# Deck

> Deck is an AI-powered customer feedback platform that consolidates feedback from interviews, surveys, support tickets, sales calls, and messaging tools — then uses multi-agent AI pipelines to extract structured insights, themes, and trends so product teams can build what customers actually need.

Deck is built for product-centric teams drowning in fragmented customer feedback. If you're juggling insights across surveys, CS tickets, Slack messages, sales calls, and user interviews — and struggling to see the full picture — Deck brings it all together.

---

## Overview

### The Problem: Scattered Feedback

Modern product teams face the Scattered Feedback Problem — feedback lives in CRMs, survey tools, support platforms, and interview recordings, making it nearly impossible to have a full picture of the customer. Teams spend days manually tagging, sorting, and analyzing feedback across disconnected tools.

### What Deck Does

Deck solves this by connecting all those dots:

- **Centralized insights** — Combine survey, CRM, support, and interview feedback into one shared workspace.
- **Integrations that matter** — Sync CRMs, messaging apps, meeting tools, and survey platforms. 16+ integrations available.
- **Automatic synthesis** — Deck's AI Agents extract themes, insights, and summaries from raw feedback in minutes and organize them into topics you can make good decisions from.
- **Decision-ready context** — Integrate with CRMs to filter insights by account size, churn risk, or user segment.

### Core Capabilities

- **Feedback Consolidation**: Ingest feedback from 16+ sources including user interviews, customer support tickets, surveys, sales calls, Slack messages, and CRM data into a single platform
- **AI-Powered Synthesis**: Multi-stage agent pipelines using OpenAI, Anthropic Claude, and Google Gemini extract insights, assign themes, validate quotes, and score confidence automatically
- **Theme Detection**: AI automatically identifies and groups recurring themes across all feedback sources — pain points, feature requests, delights, usability issues, bugs, and general feedback
- **NPS Analytics**: Detect, track, and analyze Net Promoter Scores across Slack, surveys, and direct imports with weekly AI-generated trend synthesis
- **MCP Server**: Query your customer feedback from any AI client (Claude Desktop, ChatGPT, Cursor, VS Code, Windsurf, and more) using natural language via the Model Context Protocol
- **CRM Enrichment**: Connect HubSpot or Salesforce to filter and segment insights by CRM properties like company size, deal stage, or industry
- **Video Clip Extraction**: Automatically extract and store relevant video clips from user interviews linked to specific insights and quotes
- **Weekly Slack Digests**: Automated weekly feedback synthesis and NPS trend reports delivered directly to your Slack workspace
- **Multi-Region Data Hosting**: Choose where your data is stored — US, EU, or Australia — for compliance and latency
- **Unified Contacts**: Automatically consolidate customer identities across all integrated tools using email, phone, and platform-specific IDs
- **CSV Import**: Bulk import historical feedback data (NPS, interviews, tickets, surveys) with AI processing

### Who Deck Is For

- **Product managers** who need a unified view of customer sentiment to prioritize the right features
- **UX researchers** who conduct user interviews and need structured, searchable synthesis
- **Customer success teams** who want to surface recurring support patterns and escalations
- **Product-led startups** that want to build feedback loops into their workflow without hiring a research team
- **Any product-centric team** drowning in feedback spread across too many tools

---

## Getting Started

### Step 1: Set Up Your Organization Context

Before processing feedback, configure your organization context so Deck's AI agents can interpret feedback accurately for your specific business.

Navigate to **Settings → Context** and fill in three types of context:

#### Company Information

Captures the high-level story of your business: your mission, target segments, business model, and what success looks like for your customers. This is not marketing copy — it's the pragmatic context you'd give a new PM on their first day so they can immediately understand which problems matter most.

Best practices:
- Write for a new teammate — explain as if onboarding a senior PM who has never seen your product
- Anchor on customers — emphasize who your ideal customers are, what jobs they hire you for, and what outcomes they care about
- Be specific about segments — call out key slices like SMB vs. Enterprise, self-serve vs. sales-led
- Capture strategic focus — mention current bets or themes
- Keep it alive — revisit whenever your positioning, ICP, or go-to-market focus changes

#### Product Information

Explains what you actually ship: your core modules, workflows, surfaces, and the main value they deliver. This helps Deck's Agents map feedback to the right parts of your product.

Best practices:
- Describe core modules and workflows — break your product into major areas and explain what each is for
- Connect to user goals — for each module, describe what users are trying to accomplish
- Include canonical flows — call out the 3-5 most important user journeys
- Note constraints and edge cases — mention important limitations or dependencies
- Avoid roadmap details — focus on what exists today

#### Keywords

Captures the exact terms and phrases your customers and internal teams use to talk about your product, problems, and workflows. Teaches Deck's Agents your organization's language so that synonyms, acronyms, and local jargon are interpreted correctly.

Best practices:
- Start from real conversations — pull phrases from call transcripts, support tickets, and sales notes
- Include synonyms and variants — add multiple ways customers refer to the same idea
- Capture internal jargon carefully — document team-specific terms and explain what they mean
- Focus on recurring concepts — prioritize words that show up frequently across feedback sources
- Review periodically — remove outdated terms and add new ones

When you keep these three types of context up to date, Deck's Agents can group feedback more accurately, name themes using your team's language, disambiguate similar terms, and highlight what matters for your specific strategy.

### Step 2: Connect Integrations

Navigate to **Settings → Integrations** and connect the tools where your customer feedback lives. Deck supports 16 integrations across CRM, support, sales, meetings, surveys, messaging, and knowledge tools.

Each integration uses OAuth for secure authentication. Once connected, Deck automatically syncs historical data and sets up ongoing sync schedules.

