Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

Practice for Midterm

The objective of this document is to help you prepare for the midterm by leveraging material from the text chapters.

The midterm will be administered using a Peer Learning Platform that enables peer learning. More specifically, the exam will have three parts.

● Step 1 - Write Your Memo. You will have to run an R file (provided to you with the exam) and prepare/submit a memo that addresses the following three issues:

○ What data preparation was done and why?

○ What model/analysis was done and why?

○ What are the key findings of your analysis?

When you prepare a memo for a business audience, it has to be NON-TECHNICAL.

As soon as you are done you will have to submit your memo using the Peer Learning Venture software. After you submit the memo, there is going to be a break. During this time the Peer Learning software will prepare anonymized versions of each memo, and assign memos to students for review.

● Step 2 - Evaluate and provide constructive feedback to the memos that have been assigned to you.

Evaluate each memo using the following criteria:

Does the memo explain the data preparation that was done and why.

Does the memo describe the analysis that was done (model) and why.

Does the memo communicate the key findings in non-technical terms.

● Step 3 - Score feedback provided by classmates.

To do well in your exam you will need to learn/practice how to write a memo as well as learning how to provide constructive feedback. In the following paragraphs, I will provide you with examples of memos as well as feedback that was generated by students and/or myself.

Prof. Stratopoulos

Samples of Memo and Feedback PAGEREF _m0ssgqnppgoz \h 2

Example of constructive feedback (tesla: industry analysis) PAGEREF _qdceqdap8bkd \h 2

Suggested Practice Problem - work with your crew PAGEREF _yn1wnux8fpq7 \h 3

Example of complete memo and instructions on how to improve it (Tesla) PAGEREF _mfxhoa7yd4p2 \h 4

Example of constructive feedback (telco - short version) PAGEREF _7gun245byldc \h 5

Suggested Practice Problem - work with your crew PAGEREF _wdisoahpq4r \h 6

Example of complete memo and instructions on how to improve it (Telco) PAGEREF _j2d40qph5qdq \h 7

Suggested Practice Problem - work with your crew PAGEREF _qht3kcemifvr \h 8

Samples of Memo and Feedback

Example of constructive feedback (tesla: industry analysis)

What is the business problem/question that you are trying to address with your analysis (i.e., tb4IndustryAnalysis_quiz.R)? Provide a brief non-technical explanation of the business problem.

library(tidyverse)

dt0 <- read_csv("finStatements4All_FY2009-20.csv")

names(dt0)

# The ticker symbol for Tesla is TSLA

dt0 %>% filter(tic=="TSLA") %>% select(fyear, tic, conm, naics)

dt1 <- dt0 %>%

filter(naics==336111 & fyear>2009 & fyear < 2020) %>%

select(fyear, tic, conm, sale, at, oibdp, naics) %>%

mutate(ROA = oibdp/at, PM = oibdp/sale, TATO=sale/at)

dt1 %>% select(sale, at, oibdp,ROA, PM, TATO) %>% summary()

dt2 <- dt1 %>% filter(sale>=100)

dt2 %>% select(sale, at, oibdp,ROA, PM, TATO) %>% summary()

dt3_indROA <- dt2 %>%

group_by(fyear) %>%

summarise(ind_minROA = min(ROA),

ind_q1ROA = quantile(ROA, 0.25),

ind_medROA = median(ROA),

ind_q3ROA = quantile(ROA, 0.75),

ind_maxROA = max(ROA))

dt3_indROA

dt3_indPM <- dt2 %>% group_by(fyear) %>%

summarise(ind_minPM = min(PM),

ind_q1PM = quantile(PM, 0.25),ind_medPM = median(PM),

ind_q3PM = quantile(PM, 0.75),ind_maxPM = max(PM))

dt3_indPM

dt3_indTATO <- dt2 %>% group_by(fyear) %>%

summarise(ind_minTATO = min(TATO),

ind_q1TATO = quantile(TATO, 0.25),ind_medTATO = median(TATO),

ind_q3TATO = quantile(TATO, 0.75), ind_maxTATO = max(TATO))

dt3_indTATO

dt4Tesla <- dt1 %>% filter(tic=="TSLA") %>%

select(fyear, tic, ROA, PM, TATO)

dt4Tesla

dt5 <- bind_cols(

dt3_indROA %>% select(fyear, ind_q1ROA, ind_q3ROA),

dt3_indPM %>% select(ind_q1PM, ind_q3PM),

dt3_indTATO %>% select(ind_q1TATO, ind_q3TATO),

dt4Tesla %>% select(tic, ROA, PM, TATO))

dt5

In response to this question, a student prepared the following (mini-memo):

“The business problem that we are solving is we are looking at Tesla’s sales, asset turnover, operating income before depreciation, return on assets, profit margin, and total asset turnover from its fiscal years 2010-2019 and comparing it to the industry average. We are doing this because we want to see if Tesla is at a disadvantage, parity, or competitive advantage.”

