On the significance of statistically insignificant results in consumer behavior experiments