Instructor: David Schuff

Weekly Question #9: Complete by November 16, 2017

Leave your response as a comment on this post by the beginning of class on November 16, 2017. Remember, it only needs to be three or four sentences. For these weekly questions, I’m mainly interested in your opinions, not so much particular “facts” from the class! If you sign in using your AccessNet ID and password you won’t have to fill in the name, email and captcha fields when you leave your comment.

Answer one of the two questions below (not both):

  1. Name and describe a business question that you could answer using a decision tree. What data would you collect to perform the analysis? Don’t use an example we’ve covered in class.
  2. What advice would you give someone regarding how to select the right predictor variables for a decision tree analysis?

53 Responses to Weekly Question #9: Complete by November 16, 2017

  • A decision tree could be used to determine whether to offer someone a job or not. The data to collect would be the applicant’s age, experience, education level, pertinent skills, background, and salary requirements. The analysis of the data should determine the ideal candidate and whether to offer the applicant a job offer.

  • A business can use a decision tree to decide whether or not a customer would purchase from them again and if they should send them a catalog or not. To perform this analysis, they can collect information like the customer’s gender, recent purchases, total dollars spent, etc. To select the right predictor variables, I would suggest using variables that most impact the response rate.

  • There is a countless list of things that decision trees can be used for. For example, a company could be questioning how they can increase sales. This is when they would ask do we lower prices? Should we offer a loyalty discount? or do nothing? We can list out how much it would cost for each option and then compare them as in higher sales, or lower sales and the percent of each occurring.

  • What percent of Fox School of Business graduates will get a job after graduation?
    You could observe variables such as major, GPA, employment history, etc. This could prove if students with a GPA above a certain value are more likely to have a job offer after graduation. Additionally, it could prove whether or not having an internship influences your post-college employment.

  • Who buys premium vehicles? A car dealership could use a decision tree to sell vehicles more efficiently. They would need to collect demographic and financial data from sales to see what factors have previously mattered when buying a specific car. Maybe current salesmen are weary to let people test drive the premium model, and therefore lose the higher profit sale. Conversely, they could be wasting time trying to sell a premium car to someone who is not likely to purchase. With a decision tree, they can hone in on finding the most appropriate vehicles for prospective customers.

  • A business such as a gym can use decision trees to see what kinds of people go to the gym, and therefore figure out who to market to. For example, if a common thread is men, under the age of 45, with an income greater than $65,000, (of course there will be other factors as well, such as other genders, and other ages) then they could market to those people in the future. The gym will have a better understand of their customer base so they don’t waste their money on useless marketing.

  • My advice for determining predictor variables would be to take your time and examine the problem description. This will make it much easier to choose the correct variables that have the most impact and will provide useful information.

  • Based on a number of factors, auto insurance companies must chose who they want to insure to properly spread their risk. A business question you could answer using a decision tree is whether or not to insure someone. Data one would want to collect to perform the analysis for the decisions would include: age of the driver, number of miles on the car, gender and ethnicity of the driver, geographic location, number of past claims, and number of past automobile accidents.

  • A business question such as, “Does a visiter on our site buy a product?” Factors taken: previous visitor, previous purchases, how many products viewed, time of day visiting? All of these questions determine a buyer from non-buyer. If statistics are present in your decision tree analysis, pay close attention to the importance of the variables being tested. If any variables are far less significant that others, be hesitant to use these variables.

  • A company like Spotify could use a decision try to see how likely a person is to buy the premium feature. I would collect age, income, gender, and educational level. I picked these types of data because I think more college kids have Spotify than middle-aged people. I picked income because I think people who have an income or make more money are more likely to pay for Premium Spotify.

  • To answer the question of how to select the right predictor variables, the first thing to do is determine what is the top purpose of the decision tree analysis, such as having the relatively accurate result while saving the cost most or having the most accurate answer no matter how much it costs. Then we look at the difference leaf nodes to see which predictors bring the closest results we want in the purpose. Good predictors benefits the analysis most and stays closest to out goal.

  • One business question that can be answered is for the insurer whether to offer insurance policy for person who apply for their insurance product. Data can be collected from the applicants is their age, weight, income, pass medical history, or whether the person smoking or not. To find a right predictor, always compare the expected and observed results and used chi-square analysis and choose the one with the lowest p-value to be the best predictor.

