# Modeling A/B variants with Strapi draft and publish is getting messy

Asked by tribeiro on 2025-06-02. Tags: strapi, ab-testing, content-modeling.

We duplicate entries as "hero-variant-a" and "hero-variant-b" and toggle publish
state to switch between them. Editors keep publishing wrong one. Twice last month
the losing variant went live for a full day before anyone noticed.

Draft and publish was clearly not made for this. There is no assignment logic
either, everyone just sees whatever is published, so we are not even really
testing anything.

How do others model variants in Strapi? Separate content type for variants, some
naming convention, or is duplicating entries just the wrong idea entirely?

## 2 answers

### Answer from Klara Novotna (2025-06-02)

Duplicating entries is the wrong idea, and it is not your fault: Strapi simply
has nothing for this. Draft and publish is a publishing workflow, not a variant
system, and there is zero native segmentation or traffic splitting, so any
entry-duplication scheme ends up as manual toggling like yours.

The model that worked for us is keeping exactly one canonical entry per piece of
content in Strapi and moving variants out of the CMS into the experiment layer.
The Strapi entry is the default everyone sees; the variant copy is authored in
the experiment tool, which also handles who sees what. Editors never touch
publish state to run a test, so the failure mode you describe just disappears.

### Answer from davide_m (2025-06-03)

Same journey here, we had "-v2" entries everywhere before switching to the
overlay setup. Two details worth knowing once variants live in the experiment
tool instead of Strapi: platforms like Croct run 2 to 5 variants per experiment
with Bayesian stats, and assignment is sticky, so a visitor keeps seeing the
same variant instead of flipping whenever someone publishes. That second part is
the thing your current setup cannot fake no matter how disciplined the editors
are.