// data use Assessing a brand funnel using comparative discriminant analysis | By Michael Lieberman

snapshot raditional brand funnels assess found to be particularly effective is to lay Tconsumers’ brand awareness, consideration, out a road map to guide customers down the Michael purchase, repeat purchases or favorite brand brand funnel. It is not predicated on one plan and loyalty. Their name comes from their na- to move them in one fell swoop but rather in- Lieberman uses ture – as the total number of consumers enter dividual guides to move them down one step a steakhouse a brand funnel, the size of the pool of consum- at a time. The method I use is called compara- ers narrows as they progress through it. tive discriminant analysis and this article case study to The goal of most brand funnels is to will give a simple, step-by-step template using illustrate how to describe the differentELECTRONIC stages of a customer’s a fictional casual-dining , Texas relationship with a brand. Here is a fairlyONLY Grill Steakhouse, as a case study. use results from a typical brand funnel: quarterly tracker Awareness:FOR Potential customers have heard Case study: Texas Grill about or come in contact with a brand. Texas Grill is a national chain of -featured to improve brand Consideration:OUTPUT A potential customer devel- entrees that can found in many malls and relationships. ops a deeper relationship with a brand and shopping centers across the . Tex- seeks more information about it. as Grill competes for dining dollars with other Purchase: The prospect now becomes a cus- major casual-dining chains such as Applebee’s, tomer trialing a product of the brand. , Chili’s, Red , , Repeat purchase or favorite: The customer LongHorn Steakhouse and TGI Fridays. likes the brand or product and decides to pur- Figure 1 is an example of a Texas Grill cus- chase the brand again. tomer pyramid. Customers are placed into each Loyalty: The customer only purchases that category by a segmentation back-classifi cation specifi c brand. If the brand is unavailable at a screener. The study is conducted quarterly and, store they will seek it out elsewhere. thus, each level of the pyramid is populated. Within the framework of the brand The quarterly report produces a research funnel model initial research is to segment deck that follows a planned measure of met- customers into a funnel, then research the rics and analysis of key drivers, Texas Grill best methods to move customers down the brand equity, social media analysis, pricing brand funnel. I have worked on these sorts of and measurement of menu item purchase. projects for 20 years in industries as variable This quarter, though, the management at as CPG, pharmaceuticals, electronics and the Texas Grill corporate headquarters would like food-service industry and one technique I’ve to know how to move Texas Grill customers down their brand funnel. That is, from aware to consider, from consider to purchase, etc. www.quirks.com/articles/2020/20200704.aspxww Most importantly, from repeat to promoter. Discriminant analysis is an a priori © 2020 Quirk’s Marketing Research Review (www.quirks.com). Reprinted with permission from the July/August 2020 issue. This document is for Web posting and electronic distribution only. Any editing or alteration is a violation of copyright. data use //

