GET INSPIRED 2012 ANALYTICS & INNOVATIONNEW STATISTICAL METHODS IN    MARKET RESEARCH       solutions-2, Nicole Huyghe    ...
New methodsRe-instating existing methodsPushing boundaries                                2
MENU BASED CONJOINTCLUSTER ENSEMBLESK. CHRZAN MAKE/BREAK MODELKANO RESEARCHMULTI / SPARSE / EXPRESS MAXDIFFCBC + MAXDIFFTU...
NPD/Voice of customer    Dashboard/Score card                         How are we doingWhat are customers                  ...
Menu based conjoint
Conjoint objectiveUnderstanding which product  features drive purchase         decision
Business outcomeMaximising sales and profitthrough optimising product      characteristics
There are different types of    conjoint methods                               9
Choice-Based-Conjoint                        10
Choice-Based-Conjoint                        11
12
Choice-Based-Conjoint   Assumption: compensatory approach   Natural choice task   None option makes it even more realis...
Choice-Based-Conjoint   All respondents do get to see all attributes    and levels   Risk of focusing on a few attribute...
Non compensatory approach  Adaptive CBC (ACBC)Menu based conjoint (MBC)                            15
Adaptive CBC (ACBC)CBC tasks are tailoredBut still bundles                         16
Menu Based Conjoint (MBC)Extension of CBCFor multi-check/configurator choice tasksA la carte product and service configura...
Below you will see the Land Rover L560 together with all of the additional features at different prices.For each feature, ...
19
20
CBC              ACBC                   MBC                             Compensatory     (Non) compensatory   (Non) compen...
Product/Portfolio       Share of       Price sensitivity,   Impact of or on  optimisation        preference,     value of ...
23
24
Cluster ensembles
Segmentation objectiveIdentifying groups of customers with different needs, attitudes         and behaviour
Business outcome  Optimising products and  communication by betterunderstanding the (potential)         customers
Segmentation approach: cluster ensemble  answers the who and the why question
Traditional                Cluster ensemble  segmentation                  segmentationSegments differ on few         Segm...
The cluster ensemble process                               TraditionalTheme 1                        segmentations are    ...
A succesful segmentation fulfills 4characteristics Stability   Integrity   Accuracy   Size
StabilityPeople do not move around much from one solution toanother
Cluster Integrity – HeterogeneityThe segments do differ significantly on key dimensions
Cluster Integrity – HomogeneityThe people within a segment are very similar on keydimensions
Accuracysegment membership can be easily predicted – i.e. targetablesegments           Reality                            ...
SizeIdeally, there are no very large, nor very small segments
K. Chrzan’s make or    break model
Model objective  Quantifying the drivers ofoverall satisfaction/loyalty/NPS
Business outcome   Increasing sales throughimproved performance/loyalty      /advocacy scores
***** hotelDrivers of the customer experience  Staff                Restaurant  Room                 Lounge area  Cleanlin...
***** hotelDrivers of the customer experience  4 Staff                       8 Restaurant  9 Room                        9...
Regression/correlation analysis- drivers of overall customer experience- non-compensatory impact is not measured Keith Ch...
1   Overall experience score: 6    If < 7  ask if there were aspects so bad that they made    the whole experience awful ...
3       x Staff              8   Restaurant        9 Room               9   Lounge area        9 Cleanliness        7   Ho...
RESULTS – FOR EACH ASPECTStandard weightsPenalty weights for bad experiencesBonus weights for wonderful experiences Riche...
KANORESEARCH
Kano objectiveIdentifying the delighters and dissatisfiers of the customer           experience
Business outcome   Increasing sales throughimproved performance/loyalty      /advocacy scores
Understanding how performance drives satisfaction/loyalty                  Overall                Satisfaction            ...
Identify the ‘Must Haves’….                      Overall                                     Must haves                   ...
Identify the ‘Added Bonuses’….                      Overall                    Satisfaction                               ...
….and the ‘Key Desired’ elements                       Overall                     Satisfaction                           ...
Kano AnalysisCreating a better customer experienceEstablish customer driven action plan   Identify Critical Fixes   Tail...
Multi, sparse, express        maxdiff
Maxdiff objective Understanding which productfeatures are most important for           customers
Business outcomeMaximising sales and profitthrough optimising product      characteristics
Average number oftested items : 15 – 30 if 5 items on a screen 9 to 18 screensWHAT IF 120 ITEMS ??
EXPRESS MAXDIFFEach respondent only ~ 30 itemsThe 30 items are seen 3 timesEach respondent a different set of 30Fully rand...
SPARSE MAXDIFFEach respondent sees all itemsAll items are only seen onceFull utiliy set for each respondentShorter questio...
Only asks about persuasivenessWhat about uniqueness, believability?
MULTI MAXDIFF
Persuasive   1       2                                     4                  3       5                                 6 ...
Maxdiff + CBC
Weekday                On ipadWeekend                On iphoneBoth                   On pc                       On all1 m...
Online paper option 1   Online paper option 2   Online paper option 3  Weekday                 Weekday                 Wee...
Online paper option 1   Online paper option 2   Online paper option 3  Weekday                 Weekday                 Wee...
Most likely                    Least likelysubscribe                       subscribe             Headline news           ...
Most likely                    Least likely     subscribe                       subscribe                  Headline news ...
Turf + Surf
TURF objectiveIdentifying the set of products / product characteristics which   will reach most customers
Business outcome  Maximising sales and profitthrough optimising the product            range
Gelati& Sons
Gelati   Which 4 our of these 8 should we chose?& Sons         Respondents have a 1 if they would buy the flavour.        ...
GelatiRe        Results from all 100 respondents& Sons                        Unduplic                 #        ated      ...
< 30 items  TURFIf ≥ 30     TURF + SURF(SURF: Successive Unduplicated Reach and Frequency)
Nicole Huyghe                           nicole@solutions2.be                             www.solutions2.be         If you ...
Presentation nicole huyghe (advanced analytics) get inspired 2012
Presentation nicole huyghe (advanced analytics) get inspired 2012
Presentation nicole huyghe (advanced analytics) get inspired 2012
Presentation nicole huyghe (advanced analytics) get inspired 2012
Upcoming SlideShare
Loading in …5
×