The following is a example of constructive feedback:

“Overall, your answer to the question is correct in the sense that you have identified the most important issue, i.e., the relative performance of Tesla versus its peers (industry). The area where there is room for improvement is regarding the measures that we will use to make this comparison. In your answer you list a lot more than what we need for the analysis. For example, we list sales and operating income, instead of just ROA, profit margin, and asset turnover.

While the question does not ask for this clarification, there is a reason why we don't want to use sales and operating income. They don't control for size (i.e., larger companies will have more sales and operating income).This means that it does not make sense to compare firms of different sizes.

On the other hand, using financial ratios by nature of their design, i.e., by dividing by sales in the cases of profit margin or dividing my assets in the case of ROA and asset turnover, make it easier to compare firms of different sizes.”

Example of complete memo and instructions on how to improve it (Tesla)

A student prepared the following memo:

“The analyst is looking to see how well Tesla is performing in comparison to their competitors in the industry. The ratios the analyst will use to compare Tesla to industry competitors include: Return on Assets, Profit Margin, and Total Asset Turnover. The analyst initially started by looking to see what industry classification number that Tesla was in in order for the analyst to extract information from other companies in the same industry. With the other competitors extracted, the analyst can properly compare Tesla against competitors in the same industry.

Before comparing them, the analyst had to create the ratios of Return on Assets, Profit Margin, and Total Asset Turnover for the industry and for Tesla in order to compare. Once that was created, the analyst combined the data sets in order to easily compare.

The analyst noticed that Tesla was in the bottom quartile in ROA, PM, and TATO for most years, with some exceptions. Therefore, in most years, Tesla has not been doing well when comparing itself to the industry. However, in 2019, Tesla moved in the middle of the industry. This trend upward seems to show that Tesla is moving in the right direction.”

The following is a example of constructive feedback:

“May I suggest you think in terms of what matters to your audience and allocate time/space to reflect this.

1. The discussion on the data preparation could become simpler by removing the sentence that talks about "... by looking to see what industry classification number that Tesla was in in order for the analyst to extract information from other companies in the same industry," and state that you are going to focus on other firms in the same industry classification like Tesla. At the same time you may want to tell your audience what specification of the ROA and profit margin you have used (e.g., are you using net income, operating income, or any other specification in the numerator). This matters because the specification will let you focus on different accounting elements. Otherwise, the audience will have to guess.

2. The discussion on model/analysis could benefit from a bit more detail. For example, you can compare Tesla versus the industry min, max, median, q1, q3 or any other combination that you can think of. Which one did you pick and why? For added detail you may want to say that when you created the industry measures you have excluded Tesla. The part about bringing the data sets together, I would consider as optional.

3. The discussion of results assumes that your audience knows that you have chosen to focus on the quartiles, but this was not included in the model/analysis. You may want to explain either here or in the model discussion the reasoning behind your thinking. For example, you could say that I am going to divide the industry to the bottom quartile that captures firms that are in the position of competitive disadvantage. The middle section (i.e., above the q1 but below the q3) are firms that are in competitive parity. The top quartile are firms that are in a position of competitive advantage. Having this explanation will make the presentation of results easier. The added advantage is that you let the audience draw their own conclusions based on just the presentation of the results.”

Example of constructive feedback (telco - short version)

What is the business problem/question that you are trying to address with your analysis (see R file below)? Provide a brief non-technical explanation of the business problem.

library(tidyverse)

dt0 <- read_csv("telcoData.csv")

glimpse(dt0)

dt1 <- dt0 %>% mutate(SeniorCitizen = ifelse(SeniorCitizen == 0,

"notSenior", "Senior"),

Churn = ifelse(Churn == "Yes", "Churn", "noChurn"),

hghTenure=ifelse(tenure>median(tenure),"Yes", "No"),

hghMBill=ifelse(MonthlyCharges>median(MonthlyCharges),

"Yes", "No"))

dt2 <- dt1 %>% mutate(contractLength=ifelse(

Contract=="Month-to-month", "shortTerm", "longTerm"),

autoPayment=ifelse(

PaymentMethod=="Electronic check" | PaymentMethod=="Mailed check","manual", "automatic"))

dt2 %>% select(Churn) %>% table() %>%

prop.table() %>% addmargins()*100

dt2 %>% select(contractLength, autoPayment) %>%

table() %>% prop.table() %>% addmargins()

dt2 %>% select(contractLength, Churn) %>%

table() %>% prop.table(1) %>% addmargins(2)

dt2 %>% select(autoPayment, Churn) %>%

table() %>% prop.table(1) %>% addmargins(2)

model_Α <- dt2 %>%

count(SeniorCitizen, contractLength, autoPayment, Churn) %>%

mutate(p=n/sum(n)) %>%

group_by(SeniorCitizen, contractLength, autoPayment) %>%

mutate(cp=p/sum(p))

model_Α

In response to this question, a student prepared the following short answer:

“ The business problem that we are solving is we are looking at the data of customers and determining based on their tenure and how their billing works if they will churn or not churn.”

