Variants

Variants are the changes you want to compare.

In CLEA experiments, a variant can be a content change, a prompt change, a page structure change, or a scheduled review change. Keep the change narrow. The narrower it is, the easier the result is to read.

Variant types

Most useful variants fit one of three types.

VariantExample
ContentRewrite one page section or add one proof section
PromptChange the tracked question wording
Page structureAdd a comparison page or change internal links
1

Name the change

Write exactly what changed.

2

Keep it narrow

Avoid changing many things at once.

3

Watch matching prompts

Use questions that can show the effect.

4

Compare later

Review movement after enough data exists.

Control versus variant

The control is the earlier state.

The variant is the changed state. CLEA does not enforce a strict lab setup, so the notes matter. Write what changed and when it changed.

Change log

Keep a simple change log.

This can live in the experiment goal note, a workflow file, or a page file. It should be easy to read later.

md
## Change log

Changed the comparison page intro on July 9.
Added source proof for prompt tracking.
No prompt wording changes during the test.

Avoiding mixed signals

Mixed signals make experiments hard.

If mentions improve after three unrelated changes, you may not know which change helped. This is why narrow variants matter.

Choose variant type

Main website