Skip to content

PRD.md

Latest commit

 

History

History
818 lines (631 loc) · 26.5 KB

PRD.md

File metadata and controls

818 lines (631 loc) · 26.5 KB
Raw

TrustBlade - Product Requirements Document

Document Version: 1.0
Last Updated: February 4, 2026
Author: Product Team
Status: Draft

Executive Summary

TrustBlade is a credibility scoring platform for social media users, functioning as "SocialBlade for truthfulness." The platform tracks the factual accuracy of Twitter/X users over time, building cumulative trust scores based on AI-powered fact-checking. Users can look up any Twitter/X profile to see their credibility rating, historical accuracy trends, and verified claims.

Target Launch: Q2 2026
Domain: trustblade.xyz

Table of Contents

  1. Problem Statement
  2. Solution Overview
  3. Target Users
  4. Product Goals & Success Metrics
  5. Competitive Analysis
  6. Feature Requirements
  7. Technical Architecture
  8. Trust Score Algorithm
  9. Data Collection Strategy
  10. User Experience
  11. Monetization
  12. Launch Roadmap
  13. Risks & Mitigations
  14. Appendix

1. Problem Statement

The Challenge

Misinformation spreads 6x faster than accurate information on social media. Users currently have no easy way to assess the credibility of accounts before engaging with or sharing their content. Existing fact-checking solutions are:

  • Reactive: Community Notes and fact-checkers respond after content goes viral
  • Per-tweet: No cumulative tracking of user accuracy over time
  • Fragmented: No centralized credibility database
  • Inaccessible: Requires users to actively search for fact-checks

User Pain Points

User Type Pain Point
General Public Cannot quickly assess if an account is trustworthy
Journalists Need to verify sources before citing social media
Researchers Lack quantitative credibility metrics for studies
Advertisers Risk brand safety when partnering with influencers
Platforms Struggle to surface authoritative voices

Market Opportunity

  • 500M+ daily active users on Twitter/X
  • 70% of users have shared content they later learned was false
  • $2B+ spent annually on misinformation research and mitigation
  • No direct competitor offering profile-level credibility scores

2. Solution Overview

TrustBlade provides a cumulative credibility score (0-100) for Twitter/X users based on the historical accuracy of their factual claims.

Core Value Proposition

"Know who to trust before you share."

Key Differentiators

Feature TrustBlade NewsGuard Bot Sentinel Community Notes
Profile-level scoring ✅ ❌ (domains only) ❌ (bot detection) ❌
Historical tracking ✅ ❌ ❌ ❌
Topic breakdown ✅ ❌ ❌ ❌
Real-time integration ✅ (extension) ❌ ✅ ✅
Public API ✅ ❌ ❌ ❌
AI-powered ✅ ❌ (human reviewers) ✅ ❌ (crowdsourced)

Product Components

  1. Web Application - Search, profile pages, leaderboards, trends
  2. Chrome Extension - Trust badges on Twitter/X, one-click fact-check
  3. Public API - Integrate trust scores into third-party applications
  4. Embed Widget - Display trust scores on external websites

3. Target Users

Primary Users

3.1 Information-Conscious Consumers

  • Demographics: 25-54, college-educated, news consumers
  • Behavior: Active on social media, shares content, skeptical of sources
  • Need: Quick credibility assessment before engaging/sharing
  • Value: Free web lookup, Chrome extension

3.2 Journalists & Researchers

  • Demographics: Media professionals, academics, think tanks
  • Behavior: Cite social media sources, require verification
  • Need: Quantitative credibility metrics, historical data
  • Value: API access, exportable reports, detailed claim history

3.3 Brand Safety Teams

  • Demographics: Marketing agencies, PR firms, advertisers
  • Behavior: Vet influencers before partnerships
  • Need: Risk assessment, bulk profile analysis
  • Value: Team plans, enterprise API, custom reports

Secondary Users

  • Platform Trust & Safety Teams - API integration for content moderation
  • Educators - Media literacy curriculum integration
  • Browser Extension Users - Passive credibility awareness

User Personas

Persona 1: "Skeptical Sarah"

  • 34-year-old marketing manager
  • Sees viral tweets, wants to verify before sharing
  • Uses extension to see trust badges while browsing
  • Free tier user

Persona 2: "Journalist Jake"

