Merkur Group Case Study

1
Merkur GroupCase Study

• Insights on how to prepare a research paper
• Petra Lapajne

2
Aims of this lecture

• Illuminate the research process
• Show on the Merkur Case how to carry out research
  
• Practice (Murphys law )

3
Why research?

• Identify opportunities and problems
• (exploratory research)
• E.g. SWOT analysis, new product development,
  competitor analysis
• Generate and/or refine company actions
• (problem solving research)
• E.g. marketing efficiency, investment options

4
Research process (1)

• Problem definition
• Development of research approach
• Development of research design
• Data collection and fieldwork
• Data preparation and analysis
• Report preparation and presentation

5
Problem definition business problem

• What problem does a company face?
• BUSINESS PROBLEM
• Merkur Group
• Division of the communications budget
• - among different communications
  channels/tools
• - between Merkur and Big Bang
• Among which communication channels should Merkur
  Group split the
• communications budget and how to account for the
  interrelatedness of
• Merkur and Big Bang communications?

6
Problem definition research problem

• What am I supposed to accomplish with my
  research?
• RESEARCH PROBLEM
• Merkur Group
• There are more ways to solve a business problem
• Efficiency of a promotion policy is only one of
  them
• Was the promotional policy effective in the past
  and what
• were the major determinants of in-/effeciency?

7
Research process (2)

• Problem definition
• Development of research approach
• Development of research design
• Data collection and fieldwork
• Data preparation and analysis
• Report preparation and presentation

8
Research approach

• I know what my task is. What do I do now?
• RESARCH APPROACH
• Deductive vs. inductive
• Merkur Group
• Find theory on sales promotions
• Refine research problem
• Develop hypotheses
• Test hypotheses
• Analyze results in the light of the existing
  theory

9
See what others have done

• http//www.nuk.uni-lj.si/nuk/odd_dostop_viri.asp
• - List of password-protected databases
• Useful databases EBSCO Host, Proquest, Emerald,
  Science Direct, etc.
• www.ef.uni-lj.si/enote/cek/menuzbirke.asp
• Undergraduate and graduate theses
• Working papers of the EF Research Center

10
Sales promotion theory

• Price and non-price promotions
• Different brands exhibit different response to
  promotions
• Different brands promotions have different
  impact on competitive brands sales
• Promotions affect competitive retailers sales
• Category, retailer, promotional consumer
  characteristics affect promotional efficiency

11
Refine problem definition

• What am I really supposed to accomplish with my
  research?
• RESEARCH QUESTIONS
• Merkur Group
• Does the promotional pricing policy affect sales?
• Which brands are affected by the promotional
  pricing policy the most?

12
Refine problem definition

• Promotional pricing policies of which brands have
  the largest effects on the sales of competitive
  brands?
• Do competitive retailers promotional policies
  affect the selected retailers sales?
• How do promotions of the two examined retailers
  interwind?
• Which are other factors that influence the
  success of a promotional pricing policy?

13
Hypotheses

• I understand what my task is. But how do I find
  answers to my questions?
• DEVELOP HYPOTHESES
• Merkur Group (16 hypotheses)
• Promotional effects (price, featuring effects)
• Brand switching (cross-price, cross-featuring
  effects)
• Store effects (size, cross-retailer effects)
• Day-of-the-week effect

14
Hypotheses (Example 1)

• Null hypothesis
• H0 Sales of brand i are not affected by
  promotional pricing (i.e., price effect is not
  significant).
• Alternative hypothesis, one-sided
• H1 Sales of brand i are positively affected by
  promotional pricing (i.e., price effect is
  significant and negative).

15
Hypotheses (Example 2)

• Null hypothesis
• H0 Brands in different price tiers are affected
  by promotional pricing to the same degree (i.e.,
  promotional effects are the same across brands in
  different price tiers).
• Alternative hypothesis, two-sided
• H1 Brands in different price tiers are affected
  by promotional pricing to a different degree
  (i.e., promotional effects are different for
  brands in different price tiers).

