‘Zero scent’: could negative reviews of smelly candles hint at a Covid surge?

Yankee Candle calls itself “America’s favorite brand of premium scented candles”, and offers over 600 fragrances, like “Spiced pumpkin” and “Warm apple pie” designed to fill your living room with homey holiday vibes.

Days before Thanksgiving 2020, a Twitter user pointed out a sharp rise in negative reviews complaining the famously pungent candles had no smell. Could it be a hidden sign of the Covid wave, the user wondered? In total, the candles have well over 100,000 reviews on Amazon – a potentially rich trove of epidemiological data.

That off-the-cuff tweet led to a flood of jokes, but has since been validated by scholarly research: there is indeed a correlation between Covid cases and the number of reviews complaining that Yankee Candles don’t have a smell. In early 2022, the rise of negative reviews mirrored official case counts.

So it’s concerning that in the last few weeks, there has been a notable uptick in reviews complaining that the candles don’t seem to work. (Yankee Candle did not respond to a request for comment for the story.)

One reviewer wrote on 1 October that the smell of the “Mountain Lodge” candle was “too subtle, not like any other Yankee candle I’ve burned in past 35 yrs. Doesn’t even fill one small room.” Someone who bought “Sparking cinnamon” wrote on 9 October they were “excited to light this candle … it was a disappointment … zero scent.”

As more Americans seem to be giving up on Covid measures, the candles are again attracting attention – including from Dr Jorge Caballero, the founder of volunteer group Coders Against Covid who published a viral tweet about the candles on 9 October: “Yankee Candle reviews indicate that Covid is about [to] surge again … Mask up and get boosted. Please.”

Caballero’s tweet cited the work of Nick Beauchamp, the Northeastern University political scientist and statistician who published a peer-reviewed paper about the correlation between Yankee Candle reviews and Covid data earlier this year. Beauchamp, whose main work involves analyzing online posts to predict political outcomes, told me he got involved accidentally after he tweeted about the candles last Christmas break and went viral: “I thought, ‘Oh, shoot, I gotta really do this.’” So he completed the study and submitted it three weeks later – “faster than I’d ever written a paper”.

In his study entitled, Detecting the Effect of Covid Anosmia on Amazon Reviews Using Bayesian Vector Autoregression, Beauchamp used nearly 10,000 Yankee Candle reviews from Amazon and found that each additional 100,000 new Covid cases per week resulted in an estimated 0.25% increase in the number of “no smell” reviews.

He controlled for seasonality to make sure the increase in “no smell” reviews wasn’t simply because more people buy candles in the winter. He also looked at whether there was any correlation between “good smell” reviews and Covid cases (no) or reviews and flu cases (no).

Beauchamp repeated his study on more than 6,000 Amazon reviews of Vera Wang perfume, and found a similar, though slightly weaker relationship between reviewers complaining about “no smell” and official Covid data.

That was enough for him to conclude that cases did affect ‘‘no smell’’ reviews, but the study found no significant predictive relationship between reviews and Covid cases, meaning that the reviews only mirrored rather than foreshadowed the pandemic data.

That seems to have already changed. In June, Beauchamp redid the analysis to add data from the Omicron wave. “What I found is that the parameters of the model had changed a little bit, that the reviews really had become more predictive of Covid cases than before,” he said.

“My hypothesis was that as measurements get more irregular and poorly reported, the official accounts were sort of lagging the actual cases and it was taking longer for these effects to turn up. And as a result, the reviews – which are still chugging along – become more useful as a signal.”

On Monday night, Beauchamp ran the analysis again to add the last four months of data. This time, he found something strange: while “no smell” complaints for Yankee Candles have increased over the last couple months, the case counts for Covid have moved in the opposite direction.

“All of a sudden now, these two curves that have been tracking each other pretty well, for almost three years now are starting to diverge fairly dramatically.” That means Yankee candles reviews are no longer able to predict official case counts – “But the question is, is that because the official case counts are becoming so poor, because nobody’s reporting those cases, or even testing themselves?”

Perhaps that’s why doctors like Caballero are starting to look at Yankee Candle reviews, Beauchamp said. But while the reviews are “worth adding to our collection of indirect data sources,” he cautioned that the candle reviews were “fairly minor and idiosyncratic”, and not something people should use to make decisions. “If we have to rely on Yankee Candle data, that’s not great.” (Caballero did not respond to a request for comment.)

There’s one last caveat when it comes to linking candles and public health: as more people hear about this , it probably becomes less useful as a data source.

Perhaps the upside is this method can serve as some kind of indication that people are still trying to stay safe from Covid at a time when it feels we’re ready to give up.

“The scent of these candles are so strong that I can’t walk within 50 feet of a display without feeling like a monster pumpkin pie assaulted me, complete with massive headache,” wrote a reviewer on 9 October. “If you can’t smell this pie beast then you have Covid. Get tested, get booster, and mask up!”