Intelligently split click traffic across landing page variants to maximize conversion rates.

Percentage routing (A/B Testing) directs a defined proportion of traffic to different page variations (e.g., 50% to Variant A, 50% to Variant B, or 70% / 15% / 15%).
Routing decisions execute directly on Cloudflare Edge servers during the network handshake, with zero client scripts or visible layout flicker.
Core architectural pillars ensuring high performance, isolation, and security.
Ready-to-use TypeScript code snippets and configuration examples for your integrations.
Associate variations with their respective percentage weights:
Direct HTTP request methods, parameters schema, authentication, and responses.
Configures or updates percentage traffic split weights across multiple landing page variants.
| Parameter | Type | Required | Description | Example |
|---|---|---|---|---|
| id | string | Required | Short link slug or ID. | promo-split |
| Parameter | Type | Required | Description | Example |
|---|---|---|---|---|
| mainWeight | number | Required | Percentage of traffic sent to the primary target URL (0 to 100). | 50 |
| abVariations | object[] | Required | Array of variant configurations. | [{"id": "var_b", "targetUrl": "https://store.com/landing-b", "weight": 50, "name": "Variant B (Green Button)"}] |
| abVariations[].id | string | Required | Unique variant key. | var_b |
| abVariations[].targetUrl | string (url) | Required | Destination URL for this variant. | https://store.com/landing-b |
| abVariations[].weight | number | Required | Percentage weight allocated to this variant. | 50 |
| abVariations[].name | string | Optional | Human-readable label for dashboard comparison. | Variant B (Green Button) |
Payload confirming percentage routing weights:
Expert tips to maximize performance, resilience, and link security.
Ensure total weight sums to exactly 100% (mainWeight + sum of variation weights = 100).
Run tests until reaching statistically significant sample sizes (at least 500 to 1,000 clicks per variant).
Test only one major variable per experiment (e.g. headline vs hero image) to clearly identify causality.