A one-sided test or a one-tailed test, lets you test for a metric moving in only in one direction which you specify in advance. This tradeoff gives you additional sensitivity (or power). Learn more
Examples of this include testing for regressions in guardrail metrics and testing for a change in which only one direction has any meaningful business impact. e.g. You're less interested in detecting if new features reduce crash rates, but are very interested in learning if new features are increasing crash rates.