### Step 3: Start Processing Feedback

Once integrations are connected, Deck's AI agents begin processing your feedback automatically. You can also:
- Upload user interview recordings or transcripts manually
- Import historical feedback via CSV
- Configure Slack channels for real-time feedback collection

All processed feedback appears in the **Visualize** environment with themes, insights, and trends.

---

## Core Concepts

### Themes

Themes are the overarching areas of your product that Deck's AI Agents detect based on your customer feedback.

For example, for a streaming company, recurring themes could be "Offline Streaming", "Content Recommendation Experience", or "New Content". These may all be recurring themes in the customer's mind. Once Deck detects customers are speaking about it, Deck creates these themes and every time customers mention it, Deck automatically adds the feedback to the created theme.

### Insights

An insight is the atomic unit of feedback in Deck. They can be findings from user quotes, customer Slack messages, support tickets, or survey responses. Insights are the foundation of how Deck works.

A Theme is made out of a collection of insights that relate to the same topic.

Each insight has the following data attached:
- **Name**: A descriptive label for the user's intention (picked by you or Deck's Agents)
- **Customer quote**: The verbatim quote from the customer that led to the insight
- **Sentiment**: The tone of the user's quote (Positive, Negative, or Neutral)
- **Category**: A label that buckets the insight into a pre-defined set of insight types
- **Theme**: The overarching subject the insight is a part of

### Categories

Categories are labels added to each insight about what that insight is about. The six categories are:

- **Pain Points** — Customer frustrations, difficulties, and negative experiences
- **Delights** — Positive experiences, features customers love
- **Feature Requests** — Requests for new features or improvements
- **General Feedback** — Neutral observations, suggestions, general comments
- **Usability Issues** — Problems with UI, navigation, accessibility
- **Bugs & Errors** — Technical issues, crashes, unexpected behavior

Categories help you filter insights based on your customer's intention behind the feedback. You can filter insights by category to find pain points to solve, feature requests to prioritize, or delights to amplify.

Categories are tightly connected to Sentiment — Pain Points will always have negative sentiment, but the distinction is important the more data you have, as you're able to better prioritize what to build based on categories and use sentiments to help with it.

---

## Feedback Methods

### User Interviews

User interviews are the art of talking to your customer to uncover improvement opportunities. In Deck, you can:

1. Upload a customer interview video or audio file — or paste/upload a raw transcript directly
2. Add your research goal and discussion guide
3. Get the interview auto-tagged by Deck's Agents (in less than 5 minutes)
4. Visualize and edit the interview post-agentic synthesis

#### Upload Options

| Method | Best for |
|--------|---------|
| **Video or audio file** | Recorded interviews — Deck transcribes automatically via AssemblyAI |
| **Transcript file or paste** | When you already have a transcript (`.vtt`, `.txt`, Google Meet export) |

Both paths produce the same AI-powered insights, themes, and quotes.

#### Transcript Upload

Deck lets you upload a raw interview transcript and get the same AI-powered synthesis, themes, and insights as a full video interview — in seconds.

**Supported formats:**

| Format | How to export | Auto-detected? |
|--------|--------------|---------------|
| **WebVTT** (`.vtt`) | Export from Zoom, Teams, Google Meet, or most transcription tools | Yes |
| **Google Meet** (`.txt`) | Download the transcript from a Google Meet recording | Yes |
| **Plain Text** | Any unformatted text separated by blank lines — no speaker labels or timestamps required | Yes (3+ paragraphs) |

You can also paste the transcript text directly — Deck handles format detection automatically.

**How to upload a transcript:**

1. Navigate to **Manage → User Interviews**
2. Click **Add Interview** — a two-step upload wizard opens
3. In the **Upload** step, select the **Transcript** tab
4. Choose how you want to add your transcript:
   - **Upload a file** — drag and drop a `.vtt`, `.txt`, or `.docx` file
   - **Paste text** — paste the transcript directly into the text area
5. Deck automatically detects the format and shows a live preview of the parsed segments
6. Enter a required **Interview Name** and **Interview Date**, and optionally a customer email
7. Click **Next** to proceed to the **Review** step
8. Edit the transcript segments as needed (rename speakers, fix text, add or remove segments)
9. Click **Upload Interview** to start AI processing

**What happens after upload:**

1. **Insight extraction** — key themes and pain points are identified
2. **Quote mapping** — verbatim quotes are linked to each insight
3. **Theme tagging** — insights are grouped into existing themes or new ones are created
4. **Summary generation** — a narrative summary of the interview is produced

Processing typically completes in under 2 minutes for transcripts, ~5 minutes for video/audio.

**Tips for best results:**
- Minimum length: 100 words and 10 lines
- Maximum length: up to 50,000 words
- Use consistent speaker names throughout the transcript
- Remove introductory notes or boilerplate that isn't part of the conversation

**Transcript vs. Video comparison:**

| | Video/Audio | Transcript |
|--|-------------|------------|
| Transcription | Automatic via AssemblyAI | You provide the transcript |
| Video clips | Available (jump to moment) | Not available |
| Processing time | ~5 minutes | ~2 minutes |
| Insights & themes | Yes | Yes |
| Speaker identification | Yes | Yes |
| Quote extraction | Yes | Yes |

### Surveys

Deck automatically imports and analyzes survey responses from integrated platforms, extracting structured insights from both free-text and structured responses.

