Diet Crystal Pepsi was launched in 1992 after getting a strong response in test markets, but a year later the product was pulled from the shelves after weak sales. Similarly, Frito Lay Lemonade and countless other products that passed the pilot phase went on to be failures. How could there be such a mismatch of early promise and subsequent performance?

Eric Anderson, Song Lin, Duncan Simester and Catherine Tucker offer an unusual answer: because of the customers who bought those products.

In a study that covered more than 10 million transactions from 111 outlets of a large chain of convenience stores in the US, they show that when outlier customers with niche preferences adopt new products, it can signal a greater likelihood that a product will fail.

“At first glance the result is striking,” they said. “But we find that there is an unrepresentative subset of customers whom we call Harbingers of failure who are more likely to buy products that other customers do not buy. A purchase by these customers therefore may indicate that the product appeals to a narrower slice of the marketplace, and so will be more likely to fail.”

The dataset they worked with allowed them to compare purchases by individual customers who used the retailer’s frequent shopping card from November 2003 to November 2005. The authors looked at whether they purchased new products, how many they bought, and also which of these products were still selling three years after they were launched.

Customers who bought at least two new products during the products’ first year were classified according to whether those products flopped. The Harbingers were those more likely to buy such products and, most importantly, to make repeat purchases of them.

“A one-time purchase of Diet Crystal Pepsi is partially informative about a consumer’s preferences. But a consumer who repeatedly purchases the drink is even more likely to have unusual preferences and more likely than others to choose other new products that will fail in the future,” they said.

This led the authors to suspect that such customers also tended to purchase existing products that other customers did not buy. Indeed, Harbingers were found to be more likely to purchase niche items, which collectively contribute only one per cent of total unit sales in the retailer, and very niche items (0.1 per cent).

“The result is consistent with the explanation that Harbingers have preferences that are systematically different from other customers. If they adopt a new product, it may signal that other customers will not be attracted to it,” the authors said.

“This is essentially the opposite of the argument made in other research that firms can benefit by distinguishing ‘lead users’ from other customers. Whereas lead users provide a positive signal of product success, Harbingers provide the opposite signal.”

Firms should therefore take heed and distinguish between Harbingers and other customers when making product line decisions, they said.

Having said that, the identification of Harbingers of failure means there should exist their opposite: Harbingers of success. The study provided some evidence that the latter may exist but further research will be needed to confirm this.