Presentation nicole huyghe (advanced analytics) get inspired 2012

1,237 views

Published on

On Get Inspired 2012, Nicole Huyghe (solutions-2) gave a presentation about analytics and innovation: new statistical techniques in market research.

Published in: Business, Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
1,237
On SlideShare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
25
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Presentation nicole huyghe (advanced analytics) get inspired 2012

  1. 1. GET INSPIRED 2012 ANALYTICS & INNOVATIONNEW STATISTICAL METHODS IN MARKET RESEARCH solutions-2, Nicole Huyghe June 27, 2012
  2. 2. New methodsRe-instating existing methodsPushing boundaries 2
  3. 3. MENU BASED CONJOINTCLUSTER ENSEMBLESK. CHRZAN MAKE/BREAK MODELKANO RESEARCHMULTI / SPARSE / EXPRESS MAXDIFFCBC + MAXDIFFTURF (+ SURF) 4
  4. 4. NPD/Voice of customer Dashboard/Score card How are we doingWhat are customers and how can we looking for when increasebuying a product? satisfaction/loyalty?Are there customer How is my brand segments with positioned versus different needs? the competition? Segmentation Brand positioning Perceptual mapping
  5. 5. Menu based conjoint
  6. 6. Conjoint objectiveUnderstanding which product features drive purchase decision
  7. 7. Business outcomeMaximising sales and profitthrough optimising product characteristics
  8. 8. There are different types of conjoint methods 9
  9. 9. Choice-Based-Conjoint 10
  10. 10. Choice-Based-Conjoint 11
  11. 11. 12
  12. 12. Choice-Based-Conjoint Assumption: compensatory approach Natural choice task None option makes it even more realistic Easy for the respondent Fairly short exercise 13
  13. 13. Choice-Based-Conjoint All respondents do get to see all attributes and levels Risk of focusing on a few attributes only Might result in poor data Less engaging exercise Assumptions: features are pre-bundled 14
  14. 14. Non compensatory approach Adaptive CBC (ACBC)Menu based conjoint (MBC) 15
  15. 15. Adaptive CBC (ACBC)CBC tasks are tailoredBut still bundles 16
  16. 16. Menu Based Conjoint (MBC)Extension of CBCFor multi-check/configurator choice tasksA la carte product and service configuration  ACBC and CBC: pre-bundledExamples: Restaurant menus, Cars, Telecombundling, Insurance policies, Banking options 17
  17. 17. Below you will see the Land Rover L560 together with all of the additional features at different prices.For each feature, please indicate whether or not you would subscribe to that feature at that price.If you are not interested in any of the features at any of the given prices, please tick ‘None of the above’. Land Rover L560 £70,500 Dual View £800 Bluetooth with seat belt microphones £1,100 Bluetooth phone audio connection £900 Rear seat phone with cordless handset £800 None of the above TOTAL PRICE: £71,400
  18. 18. 19
  19. 19. 20
  20. 20. CBC ACBC MBC Compensatory (Non) compensatory (Non) compensatoryAssumption approach approach approach Choice between Choice between No bundling –Principle pre-bundled pre-bundled respondent can chose concepts concepts between levelsMimicks real purchase   processAverage length ~7 ‘ ~15 ‘ ~7’Engaging   Programming flexibility of   SawtoothEase of programming   design/questionsAnalysis options   Simulator:   What if, price sensitivty,optimum productSimulator:   Optimising bundlingProgramming and analysis average average expensivecost 21
  21. 21. Product/Portfolio Share of Price sensitivity, Impact of or on optimisation preference, value of features competition revenue, profit 22
  22. 22. 23
  23. 23. 24
  24. 24. Cluster ensembles
  25. 25. Segmentation objectiveIdentifying groups of customers with different needs, attitudes and behaviour
  26. 26. Business outcome Optimising products and communication by betterunderstanding the (potential) customers
  27. 27. Segmentation approach: cluster ensemble  answers the who and the why question
