How does this store use your likes and comments? Rather, he uses them to shape his social media marketing strategy. But it’s much less likely that the retailer will use that data to make operational decisions, like how many pairs of those jeans to make or whether to cut prices.
That could change. In a recent study, Antonio Moreno, associate professor of operations at Kellogg, found that social media data can improve sales forecasts. When the researchers incorporated information about a clothing company’s Facebook interactions into predictive models, they were able to more accurately estimate purchases the following week.
The use of advanced algorithms was key to the improvements, which means it’s not enough to just collect social media data: companies will also have to upgrade their forecasting techniques.
“It’s important to get new data but also to use more sophisticated forecasting methodology,” says Moreno.
The study did not reveal why Facebook information improves predictions. Moreno speculates that the data may reflect how much attention customers are paying to the brand, as well as good or bad word of mouth.
But companies may be more interested in the effectiveness of the technique than the mechanism behind it.
“If something works,” he says, “sometimes they can be able to live without knowing why it works.”
Use of social media data
The idea of mining social media data to guide operations is in its infancy.
For example, companies might display a subsequent shirt in two colors and see which generates more clicks. The company could then use this information to decide which color to produce. “But it’s not mainstream yet,” says Moreno.
And while academic studies have explored whether social media posts boost sales, little research has been done on using the data for internal operations decisions.
Moreno decided to explore this idea with Ruomeng Cui, at Emory University, and Dennis Zhangat the University of Washington, who are both former Kellogg PhD students, along with Santiago Galino at Dartmouth College.
The team partnered with an online clothing company. Most of the company’s social media traffic came from its Facebook page, which had more than 300,000 followers at the time of the study. But to forecast sales, the company relied heavily on basic information such as its overall sales growth and weekly or seasonal patterns — such as a tendency to sell more on weekends.
Moreno’s team wrote software to extract information about the company’s Facebook posts from January to July 2013. The final data set included more than 171,000 users, 1,900 company posts, about 25,000 comments and a quarter of a million likes.
The researchers then used language processing software to categorize each comment as positive, negative or neutral. In addition, the team obtained internal data about the company’s sales and advertising campaigns during this period.
Training of Prediction Models
Using what they gathered, the researchers produced two sets of sales forecast models: the base forecast, which included only internal company information, and a second forecast that combined internal and social media data.
For both baseline and social media predictions, the team experimented with a variety of prediction methods. Most of the models were based on machine learning, in which the model is trained to determine which factors are most important.
To assess accuracy, the researchers used a measure called the mean absolute percentage error (MAPE), which captures how much the estimate deviates from actual sales. For example, a MAPE of 10% would mean that, on average, the model estimates were 10% off.
The company’s existing sales forecasts for the coming week were MAPPED by 12%. The researchers’ best-performing baseline model—the one without the social media data—reduced the error to about 7-9%.
Adding social media data reduced it even further to 5-7%. However, social media data alone was not enough. When the team plugged social media information into an underperforming model, the accuracy could be even worse than the baseline model without the social media information.
The results suggest that both data and methods are important. “By bringing in social media data, we can do better,” says Moreno. “But it seems the first step should be to have better methods.”
Accurate predictions
Future research could investigate in more detail why social media improves sales forecasting. Researchers could also conduct similar studies to predict sales for individual products rather than just total sales. And if the data could be broken down by geographic region, the information could help companies decide how much of a particular product to ship to, say, Texas versus Idaho.
Moreno notes that the study’s results may not apply to all industries. Social media data is more likely to relate to products with highly uncertain sales or industries that are heavily influenced by trends, such as fashion and entertainment. But for consumer goods like breakfast cereal, sales are already fairly predictable, so adding Facebook data may not improve forecasts much.
Companies could also become more strategic about their social media posts in order to draw specific insights to help guide their operations. For example, more companies may adopt the practice of presenting potential products and deciding what to build based on customer reactions.
“They can really use this social media to learn and make better decisions,” Moreno says.
