Presentation on theme: "Prepared by: FRC Research Corporation April, 2012 SPARKLING WATER LINE DEVELOPMENT AND OPTIMIZATION - Research Proposal - Prepared for:"— Presentation transcript:
Prepared by: FRC Research Corporation April, 2012 SPARKLING WATER LINE DEVELOPMENT AND OPTIMIZATION - Research Proposal - Prepared for:
Nestle Waters North America currently markets a number of carbonated water products with varying: Sources (Domestic vs. Imported) Brands (Poland Spring, Perrier, San Pellegrino), Flavors (Unflavored versus flavored with Lemon, etc.) Sizes (Single serve vs. multiserve) Pack configurations (Single Pack vs. Multipack) Package types (Plastic bottles vs. Glass bottles vs. cans). To date, each individual brand’s marketing efforts have been driven by each brand’s needs, and the different brands have been operating independently of each other. As the division has become more focused on sparkling products, and There is a desire to take more of a portfolio approach in managing the company’s sparkling entries, with the ultimate objective being to identify those products that, in combination with each other provides the maximum opportunities for both NWNA and its retail partners. BACKGROUND AND OBJECTIVES 2
The recommended approach will need to examine the many different combinations of products that could be combined in any given retail environment. The specific approach will ultimately rely on the complexity of the product array being considered for the study. If we limit the scope of the study to NWNA offerings we can proceed with a simpler, more straightforward approach. One limitation of limiting the investigation to NWNA products, however, is that we are looking to maximize potential without having any idea of competitive interactions. There might be some products within the NWNA array that are highly appealing when solely compared to the other NWNA products – but if consumers still prefer the competitive entry to the NWNA alternative the apparent incremental appeal would NOT be seen at retail. If, on the other hand, we include competitive products the sheer scope of the array would require a more sophisticated statistical design and analysis. But this will ultimately yield a potentially more powerful analysis which would maximize the NWNA potential within a competitive array – as would be the case at retail. The proposal will describe both options as well as indicating their respective differences and costs. They share many of the same philosophies and analytic approaches, and much would be common regardless of the approach ultimately selected. APPROACH 3
Based on our prior experience in the category, it seems as if consumers make a flavored/unflavored decision first. And then if they opt for a flavored product they then make the specific flavor decision. For the purposes of this research it would be beneficial if we can focus the investigation on flavored versus unflavored…we would, of course indicate what specific flavors are available under the different scenarios – but we would only need to ask about the specific flavors that a consumers would be interested in buying once. This, in effect, reduces the number of stimuli by about half and does not present any insurmountable analytic sacrifices in the process. The cost estimates to be provided will ultimately hold regardless of how we allocate the interviews. We can conduct all of the interviews equally across regions, we can allocate the number of interviews for each region as desired, or we can customize the sample allocation to best reflect NWNA needs – and do extra interviews in the key northeast and west coast markets, and fewer in the remaining regions. NOTES 4
If we utilize this approach, all respondents will be shown the array of NWNA sparkling products. They will be asked to rank all of the alternatives in terms of their relative level of appeal (as indicated on the next page). They will then be asked how likely they would be to buy the products that they were relatively more interested in using a 5 point purchase interest scale. (Continued) OPTION 1 – “TRADITIONAL” LINE OPTIMIZATION
MOST LIKELY TO BUY 8 LEAST LIKELY TO BUY 8 4 3 4 A “Q-Sort” of appeal for all items – this is a sequential series of exercises that allows creation of a preference-pyramid for each respondent. The following example assumes there are 30 individual items in the array. Respondents will be shown all of the alternative products simultaneously and asked which half they would be more interested in buying. They will then be asked which of these they would most want to buy. Then, among this top quarter, they will be asked to again sort their favorites until they rank their top 4 choices. Finally, they will be asked to reexamine the varieties that they had not originally selected as their favorites, and to choose the varieties in which they were least interested. 1 2 OPTION 1 – “TRADITIONAL” LINE OPTIMIZATION
OPTION 2 - DISCRETE CHOICE MODELING 7 The discrete choice modeling approach (DCM) is a more rigorous statistical approach that can handle a broader range of inputs than those in the “traditional” approach – and it allows greater analytic flexibility. It is, on the other hand, a little more expensive and takes a little longer to accomplish because of the need for the involvement of a statistician in both the created of the study design and in the analysis. Respondents will complete a multi-step exercise where they will be shown several products at a time and asked to distribute 10 purchases across those products with the assumption that they went to the store to buy sparkling water and the indicated items were the only ones available. They will be asked to do this a number of times, with the specific array of products varying for each iteration. If none of the products shown are of interest or consumers are unwilling to assign all 10 purchases across the items presented, they will be able to assign all or some of their 10 purchases to a “none of these” option The exercise will continue in this manner until the model runs its course The exact number of iterations will be based on the number of products that will be explored. (Continued)
OPTION 2 - DISCRETE CHOICE MODELING 8 Whereas the basic metric for analysis in the “traditional” LOT approach is purchase interest, this analysis ultimately utilizes “share of choice” as its key analytic element. “Share of choice” is a simulation of potential market share given 100% awareness, parity product performance, and no merchandising activity at retail. It is a clean measure of product potential, as it assumes all other things are equal. This approach allows us to not only determine the appeal of each alternative and every possible combination of alternatives. It also allows us to estimate the potential interaction of any given “line-up” of NWNA products with key competitive entries. It will ALSO allow us to do this among various behavioral, demographic and attitudinal subgroups.