16
Hypotheses (Example 3)

• Null hypothesis
• H0 There is no cross-price effect when brands
  price promote (i.e., cross-price effects are not
  significant).
• Alternative hypothesis, two-sided
• H1 Cross-price effects emerge when brands price
  promote (i.e., cross-price effects are
  significant).

17
Hypotheses (Example 4)

• Null hypothesis
• H0 Competitive retailers featuring does not
  affect sales in a given store (i.e., competitive
  retailers featuring effects are not
  significant).
• Alternative hypothesis, two-sided
• H1 Competitive retailers featuring affects
  sales in a given store (i.e., competitive
  retailers featuring effects are significant).

18
Research process (3)

• Problem definition
• Development of research approach
• Development of research design
• Data collection and fieldwork
• Data preparation and analysis
• Report preparation and presentation

19
Research design

• I want to test the hypotheses. How do I do
  that?
• MODEL DEVELOPMENT DATA DEFINITION
• Merkur Group
• Influential variables
• Model selection
• Competitive brands, competitive retailers
• Precise definition of influential variables
• Methodology selection

20
Which variables affect sales?

• Demand curve Q f (P) p
• 
  D
• 
  
• 
  
  q
• Demand shifters income, p
  
• substitute and complement
  D1
• prices, advertising,
  D
• competitive actions, ? in
  D2
• consumer expectations,
  q
• holidays, weather etc.
  

21
Which model would best capture the sales I am
examining?

• qi f (pi, pj, Fi, Fj, Fr)
• pi . price of a focal brand i ? price effect
• pj . price of a competitive brand j ?
  cross-price effect
• Fi . featuring of a focal brand i ? featuring
  effect
• Fj . featuring of a competitive brand j ?
  cross-featuring effect
• Fr. competitive retailers featuring on the
  category level ? cross-retailer effect

22
Define product category

• Examined category television sets
• BUT a TV is not a TV
• E.g. normal (43) TVs, LCD TVs, plasma TVs,
  projector TVs
• Select category with the biggest sales in 2005
• BUT a normal TV is not a normal TV
• E.g. screen smaller than 51cm in diagonal,
  51-55cm screens, 72cm screens, larger than 72cm
  screens
• Select category with the biggest sales in 2005
• Chosen category 72cm screen television sets

23
Which competitors?

• Selection of competitive retailers
• Merkur Big Bang
• Biggest competitors Mercator, Harvey Norman
• Selection of competitive brands
• Best-selling brands for the two examined
  retailers
• Merkur Philips, Sony, Gorenje, Grundig
• Big Bang Philips, Sony, Samsung
• Why not include all competitors in the model?

24
Define selected variables (1)

• Item sales
• Daily quantity sales for the representative
  items of the selected brands in 2005 (pieces)
• Item price
• Actual daily price of the representative items
  for the selected brands in 2005 (SIT)
• Item feature
• Was the item featured on the day for which its
  quantity and price are measured? (0 no, 1 yes)

25
Define selected variables (2)

• Brand feature
• Was the brand featured on the day for which
  the item (or competitive item) quantity and price
  are measured? (0 no, 1 yes)
• Retailer feature
• Did the retailer feature any item in the
  selected category on the days for which item
  quantitiy and price are measured (0 no, 1 yes)

26
Methodology (1)

• 1. Multiple regression analysis
• Selection of the functional form
• additive qi a b1pi b2pj b3Fi
  b4Fj b5Fr
• b1 ?qi/?pi
• log logqi a b1logpi b2logpj b3Fi
  b4Fj b5Fr
• b1 ?qi/?pii (?qi/qi) /
  (?pi/pi) ?qi/?pi pi/qi
• b1 b1 (average pi / average qi)

27
Methodology (2)

• Example of price elasticity (b1) calculation from
  the price effect (b1)
• b1 - 0.000278
• If the price increases by 1 (1000) SIT, the
  quantity sold decreases by 0.000278 (0.278)
  units on average c.p..
• avg. PPhilips 80,000 SIT
• avg. QPhilips 0.621 unit
• b1 b1 (average pi /
  average qi)
• -0.000278(80,000/0.621)
  
• - 35.8
• If the price decreases by 1, the quantity
  sold increases by 36 on average c.p..

