Focus Group

Area: Marketing Workflow: workflows/marketing/focus-group.json

Purpose

Intake step for Bewomo’s Focus Group product. A requester (client or internal user) submits an asset they want validated (logo, web, ad / content, development), describes the target audience and provides example contacts, and the workflow:

• Captures the brief.
• Persists it to Google Sheets (focus_group_requests tab).
• Notifies the Bewomo back-office team so they can source the audience and run the session.

Optionally — if human feedback ends up low — AI agents trained on the target persona’s psychology can validate alongside humans (claimed ~89% effectiveness). That AI fallback path is not implemented yet; this workflow only captures the preference.

Pipeline (setup)

Form Trigger: Receive Focus Group Request
  → Code: Normalize Focus Group Brief
  → Google Sheets: Append Request    (focus_group_requests tab)
  → Gmail: Send Internal BO Request  (to bo@bewomo.com)

Trigger — form fields

Field	Type	Required	Notes
Requester Name	text	yes
Requester Email	email	yes
Validation Type	dropdown	yes	Logo / Web / Campaign or Content / Development
What You Want To Validate	textarea	yes	E.g. comparative preference, general opinion, specific aspect.
Asset URL	text	no	For Web (or any URL-based asset).
Asset Files	file	no	Multi-file upload. For Logo / Campaign or Content.
Asset Description	textarea	no	For Development or extra context.
Target Audience Description	textarea	yes	Who should validate. Role, industry, seniority, geography, etc.
Target Examples	textarea	yes	Free text — 7 example companies + decision-maker LinkedIn URLs.
AI Agent Fallback	dropdown	yes	Yes — if human feedback is low / Always — alongside humans / No — humans only

Normalized brief shape

{
  "request_id": "<ISO-8601 timestamp>",
  "requester_name": "string",
  "requester_email": "string",
  "validation_type": "Logo | Web | Campaign or Content | Development",
  "validation_question": "string",
  "asset_url": "string",
  "asset_description": "string",
  "asset_files": [
    { "field": "Asset Files", "filename": "logo-v1.png", "mime": "image/png", "size": 12345 }
  ],
  "asset_files_count": 1,
  "target_audience": "string",
  "target_examples": "string",
  "ai_fallback": "string",
  "status": "pending",
  "submitted_at": "<ISO-8601 timestamp>"
}

Persistence — focus_group_requests tab

Lives in the existing workbook ($env.WORKBOOK_ID). One row per submitted request.

Column	Type	Notes
request_id	string	PK. ISO-8601 timestamp.
requester_name	string
requester_email	string
validation_type	enum	Logo / Web / Campaign or Content / Development.
validation_question	string	Free text.
asset_url	string	Nullable.
asset_description	string	Nullable.
asset_files_count	number	Count of uploaded files.
asset_files_summary	json	JSON array of { filename, mime, size } — metadata only, files not stored yet.
target_audience	string
target_examples	string	Free text — 7 LinkedIn URLs + companies.
ai_fallback	string	The requester’s preference for AI agent validators.
status	enum	pending → sourcing → running → completed → dropped. Advanced manually.
submitted_at	string	ISO-8601.
last_updated_at	string	ISO-8601.

Internal BO email

Sent via Gmail to bo@bewomo.com (placeholder — edit in the Gmail node after import).

The body contains the full brief so the BO team has everything to start sourcing the audience.

Tags

commercial.

Credentials

• Google Sheets: googleSheetsOAuth2Api.
• Gmail: gmailOAuth2.

Open questions

• File storage. Files are uploaded to the form but currently only their metadata is recorded. Decide where the actual binaries go (Google Drive folder? S3? n8n binary store with TTL?).
• Type-specific routing. A Switch by validation_type is not wired yet. When each type gets its own follow-up logic (e.g. comparative LLM analysis for Logo, accessibility audit for Web), add a Switch: By Validation Type after the Code node.
• AI agent fallback. The preference is captured but the AI-agent simulation stage is not built. Needs: target-persona prompting, agent personalities, scoring rubric, comparison vs. human results.
• Confirmation to requester. Currently only the form’s response text confirms. Decide if we want a follow-up email to the requester (with the request ID, expected timeline, etc.).
• target_examples structure. Free text now. Could become a repeatable field set (company + LinkedIn URL pairs) or a CSV upload.