With endless possibilities of Thanksgiving Day sides and plates (and stomachs) that are only so big, there’s bound to be some decision-making on Thursday. Turkey is a staple, but what else should you prepare?
Mashed potatoes or green beans? Sweet potato casserole or cornbread?
In the spirit of giving — Thanksgiving meal planning down to a science
To find out which sides are the real crowd-pleasers and which ones we might be making in the name of tradition but are destined to grow cold on the table, we simulated that decision-making experience and came up with the ultimate, data-based menu recommendation.
We designed a survey with 31 different side dishes — everything from ambrosia to green bean casserole — and had respondents indicate their preferences in an Anchored MaxDiff exercise. Additionally, we used the MaxDiff responses to run Total Unduplicated Reach and Frequency (TURF) and Total Reach Frequency (TRF) analyses to determine the best way to approach your Thanksgiving meal planning.
MaxDiff: The crowd-pleasers
Respondents were shown a series of five different side dishes and asked to select which was the most appetizing and least appetizing. By forcing respondents to make a decision between options, we were able to determine each person's first, second, third (and so on) favorite. This produced a list ranked 1-31 for side dish preferences relative to the other dishes, resulting in the following top 10:
MaxDiff pro tip
Make sure you compare “apples to apples” when designing a MaxDiff survey. It’s important that the options are natural trade-offs (e.g., rather than pitting casseroles against turkey) to not skew the data.
Anchored MaxDiff: Thanksgiving isn’t complete without them
Sides at the top of the MaxDiff list were ranked for being most appetizing, but would they be missed if they were not in the spread? To find out which items people consider most important to have on Thanksgiving Day, we added the following anchor question to the end of the survey:
Which of the following sides do you feel are "must haves" for Thanksgiving dinner?
By allowing respondents to choose multiple dishes from the list of 31, we were able to see which ones people would actually expect to find on the table:
Despite over half of respondents selecting gravy and cranberry sauce as “must haves,” they didn’t come in anywhere near the top in the raw MaxDiff scores. This shows that although people might not opt for them on their own or over a staple like potatoes, they are still a fundamental part of the Thanksgiving experience. Conversely, despite scalloped potatoes ranking number two in appeal, they weren’t considered a top “must have” for a Thanksgiving feast, falling out of the top 10 when ranking based on the "must have" criteria.
Anchored MaxDiff combines the appeal information from the raw MaxDiff scores with the "must have" question responses. This results in a more robust, accurate reflection of the combination most likely to both satisfy appetites and meet Thanksgiving expectations.
Anchored MaxDiff scores
Anchored MaxDiff pro tip
Make sure to write a firm anchor question that has a concrete or yes-or-no answer. Asking someone “what items would you actually purchase?” versus “what items do you like” will garner more accurate results.
TURF: Something for (almost) everyone
To make sure even picky eaters don’t end up empty-plated, we refer to TURF analysis. TURF allows us to identify items with the highest combined reach. When combining sides with the greatest reach, you have the highest chance of each guest being satisfied by at least one of the items. If we were to serve the four sides with the greatest reach — mashed potatoes, sweet potato casserole, meatless stuffing, and mac and cheese — almost 90% of our guests would be happy with at least one of the options.
TURF pro tip
Use screener questions to ensure participants meet certain criteria or so that you know who is responding. Having vegetarians or participants who are lactose-intolerant could impact your results — if you haven’t identified those traits ahead of time, you won’t understand your results as completely as you could.
TRF: The more exclusive gatherings
If you’re looking to feed a smaller crowd this year, you might not need all of the top-performing sides to satisfy your guests — you just want four that hit the spot. TRF analysis can indicate which combinations will be most popular.
Assuming cranberry sauce and gravy are already must-haves, a TRF analysis narrows down which four options will best complete the spread. This analysis revealed a list of 70 possible four-side combinations, 21 of which contain at least two items that most people would like:
Clearly, you’re going to need mashed potatoes. From there, it’s really up to your discretion which combination to choose — our data shows that there are no statistically significant differences among these top combinations. If you wanted to get more granular, we could look at the Shapley values for each item that varies between options, but that’s for digging into another time — you’ve got cooking to do. For a Thanksgiving meal, other decision-making factors could include the chef’s preferences or ingredient costs and availability. Regardless of which of these combinations you pick, it’s safe to say you’ll have a well-rounded meal that satisfies.
TRF analysis pro tip
Keep context in mind when choosing the optimal combination. While a rank order may be just fine for some business cases, additional analysis can help understand which three (four, five...all the way up to nine) options go together to optimize an offering.
Advanced analytics, with the Thanksgiving win
Thanksgiving may look different for a lot of us this year, but the holiday’s food favorites aren't going anywhere. We are thankful for our Advanced Analytics team for putting together the ultimate Thanksgiving recommendations with the data to back them up. Get connected with a team member to learn more about how they did it.
- advanced analytics