Instructor: David Schuff, Section 003

Weekly Question #8: Complete by April 13, 2017

Leave your response as a comment on this post by the beginning of class on April 13, 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?

32 Responses to Weekly Question #8: Complete by April 13, 2017

  • Using a decision tree, one could determine whether or not a current customer is going to leave. With this information, you could determine which customers are at risk for leaving, and provide incentives to stay such as discounts. Data you could use to construct the tree includes time since last purchase, time as customer, frequency of purchases, and responses to questionnaires pertaining to satisfaction.

  • 1. A decision tree could be used to predict which new customers will buy a specific product. The outcome would be buy/did not buy. The data collected would include information collected on their current customers such as: age, gender, length they have been a customer, number of purchases, total dollars spent will the company, how they purchase (online/offline), location ( where they live), and any other information the company might collect about their customers.

  • Decision trees can be used for almost any business question. An example is a restaurant recommending certain items to you when you are ordering online. If you buy one item, the website may recommend another item to go along with it based on what you chose first.

  • Using a decision tree a university could predict whether a potential student will commute or live on-campus. The university has a lot of information on all of its students and by correctly predicting the number living in dorms it can be better prepared to serve those students.

  • If I had to advise someone in using the right predictor values I would tell them to use the values that cause the most disruption in the data. In other words, the values that split the data into buckets that you want to consider. For example, if you had a predictor value that only broke the previous subset into two more nodes and one of the nodes has a very small subset, that predictor value wouldn’t lead to something you didn’t already know. It doesn’t tell you more about the data so it’s prediction value is worthless. This is a situation one must avoid when using decision trees for analysis.

  • MBA programs might utilize decision trees to determine the probability of success that a student may have in the program if they are admitted. Specific variables that the universities could collect data on could include GMAT scores, undergraduate GPA, full-time job status, current company ranking and whether or not the student wants to attend full time or part time. These are all variables that could the university could use to measure the likelihood of student success in an MBA program.

  • Decision trees can be used to increase sales. You could use it to identify which customers would buy something if they received a coupon or offer. Good predictor values would be past sales, if they used a coupon code, age, and income range.

  • A decision tree can be used by an organization looking to budget for developing a new product or expanding to a new market. If the organization focuses on the only the key decisions and events that are important to them, then they can compare all of the consequences they wish to outline. Good predicator values involving the successes and failures of past product expansions and market expansions (i.e. development costs, marketing costs, increased revenue for new product/market) should be collected to perform the estimated analysis.

  • 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.

  • Insurance companies often use a decision tree to price premiums of insurance packages. For example, insurance companies will consider a insured’s age, current health condition, income and past insurance purchases record to determine premiums and insurance payments for disability and life insurance.

  • A decision tree can be used to get better estimates on what type of people are more likely to purchase store brand types of comestibles. Some good variables to use in this decision tree could be the income and whether or not the individual has children.

    • I think that this is a great idea for decision tree application. Consumers who buy store brand of one good are very likely to buy store brand for many goods, and vice versa. This way, appropriate coupons can be given out for products the consumer has not even expressed distinct interest in. I think a few more variables that may work would be if they bought anything in bulk, and education level.

  • A decision tree could be used to find out where an organization should introduce a new product. Will a product be more popular on the east or west coast? In the north or south? Is age a determining factor? Or socioeconomic status? Dietary restrictions? You should pick predictors that will influence the outcome the most.

  • An example of a business question we can solve with a decision tree could be how can it be used to predict future sales of a single product. By determining which customers would purchase this specific product, a business can better prepare itself to make more sales and profit. They can begin by collecting basic information on their customers such as address, gender, income, purchase history, preferred methods of purchasing, etc. This data will then go into determining what channels the product should be exposed to and what customers preferences should be applied.

  • One reason someone would use a decision tree for a business decisions is for determining whether to buy or rent equipment. A business would have to gather data about the costs of buying and renting. Also, they would have to compare the different places they can rent or buy from.

  • An example of how a decision tree could be used for a business would be for a technology company (ex. Comcast) using a tree to troubleshoot problems with technology. For example, there are some hardware or even software issues that people experience especially with Comcast. People could go online and answer one question which leads to another option and so on until the computer could give them the potential cause of their problem and how to fix it.

  • Decision trees can be used for any business. Decision trees would be in the event that I worked at an auto organization(TOYOTA) and needed to know which gathering of individuals would probably purchase the 4×4. The information I would need to know to find a solution would be age, gender , pay, and in the event that they have kids.

