On most ecommerce sites, run one test per funnel stage at a time so results stay clean. Running overlapping tests on the same page or audience makes it hard to attribute what caused a change. Higher-traffic sites can parallelize more; lower-traffic ones should sequence carefully to preserve sample size.

Long enough to reach statistical significance and to cover at least one full business cycle — usually a minimum of one to two weeks, often more. Stopping early because a variant “looks like it’s winning” is the most common way to ship a false positive.
Always run through full weeks to avoid weekday/weekend bias.

It depends on your baseline conversion rate and the size of the lift you want to detect — smaller expected lifts need far more traffic. Use a sample-size calculator before launching. If the required sample is unrealistic for your traffic, the test isn’t worth running at that stage; move to a higher-traffic page or a bolder change.

Often yes. Behavior differs enough between devices that a winning desktop variation can lose on mobile. At minimum, segment your results by device even if you run a single test, so a strong result on one device doesn’t get diluted by the other.

A/B testing is one tool within conversion rate optimization (CRO). CRO is the broader practice of diagnosing where and why visitors drop off, forming hypotheses, and validating fixes — A/B testing is how you validate. Good testing starts with analytics, not with a list of ideas.