  28. 28. Traditional Cluster ensemble segmentation segmentationSegments differ on few Segments differ on manydimensions dimensionsLess diffentiated on Differentiation on bothtangeable aspects. More tangeable (who aspect)focus on the why than the and less tangeable aspectswho (why aspect)Difficult to target segments Targeting segments is essential and possible
  29. 29. The cluster ensemble process TraditionalTheme 1 segmentations are run on each of the different dimensions (suchTheme 2 as behaviour, needs, attitudes, demographics, …)Theme 3 ...Theme 9Theme 10 All individual segmentation Cluster results are used as input to an ensemble Ensemble methode to get segments differing on all dimensions
  30. 30. A succesful segmentation fulfills 4characteristics Stability Integrity Accuracy Size
  31. 31. StabilityPeople do not move around much from one solution toanother
  32. 32. Cluster Integrity – HeterogeneityThe segments do differ significantly on key dimensions
  33. 33. Cluster Integrity – HomogeneityThe people within a segment are very similar on keydimensions
  34. 34. Accuracysegment membership can be easily predicted – i.e. targetablesegments Reality Prediction 5 5 6 4 6 4 2 1 2 1 3 7 7 3 8 8 9 9
  35. 35. SizeIdeally, there are no very large, nor very small segments
  36. 36. K. Chrzan’s make or break model
  37. 37. Model objective Quantifying the drivers ofoverall satisfaction/loyalty/NPS
  38. 38. Business outcome Increasing sales throughimproved performance/loyalty /advocacy scores
  39. 39. ***** hotelDrivers of the customer experience Staff Restaurant Room Lounge area Cleanliness Hotel atmosphere Room size Internet/Wifi Breakfast Price Reservations
  40. 40. ***** hotelDrivers of the customer experience 4 Staff 8 Restaurant 9 Room 9 Lounge area 9 Cleanliness 7 Hotel atmosphere 9 Room size 2 Internet/Wifi 8 Breakfast 6 Price 7 Reservations+ Overall experience score: 6
  41. 41. Regression/correlation analysis- drivers of overall customer experience- non-compensatory impact is not measured Keith Chrzan’s Make or Break model
  42. 42. 1 Overall experience score: 6 If < 7  ask if there were aspects so bad that they made the whole experience awful If > 8  ask if there were aspects so good that they made the whole experience wonderful2 x Staff Restaurant Room Lounge area Cleanliness Hotel atmosphere Room size x Internet/Wifi Breakfast Price Reservations
  43. 43. 3 x Staff 8 Restaurant 9 Room 9 Lounge area 9 Cleanliness 7 Hotel atmosphere 9 Room size x Internet/Wifi 8 Breakfast 6 Price 7 Reservations4 1 x Staff 8 Restaurant 9 Room 9 Lounge area 9 Cleanliness 7 Hotel atmosphere 9 Room size 1 x Internet/Wifi 8 Breakfast 6 Price 7 Reservations
  44. 44. RESULTS – FOR EACH ASPECTStandard weightsPenalty weights for bad experiencesBonus weights for wonderful experiences Richer and more accurate model
  45. 45. KANORESEARCH
  46. 46. Kano objectiveIdentifying the delighters and dissatisfiers of the customer experience
  47. 47. Business outcome Increasing sales throughimproved performance/loyalty /advocacy scores
  48. 48. Understanding how performance drives satisfaction/loyalty Overall Satisfaction Kano Theory Satisfied allows us to derive how performance in an area drives overallPerformance Performance satisation – - Poor Outstanding Traditional Methods assume that there is always a Overall linear impact Satisfaction Dissatisfied
  49. 49. Identify the ‘Must Haves’…. Overall Must haves Satisfaction Satisfied No extra points if you get it perfect BUT people will be upset if itPerformance Performance – - doesn’t work. Poor Outstanding Dissatisfiers Critical to fix if Expected performance is / Must poor haves DISSATISFIERS Overall Satisfaction Dissatisfied
  50. 50. Identify the ‘Added Bonuses’…. Overall Satisfaction Added Bonus Satisfied People don’t Attractive expect it, so / Added there is no bonuses DELIGHTERS dissapointmentPerformance Performance – - if it is lacking Poor Outstanding BUT it delights people when it happens Create/ Identify USPs Overall Satisfaction Dissatisfied DELIGHTERS
