1. Market Intelligence Session 2 Luna Beer, Hypothesis testing, Chi square, Appropriate Statistical Tests

2. Agenda Luna Beer Hypothesis Testing Chi square Appropriate Stats 2

3. Luna Beer Case Summary Decision alternatives? – Vote Luna Beer – customers, how purchased? How will Gomez make decision? Inputs needed?

4. Approach to the Problem Calculate a Demand Forecast for the Company. Then calculate Break Even Volume and compare them. Demand Forecast = Industry Demand * Market Share for Luna Beer BEV = Fixed Costs / (Price – Variable Costs)

5. What information do we need for demand forecast and BEV? Demand forecast: – Market size (industry demand) – Market share Break even: – Fixed costs – Price – Variable costs

6. Luna Beer Case: Team Present Emphasis on: – What inputs did you need? – What reports did you buy to give you those? – How did you use the reports? – What was your recommendation?

8. Calculation of Industry Demand Method 1: Uses Reports A and B. Per capita beer consumption * population PopulationPer Capita Beer Consumption (gallons)** Industry Demand in 2013 Based on Entire Population 70,10031.3 gallons2,194,130 gallons Based on Population Over Age 21 45,50047.5 gallons2,161,250 gallons **Assumes straight line growth.

9. Calculation of Industry Demand Method 2: Uses Report E. “ Taxes Paid Approach ” Taxes Paid (at $.21/ gallon) Gallons Consumed 2011$399,0001,900,000 2012$435,2002,072,381

10. Market Share Projection Market Share Estimates are available in Report C. We estimate 25% market share in 2013. Demand Forecast = 25% * 2,161,250 gallons =540,312 gallons

11. Fixed Expenses (Year 1) Salaries: $450,000 Fixed, p. 3: $204,000 Interest on Loans at 10%/yr: $ 131,159 (see next slide) Total fixed, yr. 1: $785,159 – Note: does not include incentives, ads – Note: interest rate pulled out of hat to illlustrate

12. Investments The investments given in the case (Table A) fail to include estimates of cash and accounts receivable. Report F provides an estimate of the percentage of total assets needed at 16.3% $1,600,000 / (1-.163) = $1,911,589 -- will need to borrow $1,311,590 of it

13. Unit Contribution Price can be estimated using Report I. We assume that Luna is a premium beer, and can sell at a wholesale price equal to the average price of the top 4 beers listed ($3.11 for a 6-pack). This translates into $5.53 / gallon (128 ounces per gallon, 12 ounces per beer). In addition, kegs will be sold at a rate of 1/3 the gallons of bottles and cans. Price for kegs is 45% of bottle/can price.

14. Unit Contribution ClassificationRevenue Weight Wholesale Cost / Gallon Wholesale Price / Gallon Bottles / Cans3.0$4.44**$5.53 Keg1.0$2.00$2.49 Weighted Average $3.83$4.77 **The wholesale cost is calculated by multiplying the cost of goods sold (which from Exhibit F is 80.3% of sales) by the price per gallon. Unit contribution is therefore $.94 ($4.77 - $3.83)

15. Break Even Volume BEV = Fixed Costs / Unit Contribution = $785,159 / $.94 = 835,275 gallons Our demand forecast was 540,312 gallons. We will most likely not break even. Gomez should probably not invest in this business.

16. Total Research Required… 1. Reports A,B,C,F,I for total of $6400 2.Reports C,E,F, I for a total of $7300

17. Luna Conclusions Feasibility studies need data on: – industry demand, market share, investment, costs, margins. Break even analysis common. Know what will data look like before doing research (ask for dummy tables) Effort at problem formulation stage reduces later costs of doing research Secondary data is the place to start, but it’s usually flawed or not exactly what you need

18. Luna Conclusions (cont.) Trade off: can usually get 2 of 3: cost, speed, quality “Nice to know” info can not only add expense but be misleading Understand dummy tables and action standards My Excel version of solution on Sakai

19. Dummy Tables Two kinds of dummies: – Raw Data Dummy Table: require analysis, no action standard Example: Luna Reports A-I – Dummy Analysis: organizes output so that an action standard can direct a decision, conditional on data Example: profit in year 1 = [Total Volume in Market * Market Share * (Price – Marginal Cost)] - Year 1 fixed expenses

20. Action Standards Action Standards: – Prescribe actions on the basis of results from analysis dummy tables – Example: in Luna Beer, Go if NPV > 0, or 1st year Rev. > Fixed Expenses Break even by Year 5 – Other examples of action standards: send coupons to a segment if the expected response is 10% or higher Use new commercial if awareness is less than 60% Launch brand extension if we will break even in 5 years

21. Coming up… National Insurance Case: case is due on Thurs in class – individual assignment. Handwritten answers on handout in coursepack fine. – Wed 4-6, Danielle available in PC lab to help Colgate Oral Care Focus Group Case – Read “Using Focus Groups …” – Read case “Colgate Oral Care” – View steaming video on Sakai. Submit 2 slides by Wed. 10pm First quiz on Monday. Study guide coming soon.

