How long do you run something before you consider a report meaningful?

A few days rarely tells you much. In your experience, what is a fair minimum window before you would write up a field report and call the effect (or non-effect) real? Curious how people think about timeframe and honesty here.