  • A decision tree could be a popular method to solve/predict most business questions. An example would be for when a customer buys a certain product, what the likeliness would be of him or her purchasing another product that goes well together; similar to how Amazon does its “recommendations” when a customer is about to purchase something.

  • To pick the right predictor variables, you should choose the values with the most impact, ones that can divide the scenario in half, then you just keep breaking it down. For example, you can start off with their age; old or young, and then if they purchase the product; yes or no, and etc.

  • A decision tree can be very useful in making business decisions, especially when trying to gauge a potential customer. One way to gauge a potential customer could be with Apple and their collegiate discount bundle. They can test a populations gender, age, and they’re loyalty to Apple products when trying to decide if a customer will be likely to buy an Apple laptop for college. If a customer does buy the bundle, this can help Apple track what other products or potential services they would be interested in in the future.

  • A business question that could be answered by using a decision tree is whether a bank branch would pass an audit or not. For example, a financial institution or private auditing firm could collect random samples of various things, like IRA Savings accounts or cashiers checks, and see if they were set up correctly. Then, they could split their decision tree by the percentage of correct findings in each category to determine the probability of passing an audit for any given branch, depending on their results in each category. Furthermore, the advice I would give to people when picking the right predictor variables for their models would be to pick variables that have both a high correlation and causation, as related to the outcome variable.

  • You could determine what model of a car a potential customer would purchase from a decision tree. This could be a helpful marketing tool for car companies. Based on the demographics and other characteristics of the potential customer, the company could use a decision tree to determine a model of a car and a price range fits the customer.

  • Since sport franchises are essentially run as businesses we can look at some variables to determine the success of an NFL quarterback. NFL teams with an elite quarterback will be able to compete for multiple seasons and bring in millions in revenue. NFL teams can look at a college quarterback’s winning percentage, height/weight, age and division played in. The decision tree with these variables can predict if a quarterback will have a successful career or not.

  • Since sport franchises are essentially run as businesses we can look at some variables to determine the success of an NFL quarterback. NFL teams with an elite quarterback will be able to compete for multiple years and bring in millions in revenue. NFL teams can look at a college quarterback’s winning percentage, height/weight, age and division played in. The decision tree with these variables can predict if a quarterback will have a successful career or not.

  • A life insurance company could use a decision tree to answer the question: “Is it worth the risk of accepting this person’s policy and offering them coverage for life insurance?”. The data that would need to be collected in order to accomplish this would be the person’s habits, details on their diet and health and demographic information like age, job and area. The advice I would give to someone in regards to selecting the right predictor variables would be to make sure the predictor values chosen can make good inferences in context to the question being asked. I would also advise them to choose predictor variables that are relative to every person and/or customer.

  • My advice for selecting the right predictor variables for a decision tree would be to evaluate what the purpose of the model would be. Depending on what the model is going to be applied to, you can start with descriptive predictor variables that relate to the objective. Good predictive variables would also be easy to extract and analyze beyond the sampling of data points, otherwise it might make it difficult to apply the insights to all customers.

  • Choosing the right predictor variables for decision tree analysis varies depending on what the decision tree is analyzing. In general, however, the variables you chose should produce as many distinct and separate outcomes as possible. For example, depending on the situation, using gender as a variable will likely be a good variable in creating distinct and separate outcomes.

  • A decision tree can be used to determine where a company would like to set up a business; in this example, let’s say a gas station/convenience store. They could collect data to see how many people live in the immediate area, the demographics of those people, traffic patterns around the business, the other attractions in the same general area as the business, and the other gas stations/convenience stores in the area that would possibly compete for the same customers. By following through a decision tree, they could determine whether a particular spot is worth purchasing, or whether they should move to a different location.

  • If a business is looking to increase revenue, they could use a decision tree. To increase revenue, you make three options:
    1. do nothing
    2. launch loyalty card
    3. cut costs
    Next you would determine the probability of having either high or low sales in all three options. Lastly, you would determine the result of the high sale and the low sale.
    The data that you would collect to use would be the the cost, the probability, and the sale.

  • A decision tree could be used to determine whether or not to invest in a certain stock. You could collect data on the changes and fluctuations in stock price, how different classes of stocks are performing, index performance comparisons, how stocks are performing in each industry, etc. My advice to someone choosing predictor variables would be to no be fooled by strong correlations, because spurious correlations do exist. Just because two variables correlate heavily, that does not mean they are related because correlation does always mean causation.