  • A question that could be answered using a decision tree could be whether or not a person would get accepted to a life insurance policy. It can be used to determine by using various factors like age, gender, health, income, and other attributes, allowing the potential customers to be placed into different groups. If they get denied into a certain group because of these qualities, they will not have the best protection in the event in which they do pass away.

    • Dean that’s a great idea for using a decision tree in business situations. I agree that health/life insurance would be a great opportunity to use this type of strategy in determining who gets what for specific groups of people. Typically people’s qualities and characteristics of themselves can represent just how well their health is gauged. Using a system such as decision tree analysis could streamline the way for making smart decisions on insurance policies.

  • A business question that could be answered with a decision tree would be if an owner of a small chain of pizza stores in Philadelphia wanted to hire a new team of delivery drivers. They may ask themselves questions in order to determine if the driver is at a high-risk for getting in a crash while on a delivery, which the store could be liable for. These questions include: have they been in an accident before, do they have car insurance, is their car new or used, are they the title owner of the car, do they have over or under 10 years of driving experience? These questions would hopefully enable the chain of stores to determine if the driver was high-risk, and that would lead to an outcome of them being hired or not.

  • An example of a business question that a decision tree could be used to answer would be if a customer will cancel their subscription to a video streaming service (i.e. Netflix). The company would gather data from its users such as how often they watch shows or movies, age, gender, what type of membership they have, and other other data they could get about the customer. They could use this to predict how likely certain types of customers are to cancel their subscription.

  • A business question that a decision tree could be used to answer is whether or not to rent an apartment out to a person. The data that can be gathered are the income, credit score, age, and in some cases, the references from past landlords. To choose the right predictors to answer the business question, one must need to ask if the predictor in question is relevant to the question that needs to be answered. For instance, income will be relevant in answering the above question, but something like gender does not.

  • An example of a business question that could be answered by a decision tree would be if a client of a car insurance company is likely to have an accident. They could predict this by collecting data including age, gender, frequency of driving, previous driving record and other related data to then tell what kind of clients are more likely to be engaged in accidents. I would advise someone to pick statistics that relate to usage, age, past driving history and gender to obtain the best decision tree.

  • The advice I would give to someone when selecting the right predictor variables for a decision tree analysis would be to make sure the variables do 2 things: maximize distinct outcomes and separate the outcomes. The distinct outcomes would mean that “a” does this and “b” does this based on whatever the variable is. Having the variable and no variance between the groups would be meaningless. The variable should also be able to split outcomes such as “a” does this while another “a” does not. That way there is no mix-up between similar entities.

  • An example of a business question that can be solve by decision tree is would a Taco Bell’s manager hire more employees. The manager would collect information on the time customers have to wait in line, the time it take for food to be make, the number of customers, whether the restaurant is clean or not. The manager can use these to determine if Taco Bell lack employees or not.

  • An example of a business question that could be answered using decision trees is, in the case of a cable/internet services provider, the likelihood of current customers to choose to upgrade their packages if a promotion/advertisement campaign were launched. The tree would consider a wide array of data, including: marital status, number of children, income, payment history, number of televisions, television use, etc. They could use this tree to determine how much should be invested in a campaign, assess the rate of return, and decide whether it would be a profitable decision.

  • 1. Although not exactly a business question, professional sports teams can use the decision tree to evaluate whether or not a player is worth drafting. In basketball, some data to collect for the analysis are a player’s points per game, efficiency rating, win shares ratio,… . Although the decision to draft a player can be more than just using data, the decision tree can provide a good overview on a player’s ability.

  • An example of a business using a decision tee is amazon and their recommended and paired purchases. amazon collects purchasing data from all of their customers and will pair items with your order that other buys have bought together. Amazon can also recommend products for you to buy based off your purchase and search history.

  • 1. A decision tree can answer the business question of what consumers are most likely to buy when they purchase items. This is used all over the internet. A company can gather information about your past purchases or what items you clicked on when browsing a website. By collecting buying patterns of a customer, you can find what they are most likely to purchase, and then recommend those products to the customer on social media or at the bottom of the website when your browsing.

  • An example of a business question that a decision tree would be used to answer is whether to build a small plant or a large plant for a Chemical plant industry. You would need data such as, the demand for the plants usage, the volume of the plant, would the plant be economically sufficient, how much sales each plant would generate, how much each plant would cost to operate, etc. Each decision and alternative would need to be evaluated to make the proper decision.

  • You could make a decision tree showing if a student will attend class or not. You could collect data on their age, gender, gpa, and social involvements. The best predictors would be the ones that have the largest one-sided answer. For example, in the scenario above, if the students’ gpa is less than 1.0, there may be some statistic saying 90% of students with that gpa don’t attend class. Then, it would be theoretically correct in 9/10 selections.