  • 42-year-old investigative reporter
  • Needs to cite social media sources in articles
  • Uses API to pull credibility data for stories
  • Pro tier subscriber

Persona 3: "Brand Manager Bella"

  • 38-year-old at advertising agency
  • Vets influencers before brand partnerships
  • Needs bulk analysis and exportable reports
  • Enterprise tier customer

4. Product Goals & Success Metrics

Business Goals

Goal Target Timeframe
Indexed profiles 100,000 Month 3
Monthly active users 50,000 Month 6
Chrome extension installs 25,000 Month 6
Paying subscribers 1,000 Month 6
Monthly recurring revenue $15,000 Month 6

Product Goals

Goal Metric Target
Accuracy AI fact-check agreement with human reviewers >85%
Coverage % of top 10K accounts indexed 100%
Freshness Average time to index new claims <24 hours
Engagement Claims fact-checked per user/month 10+

North Star Metric

Claims Verified Per Day - Measures both user engagement and data flywheel growth.

Key Performance Indicators (KPIs)

Acquisition

  • New user signups (weekly)
  • Extension installs (weekly)
  • Organic search traffic

Activation

  • First profile lookup within 24 hours
  • Extension badge views per session

Retention

  • Weekly active users
  • Extension 7-day retention

Revenue

  • Conversion rate (free → paid)
  • Average revenue per user (ARPU)
  • Monthly recurring revenue (MRR)

Referral

  • Shared profiles / embeds
  • API integrations

5. Competitive Analysis

Direct Competitors

None identified - No existing product offers profile-level credibility scoring with historical tracking.

Adjacent Competitors

Product Focus Limitation
NewsGuard Website/domain ratings Does not rate individuals
Bot Sentinel Bot/troll detection Binary classification, not accuracy
Community Notes Crowdsourced fact-checks Per-tweet, reactive, slow
Snopes/PolitiFact Article fact-checks Manual, limited coverage
SocialBlade Follower/engagement stats No credibility metrics

Competitive Advantages

  1. First mover in profile-level credibility scoring
  2. AI-powered for scale (vs. manual fact-checkers)
  3. Historical tracking enables trend analysis
  4. Chrome extension for frictionless integration
  5. Public API enables ecosystem development

Barriers to Entry

  • Data moat (historical claim database)
  • AI model training on fact-checking patterns
  • Extension user network effects
  • API integration stickiness

6. Feature Requirements

6.1 Web Application

P0 (Must Have - MVP)

Feature Description Acceptance Criteria
Profile Search Search by @username Returns profile within 2 seconds
Profile Page Trust score, grade, claim history Displays score 0-100, letter grade A-F
Trust Score Display Visual score indicator Animated ring, color-coded
Claim Feed List of verified claims Shows verdict, confidence, date
Basic Leaderboard Top 100 most accurate Sortable by score

P1 (Should Have - V1.1)

Feature Description Acceptance Criteria
Topic Breakdown Accuracy by topic Shows scores for politics, tech, health, etc.
Score History Chart Trust score over time Line chart with 30/90/365 day views
Profile Comparison Compare 2-3 profiles Side-by-side score comparison
Advanced Filters Filter leaderboards By category, time period, topic
Embed Widget Embeddable trust badge iframe/JS snippet for external sites

P2 (Nice to Have - V1.2)

Feature Description
Trending Claims Viral claims currently being fact-checked
Correction Tracking Flag when users correct false claims
Network Analysis Credibility of who users retweet/quote
Alert System Notifications when tracked profiles post
User Accounts Save favorites, custom watchlists

6.2 Chrome Extension

P0 (Must Have - MVP)

Feature Description Acceptance Criteria
Trust Badge Injection Badge next to usernames Appears within 1 second of page load
Profile Page Badge Large badge on profile pages Shows score, grade, click to open
Popup UI Extension popup with search Search any username

P1 (Should Have - V1.1)

Feature Description
One-Click Fact Check Button to fact-check any tweet
Inline Results Show fact-check results in popup
Data Collection Passive tweet collection for database
Settings Toggle features on/off

P2 (Nice to Have - V1.2)

Feature Description
Highlight Unverified Visual indicator for unindexed accounts
Context Menu Right-click to fact-check selected text
Firefox Support Port to Firefox extension

6.3 Public API

P0 (Must Have - MVP)

Endpoint Method Description
/api/profile/:username GET Get profile trust score
/api/search GET Search profiles
/api/fact-check POST Fact-check a tweet