**Supported survey platforms:** PostHog (more coming soon)

**Question types supported:**

Free-text questions:
- Open-ended responses — AI extracts insights, sentiment, and themes
- Comment boxes — follow-up explanations and detailed feedback
- "Why?" questions — reasoning behind ratings or choices

Structured questions:
- NPS (Net Promoter Score) — 0-10 scale
- Rating scales — 1-5 stars, 1-10 scales, emoji ratings
- CSAT (Customer Satisfaction) — satisfaction ratings
- CES (Customer Effort Score) — effort ratings
- Single choice, Multiple choice, Yes/No, Dropdown

**How survey analysis works:**

1. **Automatic import** — syncs survey definitions, questions, and all responses
2. **AI-powered free-text analysis** — extracts insights from substantive feedback (10+ meaningful words), skips generic responses, preserves verbatim quotes, detects sentiment, assigns themes, categorizes feedback, and creates new themes when needed
3. **Structured data processing** — stores values as queryable structured insights, connects to customer accounts for segmentation
4. **Multi-question context** — each answer is processed individually, responses are kept together by respondent

**Privacy and data handling:**
- Survey responses are encrypted in transit and at rest
- Only authorized team members can access survey data
- AI analysis happens in secure, isolated environments
- Verbatim quotes are preserved exactly as written
- Respondent PII is only stored when explicitly collected in surveys

### CSV Import

CSV Import lets you bring historical feedback data into Deck in bulk. Upload a CSV file, map your columns to the right fields, and Deck processes each row through the same AI pipeline used for live integrations.

**Supported feedback types:**

| Type | Required Fields | Optional Fields |
|------|----------------|-----------------|
| **NPS** | `score` (0-10) | `reason`, `respondent_email`, `respondent_name`, `response_date` |
| **User Interview** | `transcript` | `interview_date`, `interviewee_name`, `interviewee_email` |
| **CS Ticket** | `subject`, `body` | `requester_email`, `requester_name`, `created_date`, `status`, `priority` |
| **Feedback** | `content` | `source`, `respondent_email`, `respondent_name`, `submitted_date` |
| **Survey** (pivoted format) | At least one question column | `respondent_email`, `respondent_name` |

**Survey CSV platform detection:**

| Platform | Detection method |
|----------|-----------------|
| **Qualtrics** | 3-row header block |
| **SurveyMonkey** | 2-row header block |
| **Typeform** | Single header row with boolean Yes/No column groups |
| **Google Forms** | Single header row; plain column names |
| **Unknown** | Single header row; generic heuristics |

**Survey question types and credit costs:**

| Type | Credit cost |
|------|-------------|
| NPS (0-10 scale) | 0 |
| Rating Scale (1-5 or 1-7) | 0 |
| Yes / No | 0 |
| Multi-select | 0 |
| Single-select | 0 |
| Free Text | 1 credit per respondent |

**How to import:**

Non-survey types use a 4-step wizard: Upload → Preview → Map → Review.
Survey uploads use a 6-step flow: Upload → Preview → Classify → Question Types → Review → Submit.

1. Prepare your file — export from your existing tool as CSV (max 10 MB)
2. Upload your file — go to **Manage → CSV Uploads** and drop your file
3. Preview — confirm data looks correct
4. Map columns (non-survey) or Classify columns + Question types (survey)
5. Review validation — see valid rows, invalid rows with errors, and duplicates removed
6. Submit — Deck queues the upload and begins processing

**Tracking progress:**

| Status | Meaning |
|--------|---------|
| Pending | Queued, processing not started |
| Processing | Rows being processed by AI pipeline |
| Completed | All rows processed successfully |
| Partially Completed | Some rows succeeded, some failed |
| Failed | Critical failure encountered |

**How Deck processes CSV data:**

1. **Contact matching** — links rows with email addresses to existing contacts
2. **AI analysis** — each row analyzed by the appropriate AI agent
3. **Persistence** — results written alongside your other feedback data
4. **Theme and insight linking** — themes auto-created or matched; insights linked
5. **Cache refresh** — dashboards update to include new data

**Deleting an upload** permanently removes all responses, insights, and theme associations created by that upload. Credits are not refunded. Active uploads (Pending/Processing) cannot be deleted.

---

## Integrations

Deck supports 16 integrations organized by category.

### CRM Integrations

#### Salesforce

Salesforce enables your organization to sync customer Accounts and Contacts directly to insights in Deck. Once connected, Deck automatically enriches your customer feedback data — letting you filter, segment, and prioritize insights by real business metrics.

**Why connect Salesforce:**
- Unify context — tie every insight to the right customer record
- Segment intelligently — filter themes and insights by Salesforce fields like ARR, Lifecycle Stage, Industry, or Churn Risk
- Prioritize with confidence — understand which customers are affected by specific problems
- Report impact faster — show how product decisions relate to revenue, retention, and CRM metrics

**How it works:**
1. Go to **Settings → Integrations → Salesforce** and connect via OAuth
2. Choose which Account and Contact fields to sync (e.g., ARR, Industry, Customer Tier)
3. Deck automatically matches new feedback to customer records and inherits selected CRM properties

Deck does not modify your Salesforce data — it only reads selected fields for enrichment. Data refreshes automatically every 24 hours or on demand. You can map custom field names to internal Deck properties for consistent reporting.

**Example questions Deck can answer with Salesforce:**
- What were our highest-paying customers' pain points last quarter?
- Which churned customers mentioned implementation friction?
- Are expansion-stage accounts requesting specific integrations?
- How do product issues differ between SMB vs. Enterprise customers?

#### HubSpot

HubSpot enables your organization to sync customer Companies and Contacts directly to insights in Deck. Functionally similar to Salesforce — once connected, Deck enriches feedback data for filtering and segmentation.

**How it works:**
1. Go to **Settings → Integrations → HubSpot** and connect via OAuth
2. Select Company and Contact fields to sync (e.g., Annual Revenue, Industry, Lifecycle Stage)
3. Deck automatically matches feedback to customer records

**Filtering insights by HubSpot properties:**
1. Navigate to any theme detail or list page
2. Open HubSpot filters (filter icon → "HubSpot")
3. Choose Company or Contact properties
4. Select criteria: equals, not_equal, contains_token, greater_than, less_than, has_property, not_has_property

Deck never writes to HubSpot — it only reads data. Data refreshes on demand when you apply filters.