  • An example of using a decision tree would be related to my internship this past summer, I worked at an insurance data company. They could use a decision tree to select which prospect clients would be good to contract with and which of their existing clients they should/should not keep doing business with. They would gather information on the company’s financial information from the previous quarter/year, number of employees, time spent as a client, location relative to headquarters, and what they need (cyber loss data, general liability loss data, media contract, etc.). They could use this information to see who would be a good client to work with and which companies would not fit into the ideal client profile.

  • One way that businesses can utilize decision tree analysis is by using it to evaluate certain alternatives to a problem, specifically deciding whether to approve a development budget for an improved product. Assuming that the development would be successful, you might be given the competitive edge. However, if you are unsuccessful it may damage your market share. Decision trees are helpful in evaluating the likelihood of these outcomes.

  • A decision tree could be used to find out whether a customer or a group of customer is willing to purchase or try out the new product. The firm may have already collected data on what products or categories a customer bought, their usage volume, whether the customer has signed up for special membership. The company can perform decision tree analysis on consumers who have agreed to trial new products before to construct customer profiles for marketing (sending free samples) or research (uncover drives behind the eagerness for new products).

  • An example of a business that would use decision trees would be verizon. They would want to know which customers are more likely to continue using its service. They would collect data on how long the customer has been using verizon, which package they currently have, age, timely payments, number of complaints, etc. This will allow them to send promotions to customers they think will stop using their service.

  • When selecting predictor variables for a decision tree, I would advise someone to pick record fields that describe a subject in a general sense. This would mean that the particular field can and should only have a limited number of possible answers such as the number of children, income level, number of purchases made, etc. If one were to select a field that had unique identifiers, the analysis wouldn’t make sense because the end result would yield a large amount of nodes because there is no way to properly categorize the data. Your decisions on the variables and resulting tree should tell a story of a group(s) and leave meaningful information.

  • A decision tree could be used to help businesses find out which products in their company are the most profitable among consumers . The decision tree can answer questions on whether age plays a factor, or socioeconomic status. Will a product be more popular among one race or another? Cultures and Values? These questions can determine product profitability as well as how the company should target, place, and market certain products. The most important key component is choosing predictors that will influence the most positive (least error) outcome.

  • A business question that could be answered would be how likely someone would be to buy a item, like a shovel, according to the weather, season and other various variables. It would be very simple to use R to see what the past data tells us about the consumers and how they will react according to the climate of the market and weather.

  • When determining the proper predictor variable, one should asses all variables, their descriptions, and their relationships to each other. By logically following the dataset, you can associate variables with each other, and actually frame a series of questions that all could be the predictor variable. I like to ask myself “what am I trying to answer?” and “what do I need to answer it?”

  • A business question that a decision could be used to answer would be whether a person will or will not purchase auto insurance. The insurance company can gather data from individuals who are likely to purchase insurance and those who are not. Some factors they would use to determine the outcome, would be age, gender, how many cars they have, income, geographic location. Good predictor values would come from looking at their current users and establish what type of individuals are already purchasing auto insurance and then use that analysis to market towards customers that are not necessarily purchasing insurance.

  • A decision tree could answer the business question: Will customers return to our store if they have a a rewards card? By assessing factors like how recent a customer has bought, how much they purchased or categorical factors like job status could be used to predict if a loyalty/rewards card member will return to the store or by a specific product.

  • Core Question: Does alcohol correlated to the increase of sales on fast food industry?
    In this scenario, the practical and useful variables would be the age, income, how many drinks, gender, overall purchase amount on fast food, and other data that may be influential to the prediction of the questions. The fast food industry could use these variables to conduct a decision tree analysis.

  • An example of a business question that a decision tree could be used to answer would be if a business is deciding on building a small plant or a large plant. A business would have to do some forecasting, but this could be beneficial in the end, to see which plant would be more profitable for the business.

  • A business question that a decision tree could answer could be whether or not consumers buy certain products in pairs. You could use this information to help with the placement of products in stores or where you show ads.

  • My tutoring business can use decision tree to predict better outcome in students’ performance. By taking into consideration the student’s consistency and practice score – progress, and preparation length, we can find out if the student will achieve their targeted score.

Leave a Reply

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

Where and when do we meet?
Alter Hall 232
11:00 - 12:20 Tuesdays and Thursdays
Office Hours
David Schuff (instructor):
10:00-11:00, Tuesdays and Thursdays
Speakman Hall 207G and email (see my site)

Nodir Zakhidov (ITA):
Monday: 1:00-2:00
Wednesday: 1:00-2:00
Speakman Hall 207 and email (see his site)