P1 (Should Have - V1.1)

Endpoint Method Description
/api/profile/:username/claims GET Get claim history
/api/profile/:username/history GET Get score history
/api/leaderboard GET Get leaderboard data
/api/claims/recent GET Get recent fact-checks

P2 (Nice to Have - V1.2)

Endpoint Method Description
/api/batch POST Bulk profile lookup
/api/compare GET Compare multiple profiles
/api/webhooks POST Register webhooks for updates

6.4 Non-Functional Requirements

Requirement Target
Page Load Time <2 seconds (P95)
API Response Time <500ms (P95)
Uptime 99.9%
Extension Performance <100ms badge injection
Mobile Responsive Full functionality on mobile web
Accessibility WCAG 2.1 AA compliance
Security SOC 2 Type II (future)

7. Technical Architecture

System Architecture

┌─────────────────────────────────────────────────────────────┐
│                        CLIENTS                              │
├──────────────────┬──────────────────┬───────────────────────┤
│   Web App        │  Chrome Extension │   Third-Party Apps   │
│   (Next.js)      │  (Manifest V3)    │   (API Consumers)    │
└────────┬─────────┴────────┬─────────┴───────────┬───────────┘
         │                  │                     │
         └──────────────────┼─────────────────────┘
                            │
                   ┌────────▼────────┐
                   │   API Gateway   │
                   │   (Vercel)      │
                   └────────┬────────┘
                            │
         ┌──────────────────┼──────────────────┐
         │                  │                  │
┌────────▼────────┐ ┌───────▼───────┐ ┌───────▼───────┐
│   Next.js API   │ │  Rate Limiter │ │  Auth Service │
│   Routes        │ │  (Upstash)    │ │  (Supabase)   │
└────────┬────────┘ └───────────────┘ └───────────────┘
         │
         ├─────────────────────────────────────┐
         │                                     │
┌────────▼────────┐               ┌────────────▼────────────┐
│    Supabase     │               │     Google Gemini       │
│   PostgreSQL    │               │   2.0 Flash + Search    │
│                 │               │   (Fact-Checking AI)    │
│  - profiles     │               └─────────────────────────┘
│  - claims       │
│  - tweets       │
│  - sources      │
│  - api_keys     │
└─────────────────┘

Technology Stack

Layer Technology Rationale
Frontend Next.js 14 (App Router) SSR, API routes, React ecosystem
Styling Tailwind CSS Rapid development, consistent design
Database Supabase (PostgreSQL) Managed Postgres, real-time, auth
AI/ML Google Gemini 2.0 Flash Built-in web search, fast, cost-effective
Hosting Vercel Seamless Next.js deployment, edge functions
Extension Chrome Manifest V3 Modern extension standard
Analytics PostHog Open-source, self-hostable
Monitoring Sentry Error tracking, performance monitoring

Database Schema (Simplified)

profiles
├── id (UUID, PK)
├── username (VARCHAR, UNIQUE)
├── trust_score (DECIMAL)
├── trust_grade (VARCHAR)
├── total_claims (INT)
├── true_count (INT)
├── false_count (INT)
├── misleading_count (INT)
└── timestamps

claims
├── id (UUID, PK)
├── profile_id (FK → profiles)
├── tweet_id (FK → tweets)
├── claim_text (TEXT)
├── verdict (ENUM: TRUE, FALSE, MISLEADING, UNVERIFIABLE, OPINION)
├── confidence (DECIMAL)
├── explanation (TEXT)
└── timestamps

tweets
├── id (UUID, PK)
├── tweet_id (VARCHAR, UNIQUE)
├── profile_id (FK → profiles)
├── text (TEXT)
├── engagement metrics
└── timestamps

sources
├── id (UUID, PK)
├── claim_id (FK → claims)
├── url (TEXT)
├── title (VARCHAR)
├── supports_claim (BOOLEAN)
└── timestamps

Infrastructure Costs (Estimated Monthly)

Service Tier Cost
Vercel Pro $20
Supabase Pro $25
Gemini API Pay-as-you-go $50-100
Upstash Redis Pay-as-you-go $10
Domain Annual $15/yr
Total ~$100-150/mo

8. Trust Score Algorithm

Formula

Trust Score = (Weighted Score + 2 × Total Claims) / (3 × Total Claims) × 100

Adjusted for confidence based on sample size.