#### Attio

Attio CRM integration for syncing customer data and enriching insights with CRM context.

### Customer Support

#### Intercom

Connect Intercom to automatically analyze customer support conversations for product insights using AI.

**What gets synced:**
- Conversations — full message history
- Contacts — customer contact information
- Tags — conversation tags for categorization

**Setup:** Navigate to **Settings → Integrations**, find Intercom, click Connect, and authorize via OAuth. Configure sync frequency (hourly, daily, weekly) and auto-sync settings.

**How it works:**
1. Retrieves conversations based on sync schedule
2. Filters for relevance using AI
3. Analyzes messages — extracts pain points, feature requests, and issues
4. Links to contacts
5. Generates structured insights with quotes, themes, confidence scores, and quality assessments

**AI synthesis pipeline (4 stages):**
1. **Relevance Classification** — filters non-actionable conversations, generates descriptive titles
2. **Ticket Analysis** — understands customer context, frustration levels, key moments
3. **Insight Synthesis** — extracts insights, quotes, and themes using Claude Sonnet
4. **Quality Review** — validates quotes, assigns confidence scores, filters low-quality insights

**First-time sync:** Automatically fetches the last 90 days of conversation history, processes in batches with rate limiting, then syncs incrementally going forward.

**Contact sync:** Optional feature (Starter plan+) that imports full contact details from Intercom (name, email, phone, company) and stores them in your Deck contacts.

**Privacy:** Uses secure OAuth. Email addresses are automatically redacted in the UI (e.g., `j***@example.com`). Data encrypted in transit and at rest.

#### Zendesk

Connect Zendesk to automatically analyze support tickets for customer insights using AI.

**What gets synced:**
- Tickets — full comment history
- Users — customer information
- Tags — ticket tags
- Metadata — priority, status, assignee, and other ticket details

**Setup:** Same flow as Intercom — OAuth connection, sync frequency configuration.

**How it works:** Same 4-stage AI synthesis pipeline as Intercom, applied to support tickets instead of conversations. Automatically generates descriptive subjects for tickets with generic titles.

**First-time sync:** Fetches the last 90 days of ticket history, processes incrementally going forward.

**Contact sync:** Optional (Starter plan+) — imports requester profiles from Zendesk.

**Privacy:** Secure OAuth, email redaction, encrypted data.

### Sales Calls

#### Gong

Connect Gong to automatically analyze sales calls and customer conversations for product insights.

**What gets synced:**
- Calls — recorded sales and customer calls
- Transcripts — automatic transcriptions from Gong

**Setup:** Navigate to **Settings → Integrations**, connect Gong via OAuth, choose which calls to sync.

**How it works:**
1. Fetches call recordings and transcripts from Gong
2. Analyzes conversations to extract product feedback
3. Identifies patterns across sales calls
4. Links to themes to track feature requests and pain points

**Use cases:** Sales call analysis, win/loss insights, feature request tracking, competitive intelligence, objection handling.

### Meetings & Recordings

#### Google Meet

Sync meeting recordings and transcripts from Google Meet for AI analysis.

#### Microsoft Teams

Sync meeting recordings and transcripts from Microsoft Teams for AI analysis.

#### Zoom

Sync meeting recordings and transcripts from Zoom for AI analysis.

### Surveys & Analytics

#### PostHog

Connect PostHog to automatically analyze product survey responses for customer insights.

**What gets synced:**
- Surveys — configurations, questions, and question types
- Responses — individual user responses
- Free-text answers — analyzed by AI for structured insights
- Structured data — ratings, NPS scores, multiple choice, and other structured responses

**Setup:** Connect via OAuth, choose which surveys to sync.

**How it works:**
1. Retrieves survey responses from PostHog
2. Maps question types — identifies free-text, ratings, NPS, single/multiple choice
3. Analyzes free-text feedback — extracts insights, sentiment, themes from open-ended responses
4. Processes structured responses — captures ratings, selections, and numeric data
5. Links to user data — matches responses to customer accounts via email
6. Detects NPS responses — automatically identifies and processes NPS questions

**Automatic NPS detection:**
1. Explicit detection — questions with type "NPS" are automatically recognized
2. Pattern-based detection — rating-scale questions analyzed for NPS indicators in survey name or question text

**Score-only vs. Score + Feedback:**
- With follow-up text — score and feedback analyzed by AI for themes, sentiment, and insights
- Score-only — numeric score recorded deterministically without LLM processing

**Management:** Settings page for connection/sync config, plus a dedicated **Manage → PostHog** page for viewing synced surveys, responses, AI insights, quotes, sentiment, and categories.

#### Typeform

Sync survey responses from Typeform for AI analysis.

### Messaging

#### Slack

Connect Slack to auto-sync customer feedback messages and deliver new insights.

**What you can do once connected:**
- Connect the Deck Bot to Slack channels where you receive customer feedback
- Auto-sync and synthesize those feedback messages into Deck
- Add feedback to new or existing themes
- Visualize insights in the Visualize environment
- Receive daily/weekly insight delivery back to Slack

**Setup:**
1. Go to **Settings → Integrations** and connect Slack via OAuth (requires Slack admin + Deck admin)
2. In your Slack channels, type `/invite Deck` to add the Deck Bot
3. Deck starts syncing and synthesizing customer messages

**Managing insights from Slack:** All synthesized messages appear in the **Manage → Slack** page. View individual channels and all messages for each channel.

### Knowledge & Notes

#### Notion

Connect Notion to import pages, databases, and documents — turning existing research, notes, and feedback into structured insights.

