---
name: booth-notes-to-crm
description: Converts unstructured show-floor notes (Slack threads, voice-note transcripts, phone notes, emails to self, photos of business cards or notebook pages) into structured lead notes with name, company, topic, pain, timeline, intent level and agreed next step, matched to the badge-scan list when available. Use when reps captured booth conversations outside the lead retrieval app, when notes need attaching to event leads before follow-up, or when someone pastes a booth leads Slack channel and asks for a table.
license: MIT
metadata:
  author: eventmarketingsoftware.com
  version: "1.0"
  category: leads-and-follow-up
  source: https://eventmarketingsoftware.com/skills/booth-notes-to-crm
---

# Booth notes to CRM

The context that makes follow-up work often lives outside the scanner app: a Slack message typed between conversations, a voice memo in the taxi, a business card from the dinner. This skill turns that into one structured row per conversation, matched to the badge-scan list, ready to attach to the lead.

Terms (intent level, badge scan, lead tier, flags) follow [references/shared-terms.md](references/shared-terms.md). The output `notes` text feeds event-lead-scoring and event-follow-up-sequences.

## Intake

1. Read `event-context.md` (or `event-program-context.md`) if present: event name and dates, rep names, product areas, qualifying questions.
2. Ask for anything missing, in one message:
   - The notes: pasted Slack messages or an export, transcripts, text notes, photos.
   - The badge-scan list to match against (clean file from badge-scan-cleanup is best). Optional.
   - Event name and dates, and the time zone the timestamps use.
   - Rep names if the notes only have handles or initials.

## Connectors (optional)

- **Slack connector**: read the named channel or thread for the event dates only. Read only.
- **HubSpot or Salesforce connector**: look up people by email or name and company to match notes to existing records. Write notes only after the approval gate.
- **Google Sheets connector**: read the scan list or write the table to a new tab.
- **Otherwise**: work from pasted text, uploaded files and images, and return a CSV plus a Markdown table.

## Method

### 1. Split into conversations
- One row per person talked to. A single Slack message can cover two people ("Met Dana and her ops lead Priya"): make two rows and link them with the same `conversation_id`.
- Thread replies that add to an earlier message ("update: she booked Thursday") belong to the original row; keep the later timestamp in `updated_at`.
- Ignore chatter that isn't about a contact (lunch plans, booth logistics) and count what you skipped.

### 2. Extract fields
Use the field rules in [references/extraction-rules.md](references/extraction-rules.md). For each row:
- **Who**: first name, last name, company, title, email, phone, exactly as stated. Don't guess an email pattern.
- **What**: `topic` (what they asked about), `pain` (problem in their words), `current_solution`, `timeline`, `budget`, `decision_role` (decision maker, influencer, user, unknown), `competitors_mentioned`.
- **Next step**: what was agreed, who owns it, and when ("demo Thursday 10am", "send case study"). If nothing was agreed, leave it blank; don't invent one.
- **Intent level**: `high`, `medium`, `low` or `unknown`, using the rules in the reference file.
- **Business cards and photos**: read the card text. If a field is unreadable, leave it blank and mark `confidence = low`. If you can't read images in this environment, ask the user to paste the text.
- Keep a short `source_ref` (channel and timestamp, file name, card number) so every row can be traced back.

### 3. Remove what doesn't belong in a CRM
Leave out personal remarks unrelated to business (health, family, appearance, gossip, opinions about the person). Keep only what helps follow-up. Note how many items you dropped, without repeating them.

### 4. Match to the badge-scan list
In this order:
1. Email exact match: `matched`.
2. Full name plus company (cleaned, case-insensitive): `matched`.
3. First name plus company, or full name plus a scan by the same rep within about 30 minutes: `likely`. Show the evidence in `match_basis`.
4. Nothing: `unmatched`. These are often dinner or hallway conversations with no scan; they need a rep to confirm the email before they can be imported.
Never merge a `likely` match without the user confirming. Two possible matches: list both and mark `ambiguous`.

### 5. Write the CRM note text
For each row, build `notes` in one consistent format so it can be appended to the lead's notes or description:

```
[Example Expo 2027, Mar 10, Sam Ortiz] Asked about lead routing when two reps scan the same person. Pain: duplicates in CRM after every show. Current: spreadsheets. Timeline: Q3. Next: demo Thu 10am (Sam).
```

Keep it under 400 characters; no personal remarks; facts only.

## Output

**1. Summary**: messages or files read, conversations found, matched / likely / ambiguous / unmatched counts, items skipped as chatter, personal remarks removed (count only), and any rep who needs to fill gaps.

**2. Structured notes table** (CSV and, if short, Markdown) with columns:
`note_id`, `conversation_id`, `rep`, `captured_at`, `updated_at`, `first_name`, `last_name`, `company`, `title`, `email`, `phone`, `topic`, `pain`, `current_solution`, `timeline`, `budget`, `decision_role`, `competitors_mentioned`, `intent_level`, `next_step`, `next_step_owner`, `next_step_due`, `match_status`, `matched_record_id`, `match_basis`, `confidence`, `notes`, `source_ref`.

**3. Rep follow-up list**: unmatched and ambiguous rows grouped by rep, each with the question to answer ("What's Kwame's email? Card photo was blurry").

**4. Optional merged file**: the badge-scan list with the new `notes` appended (separated by ` / `) and an `intent_level` column, ready for event-lead-scoring.

An example, built from `assets/sample-slack-notes.txt` and `assets/sample-scan-list.csv`, is in [references/examples.md](references/examples.md).

## Validation checklist

- [ ] Every input message about a contact maps to at least one row or is listed as skipped with a reason.
- [ ] Every row has a `source_ref`.
- [ ] No field contains information that wasn't in the source.
- [ ] `likely` and `ambiguous` matches aren't merged without confirmation.
- [ ] Next steps have an owner; due dates are real dates, not "soon".
- [ ] No personal remarks in `notes`.

## Approval gates

- Before writing notes, tasks or field updates to HubSpot or Salesforce, show a preview (record, field, text to add) and get an explicit yes. Append; never overwrite existing notes.
- Don't post summaries back into Slack without showing the message and channel and getting a yes.
- Default to returning files.

## Privacy

- Booth notes are personal data. Keep them in the user's tools; don't search the web for people or enrich them.
- If notes mention EU or UK contacts, carry the consent flag from the scan list; notes don't create consent.
- Unmatched contacts from business cards are fine to follow up one-to-one, but don't add them to marketing lists without consent.

## Related skills

badge-scan-cleanup (clean the scan list first), event-lead-scoring (uses notes and intent level for the lead tier), event-follow-up-sequences (uses notes for personalized first lines), crm-event-import.