Verdict Weights

Verdict Weight Rationale
TRUE +1.0 Reward for accuracy
FALSE -2.0 Penalize misinformation heavily
MISLEADING -1.0 Partial penalty for deception
UNVERIFIABLE 0.0 Neutral - cannot confirm/deny
OPINION 0.0 Not factual, excluded from scoring

Confidence Adjustment

Confidence = min(Total Claims / 50, 1.0)
Final Score = (Raw Score × Confidence) + (50 × (1 - Confidence))

New profiles with few claims regress toward 50 (neutral) until sufficient data is collected.

Grade Scale

Score Range Grade Label
95-100 A+ Highly Reliable
90-94 A Very Accurate
85-89 A- Mostly Accurate
80-84 B+ Generally Reliable
75-79 B Above Average
70-74 B- Fairly Accurate
65-69 C+ Mixed Record
60-64 C Average
55-59 C- Below Average
45-54 D Often Inaccurate
0-44 F Frequently False

Future Enhancements (V2)

  • Recency weighting: Recent claims weighted more heavily
  • Virality multiplier: High-engagement false claims penalized more
  • Topic expertise: Separate scores per topic
  • Correction bonus: Reward users who correct mistakes
  • Source reliability: Weight claims by source quality

9. Data Collection Strategy

Data Sources

Source Method Volume Priority
Extension Users Passive collection while browsing High P0
On-Demand Lookup User searches trigger scraping Medium P0
Priority Queue Automated crawling of top accounts High P1
API Partners Data sharing agreements Variable P2

Priority Crawl List

  1. Politicians (federal, state, international)
  2. Major news organizations
  3. Journalists with 100K+ followers
  4. Public health officials
  5. Tech industry leaders
  6. Influencers with 1M+ followers
  7. Frequently searched profiles

Data Processing Pipeline

Tweet Collected
      │
      ▼
┌─────────────────┐
│ Claim Extraction│ ◄── Gemini: Extract verifiable factual claims
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│  Fact-Checking  │ ◄── Gemini + Web Search: Verify claims
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│ Score Update    │ ◄── Recalculate profile trust score
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│ Store Results   │ ◄── Save to database
└─────────────────┘

Data Retention

Data Type Retention Period
Profile scores Indefinite
Claims Indefinite
Tweets 2 years
Sources 2 years
API logs 90 days
User analytics 1 year

10. User Experience

Information Architecture

trustblade.xyz/
├── / (Home - Search, Featured, Recent)
├── /profile/:username (Profile Page)
│   ├── /claims (Full Claim History)
│   └── /history (Score History)
├── /leaderboards
│   ├── /accurate (Most Accurate)
│   ├── /improved (Most Improved)
│   └── /:category (By Category)
├── /trends (Trending Claims)
├── /compare (Profile Comparison)
├── /api-docs (API Documentation)
├── /extension (Download Extension)
├── /about (About & Methodology)
└── /pricing (Plans & Pricing)

Key User Flows

Flow 1: Quick Lookup (Primary)

User lands on homepage
        │
        ▼
Enters @username in search bar
        │
        ▼
Views profile page with trust score
        │
        ▼
Scrolls to see recent claims
        │
        ▼
Clicks claim to see sources/explanation

Flow 2: Extension Usage

User installs Chrome extension
        │
        ▼
Browses Twitter/X as normal
        │
        ▼
Sees trust badges next to usernames
        │
        ▼
Hovers for quick score preview
        │
        ▼
Clicks to open full profile on TrustBlade

Flow 3: Fact-Check Request

User sees questionable tweet
        │
        ▼
Clicks "Fact Check" button (extension)
        │
        ▼
Views loading state (2-5 seconds)
        │
        ▼
Sees verdict, confidence, sources
        │
        ▼
Shares result or views full analysis

Design Principles

  1. Clarity over complexity - Trust scores should be instantly understandable
  2. Speed - Results within 2 seconds for cached, 5 seconds for new
  3. Transparency - Always show methodology, sources, confidence
  4. Neutrality - No political bias in presentation
  5. Accessibility - Color-blind friendly, screen reader compatible

Brand Guidelines

Element Specification
Primary Color Purple (#8b5cf6) - "Blade"
Success/Trust Green (#22c55e)
Danger/False Red (#ef4444)
Warning/Misleading Yellow (#f59e0b)
Typography Inter (sans-serif)
Logo Shield icon + "TrustBlade" wordmark

