What Makes an Online Flash Sale Successful?
Skip to content
Operations Jun 6, 2017

What Makes an Online Flash Sale Successful?

When ratings and reviews aren’t enough, showing that a deal is popular can convince others to buy.

A customer reacts to the low inventory on online flash deals.

Yevgenia Nayberg

Based on the research of

Ruomeng Cui

Dennis J. Zhang

Achal Bassamboo

A “For a Limited Time Only!” sale sign is guaranteed to attract customers’ attention, whether it is hanging in a store window or flashing online. Retail platforms such as Amazon, Groupon, or Living Social have been quick to capitalize, offering steep discounts on thousands of limited-time deals every day. And these so-called flash or lightning deals are a hit: Americans spend $2.5 billion a year on them.

Yet what predicts a deal’s success?

It can be hard to tell. There are a lot of potential factors: a product’s ratings, its customer reviews, and, of course, how much the price has been discounted. There is also that ticker that shows the customer the fraction of deals that have been claimed.

Achal Bassamboo, professor of operations at the Kellogg School, wanted to see whether this final piece of information affected a customer’s decision to buy. Is someone more likely to pull the trigger when claimed deals are high, he wondered? To find out, he and colleagues cleverly manipulated the number of products in stock during Amazon flash deals to see how customers reacted to dwindling inventory. The researchers found that when the number of items claimed increased, so did the rate at which customers snapped them up.

“When you look at how many customers have purchased, it tells you what other people are doing and thinking,” Bassamboo says. “The customer learns about the product and others’ perceptions, both of which affect how they proceed with a purchase.”

Popularity Drives Sales

Slashing prices in flash deals makes for big business. As of 2012, customers in the U.S. spent approximately $7 million each day on flash deals. And that number is expected to grow. So understanding why some deals sell out faster than others is crucial to retailers looking to improve sales.

Together with his colleagues Ruomeng Cui of Emory and Dennis Zhang of Washington University in St. Louis, Bassamboo looked at several different factors that customers consider when deciding to splurge on a sale. Their goal was to determine the most important factors that lead someone to click that “buy” button.

Beyond investigating how shoppers evaluate the product itself—via ratings and reviews—they measured how shoppers evaluated the quality of the deal. In other words, is the discount a good one? “When the rate of people purchasing a product increases, it tells us it is a good deal,” Bassamboo says. “It reveals the quality of the deal rather than the product itself.”

“Showing that a deal is a good one helps align buyers’ and sellers’ expectations.”

The researchers looked at how these data influenced sales on nearly 24,000 “lightning deals” on the Amazon website, lasting an average of 5 hours each. On average, the items they looked at had 130 reviews, were rated 3.91 out of 5 stars, and were discounted 22–39 percent off the original prices.

By watching the rate at which items were sold, the team found that approximately 4.5 percent of an item’s inventory was claimed each hour. But this rate of purchase increased over time, as more people added an item to their online carts. Overall, a 10 percent increase in the fraction of deals claimed appeared to lead to a nearly 3 percent increase in its sales within the next hour.

So popularity breeds popularity. But what exactly was at work here? Was it simply that as time went on, more people heard about the offer through social media or from friends? Or was it that potential customers learned something about the item simply by looking at what percentage of a deal had been claimed?

“It could just be that as time goes by, more people are stumbling across the deal and buying,” Bassamboo says. “A very simple explanation is that there is no learning from other people, it is just diffusion of news.”

To get at the answer, the team hired research assistants to claim large numbers of items on sale by adding them to their shopping carts. When they did so, the percentage of inventory available to real shoppers decreased, though the researchers did not end up actually buying the products.

The experiment had a strong impact on customer behavior. For each 10 percent increase in the number of items claimed, the rate at which they were purchased increased by 2 percent—similar to what researchers saw when they studied customer behavior without manipulating inventory. This indicates that it was not the passing of time—and subsequent spreading of information about the deal—that was at work, but the actual number of items claimed.

Seeing that a particular item was popular with other customers strongly influenced people’s subsequent decision to buy it, according to Bassamboo. “When people see that a deal is popular, it gives them additional information they did not have otherwise,” he explains. “That has a powerful effect when a consumer is thinking about what action to take.”

Learning in Uncertain Times

But popularity is not everything. Customers only relied on information about how many deals had been claimed when other, concrete information about the deal’s quality was less available.

For example, if the seller offered a very deep discount, the percentage purchased had less of an effect on the rate of subsequent sales. In contrast, if a product’s ratings were high or its reviews were excellent—meaning the quality of the product was vouched for—the impact of the other customers’ buying behavior did matter.

“It is all about when customers need the extra information,” Bassamboo says. “Learning seems to matter when uncertainty about information exists. That was an unexpected finding.”

The reason, he suspects, is that people may derive comfort from popular opinion. “The thinking is, ‘Maybe I am not able to evaluate this deal completely, but the other customers are able to, and I need to take that signal into account’,” he says.

Enhancing Sales

Shoppers tend to be more uncertain of their purchases online than in stores because they cannot touch or try out products. Although many retailers have easy return policies, the effort of sending an item back, restocking, and refunds can be tedious for both buyer and seller. Any additional information that helps a buyer feel more confident in their choices can further sales, Bassamboo says.

While reviews and discounts provide some of that confidence, these results suggest that providing information about the quality of the deal itself could be beneficial. “Our study shows that in addition to looking at what other people think of a product itself, buyers also want to know how good the current price is,” he says. “That question should get equal importance, especially in the context of these lightning deals.”

One possibility, he says, would be to show the price at which other customers purchased a product along with their reviews or ratings. That way, potential buyers can more easily evaluate whether a sale price is worth their money. For example, if a buyer who rated an item at 5 stars had purchased it for $5, but another customer paid $10 and left a 2-star rating, a potential buyer might be able to better decide how much to pay for it.

“Showing that a deal is a good one helps align buyers’ and sellers’ expectations,” Bassamboo explains. “To show that a deal is not just 20 percent off, but to know that it makes sense to buy the product at that price—this is especially important when deals are being constantly updated online.”

Featured Faculty

Charles E. Morrison Professor of Decision Sciences; Professor of Operations; Co-Director of MMM Program

About the Writer
Jyoti Madhusoodanan is a Bay Area-based science writer.
About the Research
Cui, Ruomeng, Dennis J. Zhang, and Achal Bassamboo. 2017. “Learning from Inventory Availability Information: Field Evidence from Amazon.” Working paper.

Read the original

More in Operations