I remember when I was majoring in a social science quite a while ago and performing experiments that the commonly held standard was if the standard deviation (margin of error) was greater than 4%, your results were considered worthless 😬