28
Methodology (3)

• 2. Logistic regression analysis
• 3. Non-parametric tests
• Multiple methodology approach in order to
• lessen the explected methodological problems in
  the field of sales promotions (heteroscedasticity)
  
• lessen the expected methodological problems in
  the field of durables (non-normality of data)

29
Research process (4)

• Problem definition
• Development of research approach
• Development of research design
• Data collection and fieldwork
• Data preparation and analysis
• Report preparation and presentation

30
Data collection fieldwork

• Where do I get the data?
• SECONDARY PRIMARY SOURCES
• Merkur Group
• Merkur management provided data
• Statistical Yearbook of RS

31
Research process (5)

• Problem definition
• Development of research approach
• Development of research design
• Data collection and fieldwork
• Data preparation and analysis
• Report preparation and presentation

32
Data preparation and analysis

• Hurrah, I have the data!
• Yes, but do they fit the chosen methodology?
• CLEANING THE DATA ANALYSIS
• Merkur Case
• Check the data
• Transform the data
• Analyse (1) test hypotheses
• Analyse (2) make sense of the results

33
Data preparation (1)

• Check the data for inconsistencies,
  misentrances, missing data
• E.g., an abnormal nr of TVs sold, an abnormal
  TV price
• Format data to fit the model youre using
• E.g., arrange in a single database each
  column represents one variable, each row one data
  unit (daily sales of brand i, daily price of
  brand i)
• Name variables and variable values
• E.g., Philips daily quantity sold possible
  values (0, 1, 2, 3, 4, 5) or (0 no Philips
  item sold, 1 Philips item sold)

34
Data preparation (2)

• Analyze data data distribution, descriptives,
  frequency distributions, outliers etc.
• E.g., data distribution not normal, outliers
• Transform data logarithms, standardization,
  reverse variable values etc.
• E.g., to make distribution more normal, to
  satisfy methodology requirements, to make
  interpretation easier
• Check missing data
• E.g., what is the reason a datum is missing?
• See e.g. Field, 2005 Discovering Statistics
  using SPSS

35
Test hypotheses (1)

• Test each hypothesis in a way specified in the
  methodology.
• If using a multiple methodology approach, test
  each hypothesis more than once.
• EXAMPLE
• Does a competitive brand promotion affect focal
  brand sales?
• Check cross-price and cross-featuring effects
  are they significant, what is their sign? (both
  MR and LR results)
• ? Check if there is a difference in the focal
  brand sales between periods with and w/o
  competitive brand promotion (non-parametric
  tests)

36
Test hypotheses (2)

• Organize results in a clear way (H0 cannot be
  rejected / H0 rejected H1 confirmed).
• Analyze results without giving justifications or
  explanations, i.e., describe results.

37
Explain the results

• A retailer is interested in category sales not
  only brand sales - aggregation of brand level
  results to the category level.
• Results obtained from brand models are inputs for
  a category sales analysis
• promotional effects
• cross-promotional effects
• cross-retailer effects
• How can a retailer increase category sales by
  price promoting the numerous brands in the
  category?

38
Research challenges
39
Research process (6)

• Problem definition
• Development of research approach
• Development of research design
• Data collection and fieldwork
• Data preparation and analysis
• Report preparation and presentation

40
Report preparation presentation

• Im done. Now I understand everything.
• Actually no. Your job is to make other people
  understand.
• CLEAR CONCISE RESULT PRESENTATION
• Merkur Case
• Structure your report
• Use graphics
• Let somebody proof-read it
• Use more graphics
• Be aware of the limitations business
  vs.research problem!

41
Final notes

• New ideas emerge when contemplating the old.
• Data work is tedious but its worth it.
• Try to understand what the computer is doing with
  the data youre feeding it.
• No statistical method can replace common sense.
• Cooperation is key.
• Presentation is everything.