Consumer preferences are ever-evolving and always surprising. There was the pre-teen Tamagotchi trend that went from obsessive to non-existent. The resurrection of Polaroids and vinyl in a digitized world. Or, more recently and most shockingly: the comeback of JNCO jeans.

We were interested to find out if the pandemic impacted our preference for different flavors. Spoiler alert: it did.

We built AI models based on US consumer tasting data from the six months leading up to March 2020 and the six months following. Then, we used the Gastrograph AI system to map out how those two tasting groups would respond to different products and track the changes in flavor preference.

We’re sharing the results of our study to demonstrate that if you want to be the first to know about changes in consumer preference, you should use Gastrograph AI for your new product development.

We Started to Like Sour Flavors Even More

Before the pandemic, preference for sour and acidic flavors was already increasing — COVID accelerated that increase.
We noticed that there was more preference for products with a higher intensity of sour and acidic flavors, across all product categories, compared with pre-COVID.

The top two graphs show preferences for apple cider vinegar, and the bottom two show preferences for malic acid.

On the left, you can see preferences across different intensities pre-COVID. For both apple cider vinegar and malic acid, preference is generally low. For products with a high intensity of malic acid flavors, preference dips sharply.

The graphs on the right show the results from the contemporary COVID model. Higher intensity is more popular for both flavors. For the apple cider vinegar flavor, you could now increase the intensity in products to 10% before it would negatively impact how much people enjoyed them. In contrast, the peak intensity for malic acid is around 6% intensity.

We Lost Our Sweet Tooth

COVID accelerated the speed of another existing trend: a declining preference for sweetness. We analyzed high fructose corn syrup (HFCS), which is a sweetener you’ll find in candy, fast food, and soda.

Again, the graph on the left shows the pre-COVID preference, and on the right, you can see a general dip in preference from the contemporary COVID tasting data.

The contemporary COVID data set shows that preference for products with HFCS has decreased.

Although at the peak intensity (10%), preference almost matches pre-COVID levels, overall, we found the preference for products with HFCS was lower than before the pandemic.

Our theory for the acceleration in this trend is that because products that contain HFCS are typically sold on-premises (like at sports matches), quarantine impacted the sales of these products. People simply weren’t consuming as many foods and drinks with high levels of HFCS as before. Perhaps this meant they became more sensitive to sweet flavors and perceived them as lower quality.

As well as exploring preference for the intensities of different flavors within products, we were also curious to find out if our overall preference for certain types of products changed. Our system used the pre-COVID and contemporary COVID models to predict how the two groups would respond to different products, and we analyzed the results by looking at the Perceived Quality (PQ) scores they generated.

We Found Comfort in Familiar Snacks

Pre-COVID, US consumers had a relatively high preference for rare and novel flavors in salty snacks. In the contemporary COVID predictions, this trend reversed.

For our study, we defined ‘rare’ flavors as those which don’t appear in many salty snacks profiled in the US, for example, ginseng, pumpernickel, tamarind, and olive. We compared products with common flavors with those that contained rare flavors.

The x-axes show the PQ score out of seven for the two product types. The y-axes show what percentage of the population would give each PQ score. The graph on the left shows the pre-COVID results: snacks with rare flavors (blue) are generally more popular than snacks with common flavors (red). In other words, consumers pre-COVID enjoyed tasting products with flavors they weren’t familiar with.

The graphs show preferences for common snacks vs. rare snacks before and during COVID.

In the contemporary COVID results, the percentage of the population who would give the highest PQ score — seven out of seven — for common snacks increased, and it’s higher than for rare flavors. Our tendency to enjoy unfamiliar flavors switched; instead, we preferred snacks with common, familiar flavors.

We Enjoyed a Drink

When we looked at alcoholic beverages, we discovered that preference for spirits in general increased. We also found there was less of a distinction between the PQ scores for craft and mass-market spirits.

To distinguish between the spirit types, we chose a set of product examples to represent each category:

• Craft: Lagavulin Scotch Whisky 16 and Port Charlotte Heavily Peated Islay Barley 2011
• Mass market: Jack Daniel’s Tennessee Honey, Jameson Cold Brew Irish Whiskey with Coffee, Fireball Cinnamon Whiskey, and Bacardi Premium Cocktail Zombie

In these graphs, the red lines represent craft spirits, while the blue lines represent mass market spirits. The graph on the left shows the pre-COVID set: there’s a big difference in the PQ scores of craft and mass-market spirits. A larger percentage of the population would give craft spirits a PQ score of 6/7 compared with mass-market spirits.

The contemporary COVID results show less of a distinction between craft and mass-market spirits.

In the contemporary COVID predictions, we found that preference for spirits, in general, went up — for both categories, the percentage of PQ scores of 7/7 increased. What’s more, there’s not as big a difference between the scores for craft and mass-market spirits in the contemporary COVID set — the average PQ score (represented by the vertical blue and red lines) is almost the same.

How We Got Our Data

To build the Gastrograph AI platform, we collected tasting data from all over the world by getting consumers to taste products with a diverse range of flavors, aromas, and textures. We then trained AI models with the data to allow the system to understand how different consumers perceive flavor and learn to predict their preferences for different flavor combinations.

We continuously collect new data to update the system and add to our sensory database, which is the largest in the world.

For this COVID study, we first split the data into two training sets:

• Pre-COVID — tasting data from US consumers in the six months before March 11 2020, when there were no lockdown restrictions
• Contemporary COVID — tasting data from US consumers collected during lockdown restrictions, which reflects the state of people’s preferences in December 2020

Next, we created predictive models for each time period, so our system would be able to make predictions about how each set of US consumers (pre-COVID and contemporary COVID) would respond to different products. Then, we used our system to predict preferences for different products. Finally, we analyzed the results to identify shifts in preference trends.

Stay Ahead of the Trends

While there are companies that identify food and beverage trends as they emerge online, Gastrograph AI is the only platform that’s able to make accurate predictions about preference trends before they happen. Without our platform, you’re missing the opportunity to shape your product development around preference and get ahead of changing trends.

‍