22. Agenda Luna Beer Hypothesis Testing Appropriate Stats Chi square 22

23. Statistics / Hypothesis Testing: Step 1 State a null hypothesis, Ho Common nulls: – There is no demographic difference between the sample and the population – There is no difference between 2 groups – There is no association between 2 variables – Variable A has no effect on Variable B

24. Statistics / Hypothesis Testing: Step 2 Pick a significance level, e.g., a critical “p- value” at which you will reject the null H: – The P-value is the probability of finding the particular observed data assuming the null hypothesis is true “Standard” cutoffs for significant p-values are frequently cited as the following: – Significance: p <= 0.05 – Marginal Significance: 0.05< p <= 0.10

25. Statistics / Hypothesis Testing: Step 3 Observe your data, calculate your statistic and p-value Reject null or not – If the p-value is smaller than.05, we reject the null hypothesis – If the P-value is larger than.05, we “fail to reject” the null hypothesis.

26. A example with t-test 26 H o : There is no difference between men and women on attitudes toward Dove soap Test Results – Women average 6.2, men average 4.8 on 9 point scale – T-test statistic = 2.429, df=38 – P-value = 0.02 What should be our conclusion? Is this random sampling error or is there a significant difference?

27. A example with t-test 27 H o : There is no difference between men and women on attitudes toward Dove soap Test Results – Women average 6.2, men average 4.8 on 9 point scale – T-test statistic = 2.429, df=38 – P-value = 0.02 What should be our conclusion? Is this random sampling error or is there a significant difference? The prob that we would observe this large of a difference when Ho is true is the p-value. If the p-value is small, we reject Ho.

28. Agenda Luna Beer Hypothesis Testing Chi square Appropriate Stats 28

29. Chi-square Test Chi-square test is used for nominal data, to compare the observed frequency of responses to what would be “expected” under some specific null hypothesis. Two types of tests: – Goodness of fit: 1 factor, H 0 on category proportions – Test of independence: H 0 of independence in crosstabs 29

30. Nominal Data -- Observed vs. Expected Frequency Expected if random from customer base 54% M, 46%F Chi-Squared Goodness of Fit from National Insurance 30

31. 31 df = k-1 P>0.05, not significant

32. Conclusion: Fail to reject H 0 Conclude no evidence of sample bias Appears the variation is due to chance alone 32 df = k-1 P>0.05, not significant

33. Chi-squared Test of Independence In crosstab data, one type of null hypothesis is that there is no association between 2 categorical variables. Rejecting the null means the observed association is larger than would be expected if there is no association in the population Expected Proportions under independence, P(Row i AND Col j) = P(Row i) * P(Column j). Expected Frequency = Exp. Proportions*N =RowTot/N * ColTot /N * N = RowTot*ColTot / N 33

34. Example:  2 for Promotion x Purchase Promotion x purchase Observed Frequencies 34 Purchase Not purchase Promotion seen48654 Promotion not seen271946 7525100

35. Seen promotion x Purchase Expected Proportions Assuming Independence P ij = P i x P j Need to Calculate Expected Frequencies 35 Step 1 : Calculate the Expected Proportions Prob of having seen promotion =.54; not seen promotion =.46 Prob of purchasing =.75; not purchasing =.25 Purchase Not Purchase Promotio n Seen Promotio n not seen

36. # Cars x Income Expected Frequencies e.g., 0.54 x 0.75 = 0.405 x 100 = 40.5 36 Step 2 : Calculate the Expected Freq from Proportions and N Expected Proportion = Overall Row % x Overall Column % x N Purchase Not Purchase Promotio n Seen Promotio n not seen

37. # Cars x Income (Observed – Expected) Frequencies 37 Purchase Not Purchase Promotio n Seen Promotio n not seen

38. # Cars x Income Chi-Square Statistic 38 Purchase Not Purchase Promotio n Seen Promotio n not seen

39. Value of χ 2 compared to critical value of χ 2 for v degrees of information In this example = (2-1) x (2-1) = 1 x 1 = 1 df Since chi-squared = 12.08, and df=1 p <.05, Chi-Square Statistic Test 39 v = df = (# rows – 1) x (# columns -1) P-val = 0.001

40. Value of χ 2 compared to critical value of χ 2 for v degrees of information In this example = (2-1) x (2-1) = 1 x 1 = 1 df Since chi-squared = 12.08, and df=1 p <.05, Reject H 0 of no association between seeing promotion and purchase *Direction of effect? Chi-Square Statistic Test 40 v = df = (# rows – 1) x (# columns -1) P-val = 0.001

41. Back to gender bias in admissions… If these were a sample, how would I feel about drawing a conclusion from these numbers? I have 140 males accepted (14% of males) and 60 females accepted (7.5% of females accepted).

