21
Creating fragrance concepts from first principles: Identifying what drives ‘fit’ of concept elements to end-uses Howard R. Moskowitz Barbara Itty Rachel Katz Moskowitz Jacobs Inc., White Plains, New York, 10604 USA Phone 001-914-421-7400 Fax: 001-914-428-8364 E-mail: [email protected] Pieter Aarts Sensor Marketing & Research - Europe (SMR) Somme-Leuze, Belgium Phone: +32-86-322242 Fax: +32-86-323830 e-mail: [email protected] Abstract This paper deals with the fit of sets of fragrance elements (36 elements each) to 30 different end uses, varying from purely ‘fine’ to purely ‘functional’, by means of an integrated approach called a ‘mega- study’. The method used was conjoint analysis, executed on the Internet, with 150-160 respondents in each of the 30 related studies. The data suggest that respondents can identify which particularly elements in a fragrance oriented concept fit a specific end use, and that the fit to the end use differs by the specific end use. Three segments emerge from the study, based upon the pattern of the responses, with one of the three segments exceptionally responsive to the fragrance descriptions. A second recurring segment emerged that disliked fragrance messaging. A third segment emerged that liked fragrances, but liked other aspects of the concept far more. The paper shows the nature of these segments, and the business opportunity to be gained by creating fragrance concepts that appeal to the segments, rather than to the general population. Introduction - Fragrance as a key factor in many products It is clear from the products sold in stores that fragrance is a key factor in marketing. One need only go into the more prestigious department stores, and often the first people you encounter are the fragrance salespeople offering a complimentary sniff and spray of a new fragrance, or a new cosmetic line including a fragrance. Furthermore into the store are the health and beauty aids products, many of which come in different fragrances. The fragrances of the shampoos, gels, and other health and beauty aids constitute an integral part of the product, and are often used as differentiators to signal other functionalities in the product (e.g., different fragrances used for different types of hair). Finally, for the household environment (e.g., deodorizers), fragrance is key. By the judicious use of fragrance one can make the local environment both more pleasing and more individualized, albeit temporarily. Household fragrances come in a variety of different smells, and the marketing emphasis is often on the choice available to the buyer. The importance of concepts as guides to development and marketing Given the important of fragrance both for personal and household use, how can the researcher identify what to present in a strong fragrance concept? In the academic literature the work with concepts deals with the nature of how we process information. Advertising research recognizes the difference between information and emotion in concepts (Golden & Johnson, 1982; Lautman & Percy, 1983). At the practical, business level, however, there is not a published, integrated corpus of knowledge about fragrance information in concepts as one might want. Companies do a great deal of fragrance research, and have reams of data about what fragrance concepts win and what lose. Fragrance suppliers, as well, have such volumes of data. Little has appeared in the literature, and there does not seem to have been a systematic, publicly presented exploration of concepts in the way a fragrance-oriented business might find useful.

Creating fragrance concepts from first principles ...howardmoskowitz.com/wp-content/uploads/2014/04/zppd_esomar_frag... · Creating fragrance concepts from first principles: Identifying

Embed Size (px)

Citation preview

Creating fragrance concepts from first principles: Identifying what drives ‘fit’ of concept elements to

end-uses

Howard R. Moskowitz

Barbara Itty

Rachel Katz

Moskowitz Jacobs Inc.,

White Plains, New York, 10604 USA

Phone 001-914-421-7400

Fax: 001-914-428-8364

E-mail: [email protected]

Pieter Aarts

Sensor Marketing & Research - Europe (SMR)

Somme-Leuze, Belgium Phone: +32-86-322242 Fax: +32-86-323830 e-mail: [email protected]

Abstract

This paper deals with the fit of sets of fragrance elements (36 elements each) to 30 different end

uses, varying from purely ‘fine’ to purely ‘functional’, by means of an integrated approach called a ‘mega-

study’. The method used was conjoint analysis, executed on the Internet, with 150-160 respondents in each

of the 30 related studies. The data suggest that respondents can identify which particularly elements in a

fragrance oriented concