Links (A/B Testing)

Links in Fly Super are intelligent routing tools that distribute traffic across multiple destinations based on configurable weights. Perfect for A/B testing landing pages, gradual rollouts, and traffic splitting.

What Are Links?

A link is a short URL that redirects visitors to one of several destination URLs (called "variants") based on weight distribution. Each visitor is consistently assigned to the same variant.

Example:

Link: yourdomain.com/special-offer

Redirects to:
- 50% → yourdomain.com/offer-page-v1 (Control)
- 50% → yourdomain.com/offer-page-v2 (Variant)

Key Features:

• Multiple variants per link (2-10 URLs)
• Weighted distribution (e.g., 70/30, 50/50, 33/33/34)
• Consistent visitor routing (same user always sees same variant)
• Edge-based routing (zero latency, ~10ms redirects)
• Cookie persistence (variant assignment stored)
• Redis-backed for speed

Use Cases

1. A/B Testing Landing Pages

Test different versions to find highest-converting design:

Example:

• Link: yourdomain.com/product-launch
• Variant A (50%): Long-form sales page
• Variant B (50%): Short-form minimal page

Measure: Which variant drives more conversions?

2. Headline Testing

Test different value propositions:

Example:

• Link: yourdomain.com/webinar
• Variant A (50%): "Learn How to 10x Your Revenue"
• Variant B (50%): "Free Masterclass: Revenue Growth Strategies"

Measure: Which headline gets more signups?

3. Gradual Rollouts

Release new page version gradually:

Example:

• Link: yourdomain.com/pricing
• Variant A (90%): Current pricing page
• Variant B (10%): New pricing page

Gradually shift traffic: 90/10 → 75/25 → 50/50 → 0/100

4. Geographic Routing

Send users to region-specific pages:

Example:

• Link: yourdomain.com/signup
• Variant A (40%): US-specific page
• Variant B (30%): EU-specific page
• Variant C (30%): Asia-specific page

Note: Currently random distribution. True geo-routing coming soon.

5. Multi-Variant Testing

Test more than 2 options:

Example:

• Link: yourdomain.com/offer
• Variant A (33%): $99/month offer
• Variant B (33%): $199/year offer
• Variant C (34%): Custom pricing offer

Measure: Which offer generates most revenue?

6. Campaign Testing

Test different campaign messaging:

Example:

• Link: yourdomain.com/campaign
• Variant A (50%): Problem-focused messaging
• Variant B (50%): Solution-focused messaging

Share link in ads, measure which messaging resonates.

How Links Work

1. Visitor Clicks Link

User clicks: yourdomain.com/special-offer

2. Edge Routing

• Request hits Fly Super edge network
• Redis lookup for link configuration (< 10ms)
• No database query (blazing fast)

3. Variant Assignment

First-time visitor:

• Deterministic hash based on IP + User Agent
• SHA-256 hash → mod by total weight → select variant
• Cookie set: fs_variant_[linkId]=variantId

Returning visitor:

• Check cookie for existing assignment
• Route to same variant consistently

4. Redirect

• HTTP 307 Temporary Redirect
• Visitor lands on variant URL
• Entire process: ~10ms

Why 307?

• Temporary redirect (search engines don't index)
• Preserves HTTP method (POST requests work)
• Allows changing destinations

Key Concepts

Variants

A variant is a destination URL with a weight:

{
  "url": "https://yourdomain.com/page-v1",
  "weight": 50
}

Each link can have 2-10 variants.

Weights

Weight determines traffic percentage:

Examples:

50/50 Split:

• Variant A: weight 50 (50% traffic)
• Variant B: weight 50 (50% traffic)

70/30 Split:

• Variant A: weight 70 (70% traffic)
• Variant B: weight 30 (30% traffic)

33/33/34 Split (3 variants):

• Variant A: weight 33 (33% traffic)
• Variant B: weight 33 (33% traffic)
• Variant C: weight 34 (34% traffic)

Important: Weights must sum to 100!

Deterministic Routing

Same visitor always sees same variant:

How:

1. Hash visitor's IP + User Agent with SHA-256
2. Take hash mod 100
3. Map result to weight ranges
4. Store assignment in cookie

Result:

• User sees consistent experience
• No flip-flopping between variants
• Accurate test results

Cookie Persistence

Variant assignment stored in cookie:

Cookie name: fs_variant_[linkId]

Cookie value: variantId

Duration: 30 days (default)

Purpose:

• Fast lookups (no hashing needed)
• Cross-session consistency
• Works even if IP changes

Creating Links

Quick Start

1. Navigate to Links in sidebar
2. Click "Create Link"
3. Enter link slug (e.g., special-offer)
4. Add 2+ variants with URLs and weights
5. Ensure weights sum to 100
6. Click "Create"

Your link is now live! Share: yourdomain.com/slug

Detailed instructions: See Creating Links

Managing Links

Viewing Links

Links list shows:

• Link slug
• Domain
• Number of variants
• Weight distribution
• Created date
• Active/inactive status

Editing Links

Click a link to edit:

• Change variant URLs
• Adjust weights
• Add/remove variants
• Update slug

Note: Changes apply immediately to new visitors. Existing cookie assignments persist until expiry.

