ESCP Europe Turin Campus**We aren't endorsed by this school
Course
MARKETING DATA
Subject
Economics
Date
Dec 16, 2024
Pages
10
Uploaded by BailiffMusicChinchilla41
89 SESSION 10 – REGRESSION ANALYSIS AND FORECASTING 1.Recap Regression Analysis Recap Session 9 • What type of questions can we answer with a regression analyses? • What is the basic idea behind a regression analysis? • Formulate a binary regression equation. • Formulate a multivariate regression equation. • How can we interpret a significant !-coefficient? • What does the R2tell us? In a binary regression analysis, the intercept indicates:a) The observed value of the dependent variable when the independent variable is equal to zero b) The estimated value of the dependent variable when the independent variable is equal to zero c) The estimated value of the independent variable when the dependent variable is equal to zero d) The observed value of the independent variable when the dependent variable is equal to zero The total variance(for the kth observation) refers to: a) The difference between the value of the estimated dependent variable and the actual observed value of the dependent variable for the kth observation b) The regression line c) The difference between the mean of the dependent variable and the estimated value of the dependent variable for the kth observation d) The difference between the mean of the dependent variable and the observed value of the dependent variable for the kth observation Which of the independent variables (x) has a significant relationship with the dependent variable perceived stress (y) using a 95% confidence interval? a) Perceived Control b) Social Desirability c) Age d) None of the answers is correct
90 Please take a look at the table below. When social desirability is higher by one unit, how does perceived stress change? a) Given that the variable is insignificant we do not expect any change in perceived stress b) Lower by 0.149 units c) Higher by 0.149 units d) Lower by 0.468 units 2.CAUSALITY True or false? When I find a significant relationship between a dependent variable and an independent variable, then I can assume that one variable causes the other. a)True b)False CAUSALITY – REPETITION What is a causal effect? xày Identifying causal effects helps us to ➢Explain past events ➢Give appropriate policy recommendations !!! A regression coefficient !can be interpreted as a causal effect if the independent variable is ➢Exogenous (e.g., manipulated by an external force) → Recall: You can only derive policy implications based on....causal effects! 3.Robustness Hierarchical Linear Regression Models (HLM): Run different model specifications
91 To Show that Your Results are Robust... • Run different model specifications(“hierarchical linear regression models”) ➢Subsequently include your main independent variables and then possible interactions, your controls... • If available, use different measuresfor your independent and dependent variables • Do robustness checks with and without outlier Moreover, -If you are interested in causal effects, adjust your study design & analysis accordingly (as far as possible) -State R2 and adjusted R2-Use robuststandard errors (STATA “,rob” 4.Exercises Look at the slides
92 1/12/23 SESSION 12: REVISION AND PRACTICE ±Focus more on the applicationàsimilar to exercises done in class ±Definitions±Ex. Conversion rate (pick correct solutions) ±If conjoint analysis/regression based happens, then àput attention on xxx MOCK EXAM PART A EXERCISE 1 P-value: look at everything below 5% àtake these factors into account and neglect the others (age, gender, channel). If the intercept is significant, you have to count for it when predicting a value for a specific person; otherwise, you ignore it. a)Gender and Age was relevant b)b) Disregard the age, what would be the difference in spending between a male customer purchasing online and a female customer purchasing offline? [1 point] -Female/Offline spend 23.1 less -Male/Online spend 12.7 more -Female/Offline spend 35.8 more -Male/Online spend 10.4 lessFemale/offline spend 35.8 moreWe estimate the spending in the 2 cases, regardless age, and assess the difference Write down equations using the factors that matter Spending male/online = 52.3 + 12.7 * 0 – 23.1 * 1 = 29.2 Male is 0 and online is 1 Spending female/online = 52.3 + 12.7 * 1 – 23.1 * 0 = 65 Female is 1 and offline is 0 65 – 29.2 = 35.8 àfemale offline spend 35.8 more c)A customer at the age of 24, female, usually shops online, the last transaction occurred 37 days ago and is a follower on social media. How much would you forecast her transaction value to be? [1 point] -502.20 € -75.50 € -554.50 € -23.20 € Prediction purpose Check which factors are relevant and what is irrelevant is duration and social media here (you ignore them). WHAT MATTERS IS AGE, GENDER, CHANNEL. We estimate the spending for the specific case considering only the statistically significant coefficients(gender, age and channel) Spending = 52.3 + 12.7 (female) * 1 + 1.4 (age coefficient) * 24 – 23.1 (online) * 1 = 75.50 € You multiply the coefficient * variable (1 is the dummy variable = do we consider it? yes/no)