11. Monetization

Pricing Tiers

Tier Price Features
Free $0 10 lookups/day, extension access
Pro $9/mo Unlimited lookups, 1K API calls/mo, export data
Team $29/mo 5 seats, 10K API calls/mo, priority support
Enterprise Custom Unlimited API, SLA, dedicated support, custom integrations

Revenue Projections (Year 1)

Month Free Users Pro Team Enterprise MRR
3 10,000 100 10 0 $1,190
6 50,000 500 50 2 $7,450
12 150,000 2,000 200 10 $29,800

Additional Revenue Streams (Future)

  • Data licensing - Sell anonymized credibility data to researchers
  • Brand safety reports - Custom reports for advertising agencies
  • White-label solution - License technology to platforms
  • Premium API - Higher rate limits, bulk endpoints, webhooks

12. Launch Roadmap

Phase 1: Foundation (Weeks 1-2)

  • Database schema design
  • Core API routes
  • Trust score calculation
  • Gemini AI integration
  • Basic web UI (home, profile, search)

Phase 2: Extension (Weeks 3-4)

  • Chrome extension MVP
  • Trust badge injection
  • Extension popup UI
  • Passive data collection
  • Extension settings

Phase 3: Polish (Weeks 5-6)

  • Leaderboards
  • Profile comparison
  • Score history charts
  • Topic breakdown
  • Embed widget

Phase 4: Launch (Weeks 7-8)

  • Chrome Web Store submission
  • Product Hunt launch
  • Press outreach
  • Social media campaign
  • API documentation

Post-Launch (Month 3+)

  • Firefox extension
  • Mobile app (React Native)
  • Advanced analytics
  • API SDK (Python, JavaScript)
  • Enterprise features

Milestones

Milestone Target Date Success Criteria
Alpha Week 4 Core features working, 10 beta testers
Beta Week 6 Extension in Chrome Web Store (unlisted)
Public Launch Week 8 Product Hunt launch, press coverage
10K Users Month 3 Monthly active users
$10K MRR Month 6 Sustainable revenue

13. Risks & Mitigations

Technical Risks

Risk Likelihood Impact Mitigation
AI fact-check errors High High Human review queue, confidence thresholds, appeal process
Twitter API changes Medium High Diversify data sources, extension-first collection
Scale issues Medium Medium Caching, database optimization, CDN
Extension rejection Low High Follow Chrome policies strictly, privacy-first design

Business Risks

Risk Likelihood Impact Mitigation
Legal challenges Medium High Terms of service, disclaimers, legal review
Competitor entry Medium Medium First-mover advantage, data moat, brand building
User trust issues Medium High Transparency, methodology documentation, corrections
Low adoption Medium Medium SEO, influencer partnerships, viral features

Ethical Risks

Risk Likelihood Impact Mitigation
Political bias perception High High Transparent methodology, balanced coverage, advisory board
Gaming/manipulation Medium High Rate limiting, anomaly detection, human oversight
Harassment enablement Low Medium No tools for brigading, report functionality
False positives harm Medium High Appeal process, clear disclaimers, correction tracking

14. Appendix

A. Glossary

Term Definition
Trust Score 0-100 rating of a profile's factual accuracy
Claim A verifiable factual statement extracted from a tweet
Verdict The fact-check result: TRUE, FALSE, MISLEADING, UNVERIFIABLE, OPINION
Confidence AI's certainty in its verdict (0-100%)
Indexing Process of adding a profile to the TrustBlade database

B. User Research Questions

  1. How do you currently verify information on social media?
  2. Would you use a browser extension that shows credibility scores?
  3. What would make you trust a credibility rating system?
  4. How much would you pay for unlimited access to credibility data?
  5. What features would make this product essential for you?

C. Success Stories (Target)

"TrustBlade saved me from sharing a viral hoax. Now I check every account before retweeting." - User testimonial

"We use TrustBlade's API to vet influencers before brand partnerships. It's reduced our risk significantly." - Agency testimonial

"Finally, a quantitative measure of credibility I can cite in my research." - Academic testimonial

D. References

Document History

Version Date Author Changes
1.0 Feb 4, 2026 Product Team Initial draft

Approvals

Role Name Date Signature
Product Lead
Engineering Lead
Design Lead
Executive Sponsor