42. Conclusion Now? Chi-square = 19.01, p <.0001

43. Agenda Luna Beer Hypothesis Testing Chi square Appropriate Stats 43

44. Know when to use these statistics in market research: Chi Square (2 types) – Goodness of fit – Test of independence T-test – Paired sample – Independent samples Analysis of Variance (ANOVA) Regression

45. Know when to use these statistics in market research: Chi Square (2 types) – Goodness of fit: is a sample representative of population? – Test of independence T-test – Paired sample – Independent samples: Analysis of Variance (ANOVA) Regression

46. Know when to use these statistics in market research: Chi Square (2 types) – Goodness of fit: – Test of independence: is there a relationship between 2 nominal variables? T-test – Paired sample – Independent samples: Analysis of Variance (ANOVA): Regression:

47. Know when to use these statistics in market research: Chi Square (2 types) – Goodness of fit – Test of independence T-test – Paired sample: is there a difference between 2 means? (means come from 1 group) – Independent samples: Analysis of Variance (ANOVA): Regression:

48. Know when to use these statistics in market research: Chi Square (2 types) – Goodness of fit – Test of independence: T-test – Paired sample – Independent samples: is there a relationship between 1 nominal variable (2 levels) and 1 continuous (interval or ratio) variable? Analysis of Variance (ANOVA): Regression:

49. Know when to use these statistics in market research: Chi Square (2 types) – Goodness of fit: – Test of independence: T-test – Paired sample: – Independent samples: Analysis of Variance (ANOVA): is there a relationship between nominal variable(s) (>2 groups) and 1 continuous (interval or ratio) variable? Regression:

50. Know when to use these statistics in market research: Chi Square (2 types) – Goodness of fit: – Test of independence: T-test – Paired sample: – Independent samples: Analysis of Variance (ANOVA): Regression: is there a relationship between 2 or more continuous variables?

51. Know when to use these statistics in market research: Chi Square (2 types) – Goodness of fit: is a sample representative of population? – Test of independence: is there a relationship between 2 nominal variables? T-test – Paired sample: is there a difference between 2 means? (means come from 1 group) – Independent samples: is there a relationship between 1 nominal variable (2 levels) and 1 continuous (interval or ratio) variable? Analysis of Variance (ANOVA): is there a relationship between nominal variable(s) (>2 groups) and 1 continuous (interval or ratio) variable? Regression: is there a relationship between 2 or more continuous variables?

52. Examples Are awareness numbers for AudioTechnica head phones higher in Charlotte or Raleigh? Which of the following variables is the biggest driver of intention to buy Jif peanut-butter: self-reported attitude toward Jif, attitude toward Skippy, customer age, number of children? Are there enough Asian Americans in your study? Are people willing to pay more for Bratz dolls when they see it in a red package, blue package, or yellow package? Do men and women differ on brand of pizza purchased? Do customers report liking strawberry Jello or lemon Jello more?

53. Examples Are awareness numbers for AudioTechnica head phones higher in Charlotte or Raleigh?

54. Examples Which of the following variables is the biggest driver of intention to buy Jif peanut-butter: self-reported attitude toward Jif, attitude toward Skippy, customer age, number of children?

55. Examples Are there enough Asian Americans in your study?

56. Examples Are people willing to pay more for Bratz dolls when they see it in a red package, blue package, or yellow package?

57. Examples Do men and women differ on brand of pizza purchased?

58. Examples Do customers report liking strawberry Jello or lemon Jello more?

59. Examples Are awareness numbers for AudioTechnica head phones higher in Charlotte or Raleigh? (independent samples t-test) Which of the following variables is the biggest driver of intention to buy Jif peanut-butter: self-reported attitude toward Jif, attitude toward Skippy, customer age, number of children? (regression) Are there enough Asian Americans in your study? (goodness-of- fit chi square) Are people willing to pay more for Bratz dolls when they see it in a red package, blue package, or yellow package? (ANOVA) Do men and women differ on brand of pizza purchased? (test of independence chi square) Do customers report liking strawberry Jello or lemon Jello more? (paired samples t-test)

60. Agenda Luna Beer Hypothesis Testing Appropriate Stats Chi square 60

61. Qualitative research Focus groups In-depth interviews (one-on-one) Ethnography/observational – Overt – Covert All considered “Exploratory”, not decision research Outside bounds of BMR 61

62. Roles of qualitative research Insights Hypothesis generation Questionnaire development Underlying emotional benefits (“laddering”) Screening and refining ideas/concepts, etc 62

63. Roles of qualitative research Insights Hypothesis generation Questionnaire development Underlying emotional benefits (“laddering”) Screening and refining ideas/concepts, etc 63

64. Laddering exercise: in pairs Recent purchase Over $10 Went to store to purchase Not food Underlying emotional or social benefit? 64

65. Laddering Initial reason vs. deeper reason? Laddering up versus down (“why” vs. “how”)?

66. Roles of qualitative research Insights Hypothesis generation Questionnaire development Underlying emotional benefits (“laddering”) Screening and refining ideas/concepts, etc 66

67. For next time SPSS online tutorial – self-paced. – Do in computer lab with SPSS open to go through analyses as you listen to tutorial Get started on National Insurance Individual assignment