93 EXERCISE 2 Imagine that ESCP introduces a new online learning device and needs 400 chips for them. ESCP pays 6€ for each chip but they need to be changed twice per year. By selling these chips to ESCP, the supplier has a margin of 3€ per chip. A new technology of chips gets introduced to the market that has twice the endurance. The supplier’s margin for the new chips is 8€ per chips. The supplier of the chips wants to determine the value-in-use (VIU) of the chips according to different methods. a) What is the annual cost of the incumbent chip for ESCP? [0.5 point] o400 chips × 3€ = 1,200€ o400 chips × 2 changes per year × (6€ − 3€) = 2,400€ o400 chips × 2 changes per year × 6€ = 4,800€ oNone of the other answers b) Which of the following is the VIU for the new chips? [1 point] o16€ o14€ o12€ oNone of the other answers In addition to the information provided in the previous question, the new technology of chips allows a longer time between maintenance – 12 months instead of 6 months – and the cost of each shutdown is 1,000€. c) Which of the following is the shutdown cost for the new chips per year? [0.5 point] o500€ o1,000€ o1,500€ oNone of the other answers The correct answer would be 2.000€ [1000 * 2 shutdowns per year) d) Which of the following is the shutdown cost for the new chips per year? [1 point] o500€ o1,000€ o1,500€ oNone of the other answers The new chips require only 1 shutdown per year, so 1.000€ Annual cost of the incumbent chip = annual cost of the new chip 400*2*6€ = 400*1*VIU VIU = 4800 / 400= 12 €
94 e) Which of the following is the VIU for the new chips, including shutdown costs? [1 point] o14.50€ o15.00€ o16.20€ oNone of the other answers Annual cost of the incumbent chip = annual cost of the new chip (400*2*6€) + (1.000€*2) = (400*1*VIU) + 1.000€ VIU = 5800/400 = 14.50€ f) The lowest price ESCP can get for the new technology of chips is 14€ per chip. What would your recommendation be, based on VIU that considers the shutdown costs? [1 point] oPurchase the new technology of chips. oDo NOT purchase the new technology of chips. oYou cannot make a sufficient answer based on your calculations. VIU is worth 14.5 (it is you willingness to pay); if the the price is lower (because it is 14€)àpurchase; that would be a consumer surplus; in this case the value in use would be the perception of benefits and the other the perception of costs. You compare the price you can get on the market with the value in use. • If value on the market is > VIU àyou sell • If value on the market < VIU àyou purchase g) Now assume that the shutdown costs increase from 1,000€ to 1,200€. How would this affect the VIU of the new chips? [1 point] oNo proper answer is possible, based on the data. oVIU will decrease, and ESCP should not purchase the new technology of chips. oVIU will increase, but not enough to justify the price per chip. oVIU will increase, so that ESCP should purchase the new chips. Annual cost of the incumbent chip = annual cost of the new chip (400*2*6) + (1200 *2) = (400*1*VIU) + 1.200 VIU = 6000/400 = 15€ EXERCISE 3 a) What are the drivers of change fuelling this trend? (Please select all that apply) [1 point] oVirtual lives oDigital wellbeing oGlobalisation oAging population b) What new expectations are created by this trend? (Please select all that apply) [1 points]
95 oIncreased inclusivity àbeing part of the society oConvenience oDemocratisation of service oPersonalisation c) What are the basic needs satisfied by this trend (Please select all that apply) [1 point] oSelf-enhancement oInclusivity oPersonalisation oConvenience PART B: MINICASE Polkaly is a digital solution company. Their product is an interactive tool which allows the presenters to better engage with the audience. With the recent change in the market, they observed a surge in demand for the relevant tools to be used for distant meetings, especially in the education sector. They are thinking about introducing a new real-time project management tool which would allow the presenters to interact