A/B testing is simpler
It is easier to explain, implement and learn from one meaningful page-level change.
Multivariate tests need scale
More combinations divide traffic and increase the evidence needed for useful conclusions.
Use multivariate testing selectively
It can be appropriate on high-volume experiences when several element interactions are genuinely important.
Do not confuse complexity with rigor
A well-designed simple experiment is often more useful than a complicated test the traffic cannot support.