5 marks

1)a) A campaign was conducted where the sales during the campaign was $100000.

A marketing mix model indicated the following weights: 40% direct mail,30%

programmatic advertising and 30% email.

Provide a measurement template and results that would indicate that direct mail program was successful.

b) What is one advantage of direct mail over social media from an analytics perspective

5 marks

2)Answer True or False

a) The marketing research discipline and its insights should be integrated with the data science discipline and its insights.

b) RFM is a Machine Learning Technique.

c) Power BI cannot be used to produce data visualization reports

d) Deep learning models require large volumes of data to be successful.

e) Creating frequency distribution reports would occur in the first stage

5 marks

3)A) Of the 5V's in Big Data, what are the two V's which are most recent as concerns for data scientists?

b)) You need to create a report which looks at average spend of customers that earnover $50000, are male and are college educated.What are the metrics and dimensions?

5 marks

4)a) Provide one advantage and one disadvantage when comparing deep learning (AI) to traditional models.

B) A model which has been checked and validated for Q/C issues. Yet scores are vastly different between development and implementation. What would you do?

5 marks

5)Create a final response model report where you have the following information:

Response=A -B1*education-B2*spend+B3*live in Toronto+B4*income;

Partial R2

income: .05

education.03

live in Toronto: .015

spending: .005

Note: the final model variable report should contain the variable name, impact on the model, and % contribution to the model

5 marks

6)a) As a business marketer describe why in one sentence, describe why you would prefer cluster segmentation to VBS segmentation.

b)A company conduct some segmentation now during a COVID19 environment. What would you recommend (VBS or Cluster) and in one sentence describe why?

5 marks

7)a) No analytics has been done on visitors to a company's website. What would be onerecommendation and in one sentence describe why?

b) Why mightindexing be used over predictive modelling as a targeting tool?

5 marks

8)What are the 4 stages of data mining?

5 marks

9)What Stage of data mining/data science are we in?

a) Designing a measurement matrix for an upcoming marketing campaign

b) Running factor analysis

c) Operationalizing your models within a cloud environment

d) Prioritizing business opportunities

10 marks

10)You have a promotion tablewith the following information:

-Account ID

-Promotion ID

-Promotion Date

Note: this is a many to one table with multiple promotions peraccount

Create 5 derived variables and write your logic in how each variable would be created. 1-2sentences per variable.

5 marks

11)a) A response model had regions as the strongest variable accounting for over 80% of the model's power.? What might have been done to improve the overall targeting of response behaviour

b) You are conducting an analysis to a direct mail acquisition program. What would be the match key used in matching responders back to a direct mail marketing program.?

5 marks

12)a) You are evaluating the investment of whether to conduct an acquisition program. You have the total number of customers. What are the 2 metrics that you would need to determine incrementality and the Cum. ROI of the program.

b) You have 3 models that are required to be built. What two things would you look at in prioritizing these opportunities?

5marks

13)a) Why is range a weaker metric than standard deviation to measure the dispersion of data

b) In two sentences, describe how you might use factor analysis in a predictive modelling exercise

10 marks

14)The following is a responder profile

-low tenurei.enew customers

-high spenders

Meanwhile 2 variables listed below are not in the responder profile

-age

-education

Produce a correlation matrix and Exploratory Data Analysis reports for each of the 4 variables.

Note: the correlation matrix should contain the variable name, the correlation coefficient and the statistical significance. The numbers should directionally make sense given the above information.

10 marks

15)a) What technique can be used to derive new variables from character level variables that contain sparse values. Describe in one sentence why you would use it?

b) What have been the two developments to help make AI and deep learning become very successful modelling routines.

c) Why is factor analysis a critical technique in clustering?

d) What would be one useful analytical outcome in using VBS segmentation in an unstable data environment.

5 marks

16)You are asked to demonstrate that model A is a better response model than model B. Produce an AUC curve or gains chart/decile table that demonstrates this benefit. You can make up your numbers

5 marks

17a)You have been asked to validate a model. Why might a decile chart be more appropriate as the validation tool over an AUC curve.?

b) What is one difference between text mining and traditional data mining ?