**What you can import:**
- Pages — interview notes, user research documents, meeting summaries
- Database entries — structured feedback logs, NPS responses, support ticket summaries
- Nested pages — browse your full workspace hierarchy and select individual pages

**Import wizard (4 steps):**

1. **Browse** — navigate your Notion workspace, select pages/database entries
2. **Classify** — tell Deck the content type (User Interview, Survey Response, Support Ticket, or Sales Call)
3. **Map Fields** — review how Notion content maps to Deck's data model (participant, date, content)
4. **Review & Import** — preview and confirm; Deck processes through AI synthesis pipeline

**Tips:** Use clear page titles, include speaker labels in notes, import focused documents, batch related content together.

#### Confluence

Import pages from Confluence for AI-powered insight extraction.

#### Fellow

Import meeting notes from Fellow for AI-powered insight extraction.

### Issue Tracking

#### Linear

Sync issues and feedback from Linear for product insight analysis.

---

## NPS Score

Deck automatically detects and analyzes Net Promoter Score (NPS) feedback from your customer conversations, giving you insights into customer satisfaction trends.

### What is NPS?

Net Promoter Score is a customer satisfaction metric measured on a 0-10 scale:

- **Promoters (9-10)**: Loyal enthusiasts who fuel growth
- **Passives (7-8)**: Satisfied but unenthusiastic customers
- **Detractors (0-6)**: Unhappy customers who can damage your brand

NPS = % Promoters - % Detractors (ranging from -100 to +100).

### How It Works

Deck automatically:
1. **Detects NPS feedback** from integrated channels (Slack, surveys, customer conversations)
2. **Extracts scores and reasons** using AI to identify NPS responses in free-form text
3. **Links to themes** by connecting feedback to product themes
4. **Generates insights** with weekly AI synthesis of trends, patterns, and recommendations

### Enabling NPS Analysis

Automatically enabled when Deck first detects NPS feedback. Can be manually toggled in **Settings → Organization → NPS Score Analysis**. When disabled, NPS feedback is processed as regular insights.

### NPS Dashboard

Navigate to **Visualize → NPS Score** to access:

**Summary:**
- Overall NPS score and trend
- Response distribution (Promoters, Passives, Detractors)
- NPS over time chart with rolling 30-day windows
- Key insights on converting Passives and winning back Detractors

**Respondent Analysis:**
- Detailed breakdown by respondent type
- Score distributions within each category
- Trending themes for each group

**Dedicated views** for Detractors, Passives, and Promoters showing percentage trends, top themes, individual responses, and AI-generated insights.

### NPS Over Time Tracking

- Rolling 30-day windows — each data point represents NPS from responses in the prior 30 days
- 7-day intervals — new data points every 7 days
- 3-month history — default view shows last 90 days
- Automatic date handling — uses response dates when available, falls back to submission dates

### NPS Sources

- **PostHog** — automatically sync and analyze NPS survey responses
- **Slack messages** — when customers share NPS scores in synced channels
- **CSV uploads** — import existing NPS survey results
- **Survey tools** — Typeform, SurveyMonkey, Qualtrics, Hotjar, Sprig, Fillout

### Weekly Synthesis

Automatically generated reports including:
- Overall score trends and changes
- Top themes by respondent type
- Insights on what differentiates Promoters from Passives
- Recommendations for addressing Detractor concerns
- Score breakdowns and distributions

---

## Unified Contacts

Deck's unified contacts system automatically consolidates customer feedback from all integrated tools, giving you a complete view of each customer's interactions.

### How It Works

When you integrate tools, Deck automatically:
1. **Identifies customers** across platforms using email, phone, or platform-specific IDs
2. **Creates a unified contact** representing the same person across all tools
3. **Links all feedback** (support tickets, survey responses, interviews, etc.) to that contact

### How Contacts Are Matched

**Email matching:** Automatically normalizes email addresses — handles Gmail dots and plus-signs, Outlook variations. Example: `john.doe+tag@gmail.com` and `johndoe@gmail.com` are recognized as the same person.

**Platform IDs:** Matches using Intercom user IDs, Zendesk requester IDs, Slack user IDs, PostHog distinct IDs, Salesforce contact IDs, HubSpot contact IDs, and Gong IDs.

**Phone numbers:** Used for cross-platform matching when available.

### Contact Information

For each contact, Deck stores:
- Name, Email, Phone, Company
- Avatar from integrated platforms
- First and last seen timestamps
- All platforms where this contact has provided feedback

### Feedback Timeline

Each contact has a unified timeline showing:
- Customer service tickets from Intercom and Zendesk
- NPS responses from surveys
- Survey responses from custom surveys
- User interview insights from uploaded transcripts
- Slack messages from connected channels
- PostHog events from product analytics

### Automatic Backfilling

When you connect a new integration, Deck automatically processes all historical feedback, creates/updates contacts, and links historical data — all in the background.

### Privacy

- Contacts are organization-scoped — your team only sees your organization's contacts
- Contact data is automatically updated with new feedback
- Contact merging is logged for audit purposes

---

## Organization Settings

### Context Configuration

Deck lets you add high-quality organization context so AI Agents can interpret, group, and prioritize feedback in a way that matches your business. Three core types:

- **Company Information** — who you are, how you operate, who you serve
- **Product Information** — what you offer, how it works, how customers use it
- **Keywords** — exact words, phrases, and internal jargon your customers and team use

See the Getting Started section above for detailed best practices on each.

### Billing & Credits

Deck uses a credit-based billing system where AI processing operations consume credits from your monthly allocation.