with and monitor the audiences in multiple teams simultaneously. While only having one product at the moment, they separate their clients into four strategic segments:a)Based on the information above, would you conclude whether entering the education segment with a new product would be a good idea? Please explain. [1 point] Relate to BCG matrix or General Electric matrix and then look at the 2 components The average annual growth is how the market evolves (external environment evolves) Segment size: how strong the company is in a specific sector àAPPLICATION OF THE BCG MATRIX Education as a question mark àpossibility to develop it into a Star To make better data-driven decisions, the company decided to look into customer profitability and consider the customer lifetime value per strategic segment:
96 b)Assuming you need to subtract 20% of the revenue for overhead costs, which of the strategic segments would be the most attractive to enter based on their customer lifetime value. Please present your calculations to support your answer. [2 points] CLV consultancy= 5000 * 80 % * 2 (lifetime is 2 years) – 1.000 (acquisition costs) CLVeducation= (3000 * 5 ) * 80% - 1.500 = 10.500 CLV public sector= (2000*3) * 80% - 2000 = 2800 CLV retail= (2500 * 1 ) * 80% -1000 = 1000 If you have the average lifetime, you don’t need retention rate! Revenue – 20% revenue (variable costs) = 1 – 20 % = 80% If the average lifetime is 2 years, which is the retention rate that corresponds to an average lifetime of 2 ? 50 % à1 / (1 – retention rate) = average lifetimeà1. / 1- 0.5 = 2 à1 / 0.5 = 2 Alternative solution 1)Estimate retention rates using the lifetime duration & Rr formula 2)Estimate CLVs Conjoint analysis: To be sure about the decision of introducing a new product at this time, the company ran a rating-based conjoint analysis (on a 1-10 scale) with the representative respondents from each segment. The results (part-worth utilities) are as follows:
97 baseline utility àis the same as the intercept in the mock exam c)Based on the results, what would be the best combination of product levels for each of the strategic segment? Please present your calculations to support your answer. [2 points] You always pick the level that has the highest part with utility Look at photos d)What is the most important feature for each of the strategic segment? Please present your calculations to support your answer. [1 point] Important: compare which is the best and worst option within these attributes and then compare these to the differences of all attributes. Where is the biggest difference YOU NEED TO PAY ATTENTION àto what is positive and negative. You either have positive or negative, but sometimes you may have both negative or positive. e)Among all the product attributes, the unlimited group function and the in-window chat display are Polkaly’s truly competitive advantages. Taking all the information you have in the case, which group(s) would you recommend to target, with what kind of product, and at what price point? [2 points] Look at solutions ADDITIONAL PRACTICE EXERCISES EXERCISE 1 A/B TESTING (a)True (b)False (c)True Important: t-test & chi-square test (metric vs categorical data) àknow definitions and when to be applied EXERCISE 2 Clustering àSegmentation results, cluster profile: demographics and visit frequencye- Rate level of attractiveness: Look at the Size of The Segment, Satisfaction Levels and Purchase Frequency ExCluster 1: second largest (N=222); satisfaction high (5.7) and high purchase frequency (55%) f-Applying GE/McKinsey Matrix: Co’s strengths, how is competition
98 In this case answer cannot be given EXERCISE 3: REGRESSION Delete new products (last one) àp-value > 0.05 (A)Which elements have the strongest impact? Look at the higher beta àOutlet numbers (2.12) (B)Online spending (C)You look at the sign in front of the coefficients. If it is negative, the connection btw the current and the previous sales is negative and so it goes down. (D)Model is not specifying timeàperhaps spending in different months in order to have a more detailed assessing of the spending; you could also look into more specific and detailed subcategories here EXERCISE 4: CONJOINT Nb target cluster 1: difference btw best and worst is 0,56 + 0.97 (0.56 – (-0.97)) b-sum of all the parts in the question and DO NOT FORGET the intercept. Do the same for the other clusters and look at which one has the highest utility! c-if you increase the price, the utility goes up (here, they prefer the higher price, total utility goes up VS normally you would have expected the contrary)