Questioning Technique – an example 9 Imagine you went to the store to buy sparkling water and these were the products that were available. You can purchase op to 10 individual items -- you can make all 10 purchases of one product, OR you can spread your 10 purchases over as many products as you like. You can buy all 10 products, you could buy fewer than 10 products OR you can decide not to buy any of the products at all -- it is all up to you. Just make sure that ALL OF THE NUMBERS ON THE SCREEN, INCLUDING THOSE IN THE "NONE" BOX, ADD UP TO EXACTLY 10. So, if you choose to select fewer than 10 products, just make sure to indicate the number not chosen in the "NONE OF THESE" box on the screen. Option 2 - DISCRETE CHOICE MODELING (Cont’d)
ANALYTICAL APPROACH Regardless of which questioning approach we pursue, we will be able to conduct a many of the same analyses – a series of separate analytic efforts, that, in combination with each other, provide the optimum combination of alternatives that will maximize consumer appeal. Thoughtful cross tabulations will provide information about the level of appeal of each of the current and proposed varieties both overall and among key demographic, attitudinal and behavioral subgroups.
ANALYTICAL APPROACH (Cont’d) In addition, we will conduct several multivariate analyses: A Johnson’s Hierarchical Cluster Analysis Perceptual mapping Consumer segmentation based on variety preferences TURF analyses Detailed descriptions of these analyses follow.
ANALYTICAL APPROACH (continued) A JOHNSON’S HIERARCHICAL CLUSTER ANALYSIS will provide a “decision tree” reflecting the criteria consumers use in accepting or rejecting specific varieties. This is useful in understanding the sequence of decisions consumers make as they are selecting a product for purchase. An example of this analysis follows.
13 Lemon Reach = 35% Avg DWB= 25% Reach = Net % of ‘Definitely Would Buy’ Any Variety in Bucket Avg DWB = Average % of ‘Definitely Would Buy’ Across All Varieties in Bucket All Others Reach = 60% Avg DWB= 35% RSW’s Reach = 52% Avg DWB= 25% Perrier Reach = 58% Avg DWB= 11% Sam Pell Reach = 36% Avg DWB= 16% Other Familiar Reach = 35% Avg DWB= 25% Unfamiliar Reach = 28% Avg DWB= 28% Domestic Unflavored Flavored International Unflavored All Reach = 55% Avg DWB= 20% SOUP LOT Decision Tree Domestic Flavored
ANALYTICAL APPROACH (continued) A MAP OF APPEAL will also explore interactions between potential varieties – but simultaneously across the entire array rather than sequentially as in the hierarchical cluster analysis. The resulting map identifies varieties whose appeal is duplicative and those that represent untapped opportunities.
ANALYTICAL APPROACH (continued) A CONSUMER SEGMENTATION based on interest in specific varieties usually provides the most useful part of the analysis. In most flavor or variety driven categories there is no single universal model of consumer preference or interest. Rather, there are a number of different types of consumers that share the way that they tend to make purchase decisions. By identifying and isolating these separate consumer groups based on their own patterns of preference we can both gain a better understanding of category behavior and, in the process, identify incremental volume opportunities that “get lost” in more aggregate analysis. Once these segments are identified we have the option of profiling them with all of the other information contained in the study to help prioritize the opportunities that each presents.
ANALYTICAL APPROACH (continued) Finally, TURF analyses will be conducted to measure both the “breadth” (net reach) as well as “depth” (number of SKU’s would buy) of appeal of any potential line of products. These analyses will incorporate variety insularity results as well as anticipated volume in order to predict the relative incrementality of each proposed variety. The TURF analysis can also be used to determine the optimal number of varieties for the line based on an examination of the wear-out in reach as varieties are added. Examples are on the pages that follows.
Penetration Number of Varieties in the Line % = Net Penetration ( ) = Number of Choices Would Buy Net penetration starts to level off with 6 to 7 items. Lines with the highest penetration levels are shown on this chart. It is possible that a slightly less incremental line could be a better choice at any stage for marketing or manufacturing reasons. DETERMINATION OF LINE SIZE (Example)
ANALYTICAL APPROACH (continued) TURF is often not very discriminating when used to measure the appeal of product lines among the total sample. Rather, it appears to be far more sensitive within the previously developed preference segments identified in the segmentation analysis (as it allows the evaluation of potential product lines across all of the preference segments to help choose between alternatives). An example of this analysis by segment follows.
“Turf” Analysis –Potential Of Alternative Product Lines - By Segment - Baseline: Add 2: Increase Vs. Current: Add 1: Increase Vs. Current:
A draft questionnaire can delivered within a few days of project authorization. Once we receive questionnaire approval and the final stimuli for programming, our time line would be as follows: Questionnaire programming/developmet of design1 week Interviewing1 week Top line results/analytical work session2 – 3 weeks Final report (PowerPoint deck)1 – 2 week TIMING 20 FRC will be responsible for all phases of this research from questionnaire development through full analysis. As proposed, the costs would be $ COSTS