Deactivating Links

To stop using a link:

1. Open link settings
2. Click "Deactivate"
3. Link returns 404

Use case: End A/B test after winner determined.

Deleting Links

To permanently delete:

1. Open link settings
2. Click "Delete"
3. Confirm deletion

Warning: Cannot be undone. Existing links will break.

Performance

Speed

Redirect time: ~10ms average

Why so fast?

• Edge network (Vercel Edge Middleware)
• Redis cache (no database queries)
• Deterministic algorithm (simple hash calculation)
• Global CDN (request handled at nearest edge)

Scalability

Handles:

• Millions of redirects per day
• Thousands of requests per second
• Global traffic distribution

Powered by:

• Vercel Edge Runtime
• Upstash Redis (global)
• SHA-256 hashing

Analytics (Coming Soon)

Track link performance:

Metrics:

• Total clicks per link
• Clicks per variant
• Conversion rate per variant
• Traffic sources
• Geographic distribution

Use case: Determine which variant wins your A/B test.

Current workaround: Use UTM parameters and track in your analytics:

Variant A: /page-v1?utm_content=variant_a
Variant B: /page-v2?utm_content=variant_b

Best Practices

Testing Strategy

1. Start with 50/50 split for unbiased results
2. Run tests for at least 1-2 weeks for statistical significance
3. Aim for 100+ conversions per variant minimum
4. Test one variable at a time (headline OR image, not both)
5. Don't stop test early even if one variant is "winning"

Weight Distribution

1. Even splits for A/B tests (50/50, 33/33/34)
2. Gradual rollouts start small (90/10 → 80/20 → etc.)
3. Always sum to 100 exactly
4. Consider minimum viable traffic (don't use 1%, results won't be meaningful)

Variant URLs

1. Use landing pages on same domain for consistency
2. Include UTM parameters for tracking
3. Ensure all variants are published before going live
4. Test each variant URL manually before launching

Link Slugs

1. Use descriptive names (e.g., spring-sale-2024, not link-1)
2. Keep it short for easy sharing
3. Avoid special characters (use hyphens, not underscores)
4. Include campaign/test name for organization

Limitations

Current Limitations

1. No built-in analytics (coming soon)

   • Workaround: Use UTM parameters + your analytics platform

2. No true geo-routing (coming soon)

   • Workaround: Use weight distribution as proxy

3. No time-based routing (coming soon)

   • Workaround: Manually adjust weights at scheduled times

4. No device-based routing (coming soon)

   • Workaround: Create separate links for mobile/desktop

5. Cookie-dependent consistency

   • Users who clear cookies may see different variant
   • Private browsing won't persist across sessions

Weights Must Sum to 100

Invalid:

Variant A: 50
Variant B: 60
Total: 110 ❌

Valid:

Variant A: 45
Variant B: 55
Total: 100 ✅

Advanced Usage

Sequential Testing

Test variants sequentially, not simultaneously:

Week 1:

• Variant A: 100%

Week 2:

• Variant B: 100%

Compare results. Useful when simultaneous split isn't possible.

Multi-Armed Bandit (Coming Soon)

Automatically adjust weights based on performance:

• High-converting variants get more traffic
• Low-converting variants get less traffic
• Dynamically optimize over time

Personalization (Coming Soon)

Route based on user attributes:

• First-time vs returning visitors
• Logged in vs logged out
• Subscription tier
• User preferences

Troubleshooting

"Link not found (404)"

Possible causes:

1. Link slug is incorrect
2. Link is deactivated
3. Domain not configured
4. Link was deleted

Solution: Verify link exists and is active in Links dashboard.

"Visitor keeps seeing different variants"

Causes:

1. User is clearing cookies
2. Using private/incognito mode
3. Different devices/browsers

Solution: This is expected behavior. Variant assignment is per-device/browser.

"Traffic split doesn't match weights"

Causes:

1. Small sample size (< 100 visitors)
2. Bot traffic (not assigned variants)
3. Recently changed weights (existing assignments persist)

Solution: Wait for larger sample size, exclude bots from analysis.

"Redirect is slow"

Causes:

1. Variant URL is slow to load (not the redirect itself)
2. Network issues on visitor's end

Solution: Test redirect speed with curl:

curl -I -L https://yourdomain.com/your-link

Redirect should be ~10ms. Page load time is separate.

What's Next?

• Creating Links - Detailed guide to creating and configuring links
• Traffic Splitting - Deep dive into weight distribution and routing
• Landing Pages - Create variants for A/B testing

Quick Reference

Pro Tip: Always test your link variants manually before launching to production. Ensure all variant URLs work correctly and land on published landing pages!