Last week’s reports of the death of hard seltzer are greatly exaggerated. Here are three articles that show the media’s prioritization of eye-catching, “punny” headlines over accuracy: “Hard seltzer boom goes flat,” “Hard seltzer slump hits hard,” and “In focus: Is the hard seltzer boom starting to fizzle out?”

A casual perusal of these and other headlines would leave one with the seemingly innocuous takeaway—“hard seltzer: it was an interesting fad while it lasted.” However, this is an inaccurate picture of what’s actually going on because “the hard seltzer sector continues to grow apace in absolute terms.” So let’s break down how this happened and how the concept of overfitting explains why we mistake headlines for news.

Share Ah So Insights

Real analysis begs for context, and the context surrounding hard seltzer reveals that:

• Hard seltzer sales are now larger than craft beer (IRI-defined) sales for the latest 12 weeks through Jun 13 in off-premise outlets. With hard seltzer up 34% YTD, its sales are ahead of craft beer (excluding big brewer brands like Blue Moon, Leinenkugel, and Shock Top).

• Less than three weeks ago, Jessica Infante of Brewbound reported:

  Hard seltzer could account for 10% of total beverage alcohol sales by 2025, according to a report from financial services firm Credit Suisse.

  “There is too much focus on how large seltzer can be as a percentage of U.S. beer,” the Credit Suisse Equity Research Americas team wrote. “We think that is [the] wrong metric, given that 50% of share is gained from wine and spirits.”

• Hard seltzer had $280 million in sales—more than for all red table wine—in NielsenIQ off-premise channels for the July 4th holiday.

Didn’t all the headlines say hard seltzer was dead?

According to Jordan Valinsky of CNN Business:

…Jim Koch said that the "hard seltzer category and overall beer industry were softer than we had anticipated…"

Shares [of Boston Beer] plummeted 25% in afternoon trading Friday after the company's second-quarter earnings came in below analysts' expectations late Thursday.

2

 

Essentially Boston Beer admitted that it bet too big on hard seltzer. Investors balked and the stock plummeted. This, in turn, grabbed the attention of journalists, who wrote stories. The dysfunction occurred when headline writers overfit these articles with an ersatz narrative about the death of hard seltzer.

Overfitting is a concept drawn from data science. It occurs when a statistical model fits exactly against the data used to produce it. It’s not missing the forest for the trees. Overfitting is missing the trees (data) for the forest (pattern). The problem with overfitting is that it biases us to narratives where there might not be any. We, as humans, want to see patterns.

The desire to get the smaller story—Boston Beer stock dropping—to align within a larger model left us all with the mistaken idea that hard seltzer is “over.” The whole scenario illustrates why understanding overfitting is so important and how it can lead us astray. It doesn’t just happen when we scan headlines.

Concerning our own behavior, Sachin Joglekar has written, overfitting “happens any time you read too much into your notions (whether right or wrong) to optimize something in your future.” As such our very desire to fine-tune something, be it process or product, puts us at odds with truly understanding that thing. People want to use models to explain data, but not all data fits neatly into models.

Still, we must train ourselves to recognize that a headline is a road sign, not a destination. We must constantly look for the context necessary to distinguish between a singular instance of something and a broader trend. Lest we want to assume the risk of prematurely writing off a category on the rise, we must redouble our efforts in questioning what is a fact and what is merely a fancy model.