Quantifying My Love For Chipotle
Have you ever wondered how much time and money you’ve spent at your favorite restaurant?
Every couple of months or so, I try to update my family’s budget. On my last trip looking through credit card transactions, I noticed that there was an option to export all of my purchase history to a CSV, finally giving me the power I need to answer a question my wife and I have debated for years:
How much have we spent on Chipotle?
Chipotle has a very special place in our hearts. Where we went to high school, Chipotle was right down the street from us. It is our default go-to whenever we’re hungry. When we are deciding on places to live, the first question is always “Where is the nearest Chipotle?”. Our friends and family only get us Chipotle gift cards as gifts. Everyone has to have one spot where they go more than everyone else on the planet, right?
In this article, I want to look at my Chipotle spending habits from a personal perspective. I analyze attendance probabilities, forecasting my future purchases, and focusing how a corporation like Chipotle could use customer spending habits like this to make better financial decisions.
First, lets look at our purchase patters of each day from the last year and a half
From 1/1/2018 – 6/30/2019, here are a few overview statistics of my family’s purchase patterns:
- Purchased Chipotle 109 different days, which averages to 1 in every 5 days
- The longest streak of consecutive days without Chipotle was 28 days (April of 2018, which we were on vacation for the second half of the month)
- The longest streak of consecutive days with Chipotle was 2, which was surprising
- Spent a total of $1,461.15, averaging out to $13.41 / purchase
- Attended Chipotle 82% of all available weeks, 100% of all available months
From the bar chart above, you can see that during the beginning of the year and colder months, I tend to not attend Chipotle as much as I do compared to other times of the year. During summer months, I attend Chipotle at least 15% (roughly 4.5 days) of each month.
Probabilities
Probabilities are a good initial way to figure the frequency patterns of how often I go to Chipotle. Using the binomial distribution we can see how many consecutive days in a row I would eat / miss Chipotle. Starting with how many days in a row I would likely get Chipotle, you see that purchasing Chipotle 3 days in a row is roughly 1%, which lines up perfectly with my maximum streak of 2 days in a row getting Chipotle the last year in a half.
Looking at the number of days not eating Chipotle, it has a much longer right tail than the original distribution. From this distribution, it will be unlikely to go 2 weeks without getting Chipotle at least once.
Time Series Forecasting
When creating my budget for how much we spend on Chipotle, the next logical question becomes how much can we expect to spend in the future? Since most budgeting isn’t at the daily or weekly level, I decided to look at it on a monthly basis. This requires summing up the total amount spent each month and using predictive models to predict how much will be spending in the future. Should I use a more aggressive model with seasonality or one that covers the overall trend of the data? Lets look at a few examples:
The orange and red lines represent Holt-Winters time series models while the green line represents an exponential smoothing model with additive seasonality and trend of 7 months. The Holt-Winters have more of a safe, linear prediction while the exponential smoothing model takes a more aggressive approach to try and get as much of the pattern down as possible. What makes the exponential smoothing model a more comforting model is that usually January is the lowest month, which is closely mimicked in the forecast. The Holt-Winters models stay more in the bounds of the known values, but are too linear in the forecasts making them unrealistic expectations.
How Can Corporations Use This Purchase Data?
Corporations like Chipotle can use this purchase data to their advantage. Understanding fiscal projections is always useful, but knowing how likely someone is going to come back to your store provides information about when to send promotional offers and understanding food consumption patterns.
- Promotional Offer Ideas: If Chipotle knows that customers like me, who order a lot in the summer but not in the winter, might want to shoot more promotional offers off in the winter to get me to come in more often. Also, by knowing that I won’t go more than 2 weeks without purchasing Chipotle, they don’t need to send me as many buy one get one free or discount deals as someone who comes in less frequently. They can rely on my purchasing habits to be consistent, while focusing on other customers to one day have the reliability that I do.
- Food Consumption Patterns: In today’s age where food consumption is a hot topic, this is a GREAT way to determine the amount of food that should be produced on a monthly basis. By understanding consumer food consumption patterns, Chipotle will be able to know that they will need more meat and grains in the summer, and not as much in the winter (at least for my type of customer). This will not only save on food production costs, but maybe lead them in the direction to make more green decisions and even making donations to hungry of food that would go to waste.
Conclusion
Understanding personal spending patterns can have a lot of benefits. Yes, I went a little overboard here due to my high interest of my Chipotle consumption, but using probabilistic methods to understand how often you go out to eat or make other frequent purchases can give you an idea of which places you visit most often and how likely you will re-attend that location. Time series forecasting can also project how much you will spend in the future to give you an idea of how much you should be making for any future saving plans that you would like. Personal budgets and corporations should understand as much of their spending data to make the best decisions in the interest of future savings and a better overall financial awareness.