**Credit types:**
1. **Base Credits** — monthly allocation included with your plan (resets each billing cycle)
2. **Rollover Credits** — unused credits from previous month (expire after 60 days)
3. **Overage Credits** — additional credits when you exceed your allocation (if enabled)

**Viewing your balance:** Navigate to **Settings → Billing** for:
- Current Month tab — credit usage, breakdown, overage settings
- Analytics tab — usage by user and by pipeline
- Manage tab — transaction history, Stripe billing management

**Credit warnings:**
- Yellow alert at less than 20% remaining
- Red alert at less than 10% remaining
- If credits exhausted and overage disabled: AI processing paused, existing data still viewable

**Rollover example:**
- Month 1: 1,000 credits allocated, 800 used → 200 rollover
- Month 2: 1,000 new + 200 rollover = 1,200 total
- Month 3: Unused rollover from Month 1 expires

**Overage:** When enabled, Deck automatically purchases credits beyond your allocation. Charges appear on your next Stripe invoice. You can set a monthly spending cap.

### Security Features

Deck provides multiple layers of security:

**Credential Stuffing Prevention** (all tiers) — monitors login attempts for suspicious patterns, blocks automated credential testing, protects against brute force attacks.

**Bot Signup Detection** (all tiers) — analyzes signup behavior patterns, blocks automated bot signups, prevents spam accounts.

**Passkeys / FIDO2** (Starter tier+) — passwordless authentication using biometrics (fingerprint, Face ID) or hardware security keys. Phishing-resistant, faster than passwords.

**SAML/SSO** (Business tier+) — single sign-on with SAML 2.0 through your identity provider. Supports Okta, Microsoft Entra ID, Google Workspace, OneLogin, JumpCloud, and any SAML 2.0-compliant IdP. Includes force SSO login, JIT provisioning, automated role assignment, and SCIM support.

**Security by tier:**

| Feature | Free | Starter | Business | Enterprise |
|---------|------|---------|----------|------------|
| Credential stuffing prevention | Yes | Yes | Yes | Yes |
| Bot signup detection | Yes | Yes | Yes | Yes |
| Passkeys (FIDO2) | No | Yes | Yes | Yes |
| SAML/SSO | No | No | Yes | Yes |

---

## MCP Server

Deck provides a Model Context Protocol (MCP) server that exposes 16 read-only tools and 8 resources for querying customer feedback from any MCP-compatible AI client.

### What is MCP?

Model Context Protocol is an open standard introduced by Anthropic that enables AI assistants to securely connect to external data sources. MCP provides a universal interface for AI systems to read data, execute functions, and handle contextual prompts. The protocol is managed by the Agentic AI Foundation under the Linux Foundation.

### Supported Clients

Claude Desktop, Claude Code, ChatGPT (Plus/Pro/Business/Enterprise/Education with Developer Mode), Cursor, Visual Studio Code, Windsurf, Zed, Codex (OpenAI), v0 by Vercel, and any generic MCP-compatible client.

### Connection Details

**HTTP Transport (recommended):** `https://mcp.getdeck.io/mcp`

**stdio Transport (via mcp-remote):** `npx -y mcp-remote@latest https://mcp.getdeck.io/mcp`

### Setup

**Step 1:** Organization admins enable MCP access in **Settings → MCP**, toggle on, and select which roles (Admins, Members) can connect.

**Step 2:** Configure your AI assistant using client-specific instructions below.

#### Claude Desktop

Add to `~/Library/Application Support/Claude/claude_desktop_config.json` (macOS) or `%APPDATA%\Claude\claude_desktop_config.json` (Windows):

```json
{
  "mcpServers": {
    "deck": {
      "command": "npx",
      "args": ["-y", "mcp-remote@latest", "https://mcp.getdeck.io/mcp"]
    }
  }
}
```

Restart Claude Desktop, follow the OAuth flow to sign in.

#### Claude Code

```bash
claude mcp add --transport http deck https://mcp.getdeck.io/mcp
```

Authenticate via `/mcp` command. Supports local (default), project, and user scopes.

#### ChatGPT

Enable Developer Mode in settings, go to **Settings → Apps & Connectors**, click Create under Connectors, enter name "Deck Customer Feedback" and URL `https://mcp.getdeck.io/mcp`, select OAuth authentication, complete Clerk OAuth login.

#### Cursor

Quick install link available, or manually: **Settings → Features → MCP → Add new MCP server**, name "Deck", URL `https://mcp.getdeck.io/mcp`.

#### Visual Studio Code

Command Palette → "MCP: Add Server" → Command (stdio) → `npx -y mcp-remote@latest https://mcp.getdeck.io/mcp` → name "Deck".

#### Windsurf

**Settings → Cascade → MCP servers → Add custom server:**
```json
{
  "deck": {
    "command": "npx",
    "args": ["-y", "mcp-remote@latest", "https://mcp.getdeck.io/mcp"]
  }
}
```

#### Zed

Add to Zed settings:
```json
{
  "context_servers": {
    "deck": {
      "command": {
        "path": "npx",
        "args": ["-y", "mcp-remote@latest", "https://mcp.getdeck.io/mcp"]
      }
    }
  }
}
```

#### Codex (OpenAI)

Enable experimental rmcp in `~/.codex/config.toml`, then: `codex mcp add deck --url https://mcp.getdeck.io/mcp`

#### v0 by Vercel

**Settings → Connections → Add Connection**, enter URL `https://mcp.getdeck.io/mcp`.

### Tools Reference (16 tools)

All tools are read-only. Data is automatically scoped to the authenticated user's organization.