  • Some advice that I would give to someone about choosing the right predictor variables would be to choose the ones that you think will output useful information. The variables should relate directly to the outcome variable and should give additional insight into the outcome.

  • Decision tree could be used by car dealers. The question would be “What type of a car someone would buy?”. Car dealers would collect and analyze the costumer’s income, age, marital status, children/no children and maybe credit score. For example, A 19 years old student who’s not married, have no kids and have a low income would be more interested in a used, cheap car.

  • The HR department from a corporate can use decision trees to see who are the potential candidate that they should hire. The data for the HR department to collect in order to perform the analysis would be graduated degrees, GPA, employment history (previous employment and internships), age, etc. This can help to analyze if the candidate would be a potential employee and how well they would perform on their offered position.

  • Decision trees could be used by sports analysts who predict whether a sports club will sign a particular player or not. The outcome variable would be whether a player is signed or not. The data that would be collected is the youth academy the player is from, player’s age, player’s fitness, player’s income in previous clubs, player’s scoring record etc. These variables can be used to form a decision tree to predict whether a sports club will sign a player or not.

  • A business question that could be examined with a decision tree is one of eSports related approaches, say, with gambling on teams and whatnot. You could put the related variables of each team, such as the amount of time their players practice and how long they’ve been playing the game, into the decision tree, and try to get it to spit out a prediction of the winner. I don’t expect it to be super accurate, but it’s something.

  • A decision tree could be used by a company to evaluate what types of insurance plans that they should purchase. Say there are options 1 Full Insurance, 2 partial insurance, 3 no insurance. Then determine the probability of a loss occurring and comparing the expected costs for each plan to make a decision.

  • Temple might be able to make use of a decision tree to figure out what how likely a student is to enroll at the university– the question then being, “how likely is this student to enroll at Temple in the upcoming semester?” Some of the factors would include distance from campus, accreditation of the program they are applying for, financial status, and whether they are just entering college or have attended another institution. Theoretically they could use this type of analysis would help them decide which students they may want to offer better financial packages to in order to increase the chance that the prospective student enrolls.

  • Decision trees could be used by any business that is taking on a new project. In making the decision about whether or not to accept the project, I would observe things like the cost of the project, its length, the amount of employees needed to work on it, etc. Collecting this data allows the business to determine if the project is a worthy investment or if their time would be more effectively used elsewhere.

  • A decision tree can be used by phone companies. The question would be “What type of phone plan would someone buy? Phone companies would collect and analyze the customer’s income, age, martial status, whether they have children or not. In addition, companies would analyze a customer’s cell phone history such as the data used per month as well the number of minutes they use per month. For example, a college student who is not married, has no kids, and a very low income would be more interested in a plan with a lot of data yet also cheap.

  • A company may want to use a decision tree when deciding which region to launch a new product in.

    The company may want to know what the average disposable income is for each region, regional consumer purchasing habits (large purchases/frugal purchases/etc.), regional adoption to new products, regional climate, etc.

    I would advise the company that a region who has a higher-than-average average disposable income, regional openness to new products, regional tendency to make large or frugal purchases and a more tempered climate would be more inclined to be receptive to the launch of a new product.

  • A business question that could be answered using a decision tree could be a business trying to determine whether they want to lease or buy a facility. Although there are many factors to take into consideration, I think the most important would be cost, capacity, and availability.

  • A business question that could be answered using a decision tree is who is most likely to buy a season ticket to a sports team. Some of the data you would want to collect is income, age, residence, gender, number of games attended, purchase history at team store, children, etc.

  • A decision tree could be used by Apple to answer a question such as, Who will buy the new iPhone X and be able to pay it off? the data collected would be age, gender, income, if a person owned an iPhone before or not, and if a person is subjected to upgrade their iPhone every time a new one comes out, and what phone carrier’s they’re in because not all carriers allow payment plans.

  • 1.) A business question that could be answered using a decision tree is whether or not you should continue selling a certain product in a specific area. To perform this analysis you would have to collect data on each customer from the area you are selling the product and see which customers are more likely to buy this product. For example if you are selling umbrellas, maybe someone who owns a car is less likely to buy one because they don’t have to walk much in the rain.