The following is a example of constructive feedback:

“The important thing is that you have identified the key business issue, i.e., we want to know the probability that a customer will churn or not. When I am looking at your answer I feel that you are "leaving money on the table" in the sense that you don't list all the variables that we can leverage to improve our predictions. In the context of our scenario we have the following three attributes: senior citizen status, their choice to have a short term or long term contract, and the decision to have automatic or manual payments. We want to know how these three attributes affect their decision to churn or not churn.”

Example of complete memo and instructions on how to improve it (Telco)

What is the business problem/question that you are trying to address with your analysis (see R file below)? See the R code for the example above.

In response to this question, a student prepared the following memo:

“The analyst is trying to understand which type of consumers from the Telco database has the highest probability of churning based on if the are seniors, their contract length, and what type of payment method they have. By understanding this, the analyst can provide Telco with information on which type of consumers to target more.

To start the analysis, the analyst had to prepare the data for it to work properly. Initially, the analyst changed the data to show if someone is a senior citizen or not a senior citizen. In addition, the analyst adjusted the contract length to be called short term or long term based on if the consumers had a month-to-month contract or a 1 year/2-year contract. Last, the analyst adjusted the payment method to show manual for electronic check/mailed check and show automatic for a bank transfer/credit card payment. To show the results, the analyst created a model to show the probability that someone will churn based on the different factors of being a senior citizen or not, having automatic or manual payment method, and if the contract length is long term or short term.

The key takeaway is that consumers that have the highest probability of not churning are consumers who have a long-term contracts and automatic payment. The individuals that are not seniors who have these characteristics make up 22.9% of the whole population and the individuals who are seniors that have these characteristics make up only 3.01% of the population.”

Instructor’s Feedback:

“This exercise is a great opportunity to showcase the fact that in CRISP-DM there is an arrow that goes back and forth from business to data understanding and preparation. Let's see how we can leverage this to prepare the memo and communicate the results.

The first sentence of your memo sets the agenda, i.e., we want to know what the probability is that a customer will churn. You can supplement it by saying that ... After we reviewed the data set we decided to focus on three factors/variables and see their combined effect on churn. More specifically, we will focus on the age of the customer (senior or not), the contract length and method of payment. Preliminary analysis has indicated that we may want to contrast customers with month to month contracts (i.e., short term) versus one or two year contract (i.e., long term). In addition to this we may want to contrast customers with automatic payment (i.e., bank transfer or credit card) versus manual payment (i.e., e-check or mailed check). Now that we have defined the new variables we can come back and re-define our business problem as follows: What is the effect of the three variables (age, contract length, automatic payment) on churn. Alternatively, we can say that given a customer's age (senior or not), contract length and automatic payment, what is the probability that the person will churn.

The advantage of using the second specification is that it serves as a segue to the analysis/model. This is a model that lets us see the percentage of customers in each possible combination out of our entire set of customers, as well as the conditional probability for a specific group (e.g., the probability of churn given that the customer is senior with short term contract and manual payment method).

Whenever I work on the preparation of a memo like this I always try to imagine myself giving a presentation to a team of senior managers. So for this last part, I picture myself standing in front of a screen that shows the results shown in model1 of the R file. Having specified my business problem and data preparation, as well as my model (analysis) I am ready to move to the communications piece. Please notice that without having to discuss any results I have already prepared my audience on what to expect and how to interpret my results.

The audience can read or hear the statement "... The key takeaway is that consumers have ..." and can easily understand its meaning. Looking more carefully at your communication sentence, I would like you to pay attention to the following issue. In the beginning you told your audience that we will use three variables (age, contract length, auto-payment) to predict churn. Yet, in your statement you list only two. Fix this part and your memo will be complete.”

Practice for Midterm

Table of Contents

Samples of Memo and Feedback

Example of constructive feedback (tesla: industry analysis)

Suggested Practice Problem - work with your crew

Example of complete memo and instructions on how to improve it (Tesla)

Example of constructive feedback (telco - short version)

Suggested Practice Problem - work with your crew

Example of complete memo and instructions on how to improve it (Telco)

Suggested Practice Problem - work with your crew