#### Insight Tools

**list_insights** — List customer insights with optional filtering.
Parameters:
- `limit` (number, 1-100, default 50) — maximum results
- `offset` (number, >= 0) — pagination offset
- `theme_id` (UUID) — filter by theme
- `category` (enum: PAIN_POINTS, DELIGHTS, GENERAL_FEEDBACK, FEATURE_REQUESTS, USABILITY_ISSUES, BUGS_AND_ERRORS) — filter by category
- `sentiment` (enum: POSITIVE, NEGATIVE, NEUTRAL) — filter by sentiment

**get_insight** — Get detailed information about a specific insight.
Parameters:
- `insight_id` (UUID, required) — the insight ID

**search_insights** — Search insights by keywords across names and quotes.
Parameters:
- `query` (string, 1-500 chars, required) — search query
- `limit` (number, 1-50, default 20) — maximum results

**get_insight_details** — Get full details including all quotes, video snippets, and theme associations.
Parameters:
- `insight_id` (UUID, required) — the insight ID
Returns: All basic fields plus quotes[], snippets[] (with video playback URL, thumbnail, timing), theme_associations[]

**get_new_insights** — Get recently created insights from the last N hours.
Parameters:
- `hours` (number, 1-720, default 24) — hours to look back

#### Theme Tools

**list_themes** — List all themes with names, descriptions, and insight counts.
Parameters:
- `limit` (number, 1-100, default 50)
- `offset` (number, >= 0)

**get_theme** — Get detailed theme information and all associated insights.
Parameters:
- `theme_id` (UUID, required)

**get_theme_insights** — Get all insights for a theme with optional sentiment/category filtering.
Parameters:
- `theme_id` (UUID, required)
- `sentiment` (enum: POSITIVE, NEGATIVE, NEUTRAL)
- `category` (enum: PAIN_POINTS, DELIGHTS, GENERAL_FEEDBACK, FEATURE_REQUESTS, USABILITY_ISSUES, BUGS_AND_ERRORS)

**get_theme_stats** — Get sentiment and category breakdown statistics for a theme.
Parameters:
- `theme_id` (UUID, required)
Returns: total_insights, sentiment_breakdown (positive, negative, neutral counts), category_breakdown (counts per category)

#### Feedback Tools

**list_feedback** — List feedback sources (interviews, calls, surveys).
Parameters:
- `limit` (number, 1-100, default 50)
- `offset` (number, >= 0)
- `content_type` (enum: USER_INTERVIEW, SALES_CALL, SUPPORT_TICKET, SURVEY_RESPONSE, APP_REVIEW)

**get_feedback** — Get detailed feedback item including full transcript.
Parameters:
- `feedback_id` (UUID, required)

**list_interviews** — List user interviews with status filtering.
Parameters:
- `page` (number, >= 1, default 1)
- `limit` (number, 1-100, default 50)
- `status` (enum: IDLE, UPLOADING, QUEUED, PROCESSING, COMPLETE, FAILED, CREDIT_BLOCKED)

**get_interview_transcript** — Get full transcript with speaker segments.
Parameters:
- `feedback_id` (UUID, required)
Returns: feedback_id, file_name, content, speaker_segments[]

**get_interview_insights** — Get all insights extracted from a specific interview.
Parameters:
- `feedback_id` (UUID, required)

#### Organization Tools

**get_org_information** — Get organization context (company info, product details, target audience, keywords).
Parameters: None (uses authenticated organization)

**get_platform_links** — Get URL links to view resources in the Deck web app.
Parameters:
- `resources` (array of objects, 1-20 items, required)
  - `resource_type` (enum: theme, insight, interview, required)
  - `resource_id` (UUID, optional — omit for list views)

### Resources Reference (8 resources)

#### Static Resources

**deck://categories** — List of the 6 fixed insight categories with descriptions (application/json)

**deck://integrations** — Available integration types for connecting third-party services (application/json)

#### Dynamic Resources (organization-scoped)

**deck://themes** — List all themes in the organization (application/json)

**deck://themes/{theme_id}** — Get a specific theme with its insights (application/json)

**deck://insights** — List customer insights in the organization (application/json)

**deck://insights/{insight_id}** — Get a specific insight with full details including quotes, snippets, and theme associations (application/json)

**deck://transcripts** — List interview transcripts (only completed interviews) (application/json)

**deck://transcripts/{feedback_id}** — Get a specific transcript with content and speaker segments (application/json)

### Security

- **OAuth 2.0 / 2.1** — all connections use Clerk OAuth for secure authentication
- **PKCE Support** — prevents authorization code interception attacks
- **Dynamic Client Registration** — supports RFC 7591 for clients like ChatGPT
- **Read-only access** — all data access through MCP is read-only
- **Organization scoping** — connections automatically scoped to your organization
- **Role-based access** — admins control which roles can connect
- **Revocable** — sign out or revoke access at any time; admins can disable MCP instantly

### Usage Examples

Discovery & overview:
- "Give me a summary of all customer feedback themes"
- "What are the top 5 most mentioned issues this quarter?"

Sentiment analysis:
- "Show me all negative sentiment insights from last month"
- "Which themes have the most negative sentiment?"

Deep dive research:
- "Search for insights related to pricing concerns"
- "Show me the full transcript from the most recent user interview"

Product planning:
- "What features are customers requesting most?"
- "Identify the top 3 issues we should prioritize fixing"

---

## AI Processing Pipelines

Deck runs 5 specialized AI processing pipelines:

### 1. User Interview Synthesis

A 5-agent pipeline that processes recorded interviews and transcripts:

1. **Transcript Validation** — validates transcript quality and identifies speakers
2. **Discussion Guide Inference** — infers the interview's research goals and discussion topics
3. **Insight Extraction** — identifies key themes, pain points, and feature requests from the conversation
4. **Quote Verification** — maps verbatim quotes to each extracted insight, ensuring accuracy
5. **Quality Review** — assigns confidence scores, filters low-quality insights, produces final summary

Processing time: ~5 minutes for video/audio, ~2 minutes for transcripts.

