Cooking Magazine Bon Appetit's Possible Slowdown
For my web scraping data project, I chose to scrape the cooking magazine Bon Appetit's online recipe catalog. For it, I built a spider that crawled through search result pages for all twelve months for every year since 2014. The spider collected nearly 4,000 recipes of which I recorded the dish's name, ingredients, publishing date, rating, number of ratings, review text, and number reviews written.
With this information, I analyzed commonly repeated ingredients, trends in reviews and ratings, recipe output by the magazine, and finally how COVID-19 and the recent c-suite reorganization might be affecting the magazine holistically.
Initially, my expectations were that both online publication of recipes and user involvement would rise over time. I anticipated more recipes in years closer to the present as companies slowly become more and more digitally friendly. And just as companies grow with technology, I expected consumers to also become more present online, writing more reviews and rating recipes more frequently.
Surprisingly, however, only half of my expectations were true. According to the data scraped, Bon Appetit is actually decreasing in recipes published per year, and at a pretty steep decline. While my inclination was to expect more digital presence as the years went on, I might have misplaced where that presence would be.
Output in general it seems has been on the decline for a few years. The graph to the left illustrates a lessening in new material for the past three years. Bon Appetit's online recipe peak looks to fall within 2016, where a total of 747 new recipes were published.
After a peak in 2016, followed by an almost as close year in 2017 with 730 recipes posted, content output swiftly declined. 2018 saw only 503 recipes, while 2019 only made it into the low 400s. The reduction in output might be a sign of shifting priorities in the company.
While 2020 has only just passed its halfway point, there is still a considerable difference in the content output. Note, when scraped, August's content had only begun to be released to the website. August's numbers are not final.
Another feature to look at: July seems to be a very slow month for the company in terms of new material. For the past two years, and this year included, July consistently sees a drastic drop in new recipes published.
Reviews and Ratings
(Because a consumer can leave a rating without a review or a review without a rating, the two metrics are looked at individually.)
The graph below shows that, except for a few very negative people in 2016 and 2017, recipes have been mostly well-rated. In fact, the average lowest rating has had a general upward trend, with an average clustering around 4.0-5.0 stars from the end of 2018 to the present day.
It would seem that before late 2015, however, the rating system operated differently, as there is an unusual block of highly varied ratings up to 2015. After 2015, ratings seem to often snap to whole numbers, e.g. '3.0', '4.0', and '5.0'. This could be because a dish's overall rating was once averaged from a collection of questions gathering ratings on facets of the dish (a recipe's clear/poor instructions, the use of certain ingredients, overall taste, and others). Other possibilities are that users were given a sliding bar for a number out of 100 or some other scale size rather than a 1-5 star system.
The review count below shows an increase in multiple reviews being written per recipe. Previously, especially in 2015 and 2016, most recipes went unreviewed, or they received only one or two reviews. Since then, customer engagement has increased , and now recipes are more likely to receive multiple reviews.
While it is true that fewer recipes are being published each year by Bon Appetit, thus increasing the likelihood that a recipe is trafficked and reviewed, it is still the case that customers are engaging more with the material. The table to the right illustrates the stark increase in total reviews written each year. Recipes published in 2019 acquired over 7,000 reviews.
This is impressive given that time spent on the website, as often recipes are recycled and brought back into circulation, could have easily favored older recipes to gain more reviews; yet it is the newest listings that pull in higher and higher numbers of reviews.
To see the best rated or most reviewed dishes, check out the Shiny App for more, including each year's top recipe.
Most Popular Ingredients
In addition to analyzing trends in recipe output and consumer engagement, I refined each recipe into core ingredients to measure variability and ingredient fads in which the magazine might have participated. When looking at the most frequent ingredients, one ingredient takes the cake, or pasta water: kosher salt.
Kosher salt, and many of the other top ingredients, come as no surprise as they are all widely known to be pantry staples. Depending on your cooking habits, though, a few might be unusual, for example flaky sea salt (a second salt option within the top twenty list) or apple cider vinegar.
