CBSplit CMS Wiki

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Why CBSplit does not run split tests for you ======

===== The expectation mismatch =====

When people hear “split,” they assume:

  
  * Automatic A/B testing
  * Variant creation
  * Statistical winner declaration
  * Auto-traffic shifting

CBSplit does not do this.

And that is intentional.

===== CBSplit is not a page-level experiment tool =====

Classic split testing tools:

  
  * Inject scripts into pages
  * Modify headlines or layouts
  * Track event-level differences
  * Declare winners based on short-term metrics

CBSplit operates at a different layer.

It evaluates revenue systems, not page elements.

===== Running tests is easy. Measuring truth is not =====

Creating variants and sending traffic to them is simple.

Understanding:

  
  * Refund-adjusted performance
  * Rebill durability
  * Cohort-based LTV
  * Processor-safe scaling

Is not simple.

CBSplit focuses on outcome intelligence, not variant automation.

===== Automation without lifecycle awareness is dangerous =====

Auto-running split tests can:

  
  * Shift traffic based on early signals
  * Reward impulse-driven conversions
  * Scale fragile revenue
  * Amplify refund risk

Without lifecycle visibility, automation accelerates mistakes.

CBSplit requires lifecycle maturity before decisions are made.

===== CBSplit provides evaluation, not decoration =====

CBSplit does not:

  
  * Change your headlines
  * Modify your button colors
  * Inject DOM experiments
  * Optimize surface-level elements

It provides:

  
  * Refund-adjusted metrics
  * Rebill-aware analysis
  * Cohort stability tracking
  * Outcome-based segmentation

It measures what matters after the experiment ends.

===== Split tests are tactics. CBSplit is infrastructure =====

Split testing is a tactical tool.

CBSplit is infrastructure for:

  
  * Revenue durability
  * Lifecycle evaluation
  * Traffic-quality filtering
  * Risk-aware scaling

It supports testing decisions.
It does not replace strategic judgment.

===== Responsibility remains with the operator =====

CBSplit does not remove human decision-making.

It provides:

  
  * Clear outcome signals
  * Lifecycle-adjusted data
  * Segmented performance insights
  * Risk indicators

Operators still design experiments.
CBSplit ensures experiments are judged correctly.

===== Automation without understanding creates fragility =====

Many systems promise:

  
  * Auto-optimization
  * Automatic winners
  * Hands-free scaling

These systems optimize events.

CBSplit ensures that optimization aligns with net revenue truth.