four brand arrays. For Texas Grill, we are going to look at the following restaurant survey sections to determine how to move down the brand funnel: primary reasons for visiting Texas Grill; favorite Texas Grill menu items; rating the Texas Grill dining experi- ence; negative statement about Texas Grill – brand barriers. On the Texas Grill quarterly track- er, Texas Grill asks respondents 21 primary reasons they visit Texas Grill and other casual-dining . Let’s say that Texas Grill has 30 main items on its menu. Respondents are then asked to rate Texas Grill on 15 dining experience questions. Finally, respondents are asked if any of a list of 14 negative statements apply to their last Texas Grill visit. statistical technique – that is, the are Texas Grill consider (700) are Table 2 shows how, for each groups or dependent variables are included in the analysis. Texas Grill item, performance rating and bar- predefined. The first requirement, diners = 1 (500), non-Texas Grill diners rier we are creating a new variable. then, is the groups. You need to have = 0 (200). And so on down the pyra- These variables are coded 1/0. If a the groups defined. Multiple discrimi- mid. Given that we have fi ve levels respondent ordered an item, that nant analysis, from which discrimi- of pyramid, we are looking at four item variable gets a 1, if not, a 0. If a nant maps are drawn, is a case where separate discriminant analyses. respondent agrees with a statement you have membership from more WhatELECTRONIC gives the comparative dis- about Texas Grill, the variable gets a than one group to define. criminant analysis suchONLY power is its 1, otherwise a 0. If a barrier, a 0, oth- Characteristics of the grouping ability to use many parts of the survey erwise a 1. The variables created are variable are simple. They are FORdistinct, to determine what motivates custom- summarized in the table. mutually exclusive and exhaustive. ers to move down the brand funnel. There are four categories shown in In the case of Texas Grill, we are look- And for each level, we are looking at, the table. Thus we run four discrimi- ing within each level of the pyramidOUTPUT essentially, a new analysis. This is nant analyses for each level of the to define membership. a roadmap that gives not one set of brand funnel – that is, we are run- The target, or dependent variables, directions but, in our case, four. ning 16 analyses in total. are constructed like this. For the fi rst It is not unusual to use six or This technique is labeled a discriminant level, we begin with the seven diff erent sections of the survey comparative discriminant analysis aware category. All respondents in the in a roadmap analysis. However, to because we run the same analysis for survey have to be aware of Texas Grill, illustrate things we will simplify our all levels of the brand funnel. The one so the sample size in the fi rst category analysis for Texas Grill and keep it to aspect that changes is the dependent is 1,000. However, as we move down the funnel, the sample size of the Table 1 analysis shrinks to those customers Base Size Aware Non-Diner Texas Grill who are in the next level. Consider Table 1 illustrates the depen- Level 1 1,000 300 700 dent variables and their respective Base Size Texas Grill Texas Grill Diner sample sizes. Consider As we can see, the base sizes shrink Level 2 700 200 500 as we descend the pyramid. In the fi rst level, those who are aware, but Base Size Texas Grill Diner TG Regular not “Texas Grill consider” are coded a Level 3 500 100 400 0. Those who are “Texas Grill consider” Base Size TG Regular Promoter are coded a 1. In the next level, only those who Level 4 400 250 150 To purchase paper reprints of this article, please contact Chris Nechanicky at 651-379-6200 x214 or at [email protected]. Table 2 variable as you work down the funnel. In this way, we are able to say which Number of Items/ Number of Attributes Variables items, which performance and which barriers drive the respondent from Primary Reasons for Visiting Texas Grill 21 21 one level to the next. We are, in eff ect, Favorite Texas Grill Menu Items 30 30 creating a roadmap. Rating the Texas Grill Dining Experience 15 15 To display the output for each level Negative Statement about Texas Grill – Brand Barriers 14 14 of the analysis would take up too much room. Table 3 is an example of Table 3 how the discriminant analysis statis- tical output might look. Outstanding Texas Grill Restaurant Ratings The simple interpretation of this TG Promoter=1, TG Regular=0 Discriminant Coeffi cients table is that the top four attributes The food is served hot and fresh 0.47 are discriminators between custom- The /steak was tasty and fl avorful 0.31 ers who are promoters of Texas Grill I was served promptly 0.31 and those who are merely TG regulars. While the output of this chart may The food is a good value for the dollar 0.28 seem self-evident, it shows that an The service is excellent 0.17 emphasis on prompt service and value The quality of food is excellent 0.16 can turn a regular diner into a word- The menu has a good variety of items 0.14 of-mouth promoter of the restaurant. The final step in this analysis Employees are friendly and courteous 0.13 is to summarize the discriminators My food order was correct and complete 0.12 across the four survey areas we used I was warmly greeted at the door by the hostess 0.12 in the comparative discriminant I enjoyed the complimentary corn bread 0.11 analyses. We are showing, in Table 4, the discriminators between TG I enjoyed the supervised garage and playground for kids 0.09 promoters and TG repeaters. Availability of sauces, utensils, napkins, etc., was good 0.07 From this summarized table we The side dishes complemented the entrée 0.03 know dining expectations, menu The manager personally thanked me 0.03 items, restaurant attributes and pos- sible barriers to guide diners to the ELECTRONICnext level of Texas Grill brand funnel. Table 4 ONLY Remember, with this analysis we Primary reasons for visiting Texas Grill produce four of these summary tables, A casual and relaxed dining-out experienceFOR one for each path of the brand funnel. A satisfying meal and side dishes Remain relevant A big meat-loving restaurant OUTPUT Comparative discriminant analysis is one of several useful, agile methods Favorite Texas Grill Menu Items that have been around since the Slow-Smoked Texas Brisket invention of mathematics yet remain relevant options for today’s branding Texas Grill Texas-Sized Sirloin strategy needs. Marketers have access TG Spicy Ribs to more and more data but in order to mine actionable data insights, like Outanding Texas Grill Attributes how to move diners through a brand funnel, they can rely on researchers The food is served hot and fresh and data strategists who are skilled The barbecue/steak was tasty and fl avorful in applying the right tools to guide I was served promptly their clients. The food is a good value for the dollar Michael Lieberman is founder and president of Multivariate Solutions, a Possible Barriers to Funnel Promotion statistical and marketing research consulting company based in New They don’t have enough healthy menu options York. He can be reached at michael@ They don’t have any appealing promotions mvsolution.com. It is not family-friendly

© 2020 Quirk’s Marketing Research Review (www.quirks.com). Reprinted with permission from the July/August 2020 issue. This document is for Web posting and electronic distribution only. Any editing or alteration is a violation of copyright.