Further, upon research into the popularity of a few ingredients, Bon Appetit has both jumped on board and forgotten some fan favorite ingredients. Below are four charts showing the rise and fall of popularity for four particular ingredients: kale, pears, tomatoes, and lemons. More ingredients are explored in the Shiny App including pickles, red wine, and ricotta to name a few.
Each ingredient shows a very different history, as kale exploded in use in 2019 while pears, admittedly rarely used in the first place, saw a huge surge in 2015. Lemons are a staple through the years it appears. Finally, tomatoes, also a staple showing up in roughly 10% of recipes each year, appeared in double the amount of recipes in 2018.
Recipe Publishing Repercussions
March 2020 completely altered life as we knew it for most New Yorkers, and for many Americans across the country. Companies furloughed or terminated many workers, and those lucky enough to still have a job faced the challenges of working from home. With children home as schools shut down, and work hours bleeding into personal time (and vice versa), it's safe to say, productivity has not been easy to foster.
However, because Bon Appetit operates around its magazine publication, most of the work for March's reports was completed in February. According to the chart on the right, March was the most productive month of the year up to that point. April's output, however, potentially exhibits signs of dampened productivity due to Covid-19 restrictions.
By May, though, with some adjustments to shipping ingredients to personal addresses, cooking recipes with the constraints of a New York apartment kitchen, and testing dishes out on family members, Bon Appetit bounced back. May's level of recipe output matches that of March, and June follows very closely to May's high.
Consumer Engagement Repercussions
As older recipes have an advantage in gaining reviews, it's too soon to tell whether consumer engagement has been significantly affected by either Covid-19 or the recent upper management scandals. July's recipes have had barely half the time of June's recipes to acquire reviews and even less than May's recipes.
Additionally, Bon Appetit pulls together collections of older recipes in blog posts that compete for traffic on the webpage. It's very possible we see fewer reviews logged simply because these recipes haven't been in circulation as much.
Even though many Americans are stuck at home now, only time will tell if writing online reviews is where customers will spend their extra hours, or if the scandals have lost the company a significant portion of its fanbase and customers.
A Possible Decline for Bon Appetit
Though the research is still in its infancy, it does show that customers are engaging more online and are overall happier each year with the recipes being created by Bon Appetit, even though the magazine is outputting less and less each year. This could be because of a pivot in priorities; Bon Appetit might be focusing more on quality than quantity.
It is worth noting that according to the graph below, review writing has been on the rise since 2016. And recipes published during Winter 2019 have already seen record high review counts. However, while the start of a new year doesn't see as many reviews written as the end of one, 2020's review count is not encouraging of similar customer engagement patterns.
The line chart below shows that summers typically experience a dip in customer engagement. It might be a while, possibly early winter or the start of next year, before any effects of Covid-19 or the scandals can truly be evaluated.
This may be a decrease in customer engagement simply because summers historically experience low engagement. This may also be a result to tight schedules due to Covid-19. Or worst case, this could be the beginning of a shrinkage of Bon Appetit's customer base.
In late 2016, the last year before recipe output declines, Bon Appetit launched a YouTube channel featuring a test kitchen chef making fermented foods and featuring microbial food cultures. While the series' style was fairly unproduced and there were plenty of 'real-life' moments and mistakes, it was a hit.
Previously social media had been inundated with POV cooking videos featuring only hands and pans, so to see the actual chef and to connect with the person cooking the dish made Bon Appetit's channel stand out.
Following the 'It's Alive with Brad' series, Bon Appetit released more and more specialty channels, quickly gaining traction.
Bon Appetit has in fact increased their digital presence, just not in the way I had expected. By spending more time and resources per recipe published through their YouTube channel, Bon Appetit has pivoted their brand to one that has recipes consistently trending on YouTube and nearly 50 videos with more than 5 million views each. This redirection to social media may explain the fewer recipes as the magazine devotes more time and media platforms to each recipe. The decrease in output may be a result of prioritizing quality over quantity.
To conclude whether Covid-19 will hurt or help the company, only time will tell. And the same can be said for the recent scandals and how they will affect customers' engagement with the company. Summers are usually slow, and new recipes need time to accrue reviews before conclusions can be made regarding customer engagement. And as the saying almost goes, 'a watched review count never boils'.
The skills I demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.