  51. 51. ….and the ‘Key Desired’ elements Overall Satisfaction Desired Satisfied Desired Fall into both Attractive categories. / Added bonuses Delighters These are thePerformance Performance – - key areas for a Poor Outstanding company to Dissatisfiers Expected focus and / Must perfom on haves DELIGHTERS & DISSATISFIERS Overall Satisfaction Dissatisfied
  52. 52. Kano AnalysisCreating a better customer experienceEstablish customer driven action plan Identify Critical Fixes Tailor Offering to Customer Needs Create USPs Optimise Investment More satisfied, loyal & profitable customers
  53. 53. Multi, sparse, express maxdiff
  54. 54. Maxdiff objective Understanding which productfeatures are most important for customers
  55. 55. Business outcomeMaximising sales and profitthrough optimising product characteristics
  56. 56. Average number oftested items : 15 – 30 if 5 items on a screen 9 to 18 screensWHAT IF 120 ITEMS ??
  57. 57. EXPRESS MAXDIFFEach respondent only ~ 30 itemsThe 30 items are seen 3 timesEach respondent a different set of 30Fully randomised setsFull utility set for each respondentSample size >>Ideal for list with > 60 items
  58. 58. SPARSE MAXDIFFEach respondent sees all itemsAll items are only seen onceFull utiliy set for each respondentShorter questionnaireIdeal for list with < 60 items
  59. 59. Only asks about persuasivenessWhat about uniqueness, believability?
  60. 60. MULTI MAXDIFF
  61. 61. Persuasive 1 2 4 3 5 6 2 1 3Unique 6 4 5 1 4 5Believable 2 3 6 Average High Maxdiff score
  62. 62. Maxdiff + CBC
  63. 63. Weekday On ipadWeekend On iphoneBoth On pc On all1 month archive Paper copy week1 year archive Paper copy WE1 week archive No paper copyHeadline news National newsFinancial news Economic newsSports Stock exchangeCultural news BlogsLocal news …International news
  64. 64. Online paper option 1 Online paper option 2 Online paper option 3 Weekday Weekday Weekday On ipad On ipad On ipad 1 month archive 1 month archive 1 month archive No paper copy No paper copy No paper copy Financial news National news Financial news International news Blogs Local news Sports Cultural news Sports Local news Local news Local news Blogs Sports Eonomic news Stock Exchange £9/month £13/month £15/month 
  65. 65. Online paper option 1 Online paper option 2 Online paper option 3 Weekday Weekday Weekday On ipad On ipad On ipad 1 month archive 1 month archive 1 month archive No paper copy No paper copy No paper copy Financial news National news Financial news International news Blogs Local news Sports Cultural news Sports Local news Local news Local news Blogs Sports Eonomic news Best to reduce list Stock Exchange before conjoint 
  66. 66. Most likely Least likelysubscribe subscribe  Headline news Financial news  Sports Cultural news
  67. 67. Most likely Least likely subscribe subscribe  Headline news Financial news  Sports Cultural newsThe top 5 items for each respondent arethen passed to the conjoint  tailored
  68. 68. Turf + Surf
  69. 69. TURF objectiveIdentifying the set of products / product characteristics which will reach most customers
  70. 70. Business outcome Maximising sales and profitthrough optimising the product range
  71. 71. Gelati& Sons
  72. 72. Gelati Which 4 our of these 8 should we chose?& Sons Respondents have a 1 if they would buy the flavour. R1 1 1 0 0 0 0 0 0 R2 0 1 1 1 0 0 1 0 R3 0 0 1 0 0 0 0 0 R4 1 1 0 1 1 0 0 1 R5 0 0 0 0 0 1 0 0 There are 70 different ways to choose 4 flavours from these 8!76
  73. 73. GelatiRe Results from all 100 respondents& Sons Unduplic # ated Flavours Reach Flavours 1 65% 2 80% 3 90% 4 95% 5 100% = with this selection of 5 flavours, they please all respondents 77
  74. 74. < 30 items  TURFIf ≥ 30  TURF + SURF(SURF: Successive Unduplicated Reach and Frequency)
  75. 75. Nicole Huyghe nicole@solutions2.be www.solutions2.be If you have anyrisingquestions

×