### 2. Customer Service Synthesis

A 4-stage pipeline for support tickets (Intercom conversations, Zendesk tickets):

1. **Relevance Classification** — filters non-actionable conversations (greetings, simple questions, status updates) and generates descriptive titles/subjects
2. **Ticket Analysis** — understands customer context, frustration levels, and key moments
3. **Insight Synthesis** — extracts insights, quotes, and themes using Claude Sonnet
4. **Quality Review** — validates quotes against original messages, assigns confidence scores, filters low-quality insights

Only high-quality insights with verified customer quotes are surfaced.

### 3. NPS Processing

Processes Net Promoter Score feedback from multiple sources:

1. **Detection** — identifies NPS responses in Slack messages, surveys, and direct imports using explicit type detection and smart pattern recognition
2. **Score Extraction** — captures 0-10 scores, classifies as Promoter/Passive/Detractor
3. **Sentiment Analysis** — analyzes follow-up text for themes and sentiment
4. **Weekly Trend Synthesis** — generates weekly reports with score trends, theme breakdowns, and recommendations

### 4. Survey Response Analysis

Processes survey responses from PostHog and CSV imports:

1. **Question Type Mapping** — identifies free-text, ratings, NPS, single/multiple choice questions
2. **Free-Text Analysis** — extracts insights from substantive responses (10+ meaningful words), skips generic answers
3. **Theme Classification** — assigns responses to existing themes or creates new ones
4. **Batch Processing with Confidence Scoring** — processes responses in batches, assigns confidence scores to each insight

Structured answer types (everything except free text) are stored directly without LLM processing and cost 0 credits.

### 5. Cross-Source Feedback Synthesis

Weekly pattern detection across all feedback sources:

1. **Aggregation** — collects insights from all sources (interviews, tickets, surveys, Slack, NPS)
2. **Pattern Detection** — identifies recurring themes and trends across sources
3. **Synthesis** — generates a comprehensive weekly summary
4. **Slack Delivery** — delivers the synthesis report to configured Slack channels

---

## Pricing

For the canonical agent-readable pricing brief, see [Deck Pricing](https://getdeck.io/pricing.md).

### Plan Comparison

| Feature | Free | Starter | Business | Enterprise |
|---------|------|---------|----------|------------|
| **Monthly price** | $0 | $99/mo ($79/mo annual) | $249/mo ($199/mo annual) | Custom |
| **AI credits/month** | 200 | 1,000 | 5,000 | 25,000 |
| **Overage rate** | N/A | $0.05/credit | $0.04/credit | $0.03/credit |
| **Integrations** | 2 | 5 | Unlimited | Unlimited |
| **Slack channels** | 1 | 3 | 10 | Unlimited |
| **CRM integration** | No | No | Yes | Yes |
| **NPS Analytics** | No | Yes | Yes | Yes |
| **Passkey auth** | No | Yes | Yes | Yes |
| **SAML/SSO** | No | No | Yes | Yes |
| **Credit rollover** | Yes (60 days) | Yes (60 days) | Yes (60 days) | Yes (60 days) |

### How Credits Work

Credits are the currency for AI processing in Deck. Each AI operation consumes credits based on complexity:
- Interview synthesis, feedback analysis, NPS detection, survey processing, company research

**Credit sources:**
1. Base credits — monthly allocation, resets each billing cycle
2. Rollover credits — unused from previous month, expire after 60 days
3. Overage credits — additional when you exceed allocation (if enabled)

### Downgrades

Downgrades take effect at the start of your next billing cycle, not immediately. You keep your current plan's features and limits through the end of the paid period. At renewal, the lower tier's entitlements are applied.

Existing resources over the new limit are not removed. Example: 8 integrations on Business → downgrade to Starter (5 max) → at renewal, all 8 stay connected, but you cannot add new ones until under the limit.

Cancellation schedules the subscription to end at period close. When the period ends, the org reverts to the Free tier.

---

## Security & Compliance

- **GDPR compliant** — full compliance with EU data protection regulations
- **SOC Type 2 infrastructure partners** — Clerk (auth), Neon (database)
- **Customer data never used to train AI models** — your feedback data is processed for insights only
- **Multi-region database isolation** — choose US (NA), EU (EMEA), or Australia (APAC) for data residency
- **Enterprise-grade authentication** — passkeys (FIDO2), SAML/SSO, credential stuffing prevention, bot detection
- **Service-to-service authentication** — short-lived JWTs (HS256) with shared secrets
- **Encryption** — all data encrypted in transit and at rest
- **Read-only integrations** — Deck never writes to your CRM, support tools, or external services
- **Organization-scoped data** — complete data isolation between organizations
- **OAuth authentication** — secure OAuth for all third-party integrations via Nango
- **Email redaction** — customer emails automatically redacted in support ticket views

---

## Links

- [Deck App](https://app.getdeck.io): Sign in to the Deck platform
- [Documentation](https://docs.getdeck.io): Full product documentation
- [MCP Server Docs](https://docs.getdeck.io/docs/integrations/mcp): Setup guide for connecting AI clients to Deck
- [Integrations](https://getdeck.io/product/integrations): Browse all 16 supported integrations
- [Pricing](https://getdeck.io/pricing): Compare plans and pricing
- [Agent-readable Pricing](https://getdeck.io/pricing.md): Plain Markdown pricing for AI agents
- [How Deck Works](https://getdeck.io/how-deck-works.md): Plain Markdown overview of Deck for customers and AI agents
- [Security](https://getdeck.io/security): Security practices and compliance

## Optional

- [Marketing Website](https://getdeck.io): Deck homepage and product overview
- [Sign Up](https://app.getdeck.io/sign-up): Create a free account