  • A decision tree could be analyzed if a company wants to hire a person. The variables could be age, gender, skills, experience, and education level. I would tell the company to think of their ideal candidate and the requirements you want. From there you will be able to create the correct variables, so you know you are getting an employee that will work and fit properly in your organization.

  • A business question one could answered using a decision tree is how to increase sales and profit. The data that could be collected to perform the analysis would be the past record of increased sales and profit from the sales team and from advertising, the amount of budget needed to invest to help increase sales and profits, and the increased profit and sales rate from past agency/dealer or new agency/dealer.

  • Colleges can utilize decision trees to take a better look at applicants for future school years. The data could include GPA, SAT and ACT scores, gender, ethnicity, past records, and family income. Using this data, you can determine who of the applicants to accept and who to reject.

  • A decision tree could be used to answer how likely it is that a person of a certain demographic will buy concert tickets for a specific music artist. Data I would collect would include gender, age, income, number of previous concerts attended, number of songs downloaded for a certain music genre, etc. I would suggest that someone who is more interested in a certain genre will be more inclined to purchasing tickets to the concert.

  • AA business question that can be answered using a decision tree is to What type of laptop would a person buy? The company would look at the demographics and different factors to create a decision tree. Which will help them figure out what laptops different types of people would buy.

  • A business question that could be answered using a decision tree is what is the likelihood of an undergrad student getting a job after graduation? The data you could collect could be their g.p.a., if they had a internship or job before and if it was relevant to their major, their major (based on college), the university that they went to and how active and involved they are on campus. The more data you collect, the more accurate your results from the decision tree will be.

  • A business question that could be answered would be whether a person is going to continue to pay a subscription for a streaming service after the free trial has ended. The data collected could be age, income, and whether the person has other streaming services. I would conduct surveys across ages and demographics and locations throughout the United States to see if people have streaming services and how much they would be willing to pay for the services. This would be an easy way to collect the data you would need.

  • I like the video game Overwatch and it has what’s called microtransactions, which if purchased, give loot boxes that unlock skins for heroes and other fun stuff. Since very few players actually spend their money on loot boxes, Blizzard, the company behind the game, could use a decision tree to predict how likely a certain player is to purchase them during a seasonal event. They could use predictor variables like total hours played, age, income, and whether they purchased loot boxes in the past or in the company’s other popular games.

  • I work at a trampoline park and corporate has been pushing us to get customers in more than once a month. We could use a decision tree to determine which customers are more likely to come in more than one time a month. We collect customer data through our waiver database and could use this for the analysis.

  • A decision tree could answer what the likelihood is that people will pay their student debt after graduation. The data that I would collect to perform the analysis would have to be income, marital status, and amount of student debt. For example, someone who gets a $30,000 post-graduation can pay off $10,000 in student debt even with interest, but if someone who makes $30,000 and has $50,000 in student debt, then it is unlikely that they would be able to pay it off immediately.

  • Decision Trees can assist a business in determining consumer predictability using outcome statistics from variables, which normally in most cases can include credit history, income, gender, education as predictors. A business looking to predict sales based on demographic factors can help provide an estimate which can assist a company in targeting a particular audience based on those variables within the demographics. I would suggest that someone using a decision tree should use input variables derive from a relevant granular process that result from an outcome variable.

  • A business like a car dealership could use a decision tree to determine the demographic that buys each model. Data like income, age, and children could be used for the analysis and help in figuring out a more exact target market for each.

  • A decision tree could be used to determine if a high school student should be admitted to a university. Data you could collect could be GPA, class level (AP and honors), SAT scores, extracurricular activities, and recommendation letters. Advice I would give for someone collecting this data would be to ensure there is a large sample of data to work from.

  • Decision trees, if not already taken advantage of, can be utilized by all sorts of industries. In the commercial insurance industry, an insurance underwriter usually has a near endless supply of accounts to write or pass on. Some information to analyze would be their location, revenue, management, good or service being offered, and their customer demographics.

Leave a Reply

Your email address will not be published. Required fields are marked *

Where and when do we meet?
Alter Hall 232
12:30 - 1:50 Tuesdays and Thursdays
Office Hours
David Schuff (instructor):
2:00-3:00, Tuesdays and Thursdays
Speakman Hall 210E and email (see my site)

Lauren Soentgen (ITA):
1:00 - 2:00 Mondays
11:00-12:00 Fridays
Speakman Hall 210D (and see her site)