERSONA_ID --jsonRead back the ID returned by creation. To change an existing profile, use the update operation in personas. Keep detailed evidence in the research note rather than copying confidential source text into the description.
4. Connect the profile to channel-scoped questions
Map each audience's needs to awareness, comparison, and purchase questions. Register only approved questions on YOUR_CHANNEL_ID, and preview the request before saving it. The --persona-ids option on prompts create accepts existing IDs in this workspace.
orc prompts create --topic-id YOUR_TOPIC_ID --text 'YOUR_APPROVED_AUDIENCE_QUESTION' --persona-ids YOUR_PERSONA_ID --platforms YOUR_CHANNEL_ID --preview true --idempotency-key YOUR_REQUEST_KEY --json
orc prompts create --topic-id YOUR_TOPIC_ID --text 'YOUR_APPROVED_AUDIENCE_QUESTION' --persona-ids YOUR_PERSONA_ID --platforms YOUR_CHANNEL_ID --idempotency-key YOUR_CREATE_REQUEST_KEY --json
orc runs create --prompt-id YOUR_CREATED_PROMPT_ID --model-channel-id YOUR_CHANNEL_ID --idempotency-key YOUR_RUN_REQUEST_KEY --wait --json
orc answers list --prompt-id YOUR_CREATED_PROMPT_ID --platform YOUR_CHANNEL_ID --jsonSee buyer-stage prompt setup for detailed candidate review and readback guidance.
Result
Return a research note separating evidence from assumptions, approved persona IDs, associated questions, and missing audience evidence. Keep unsupported segments unknown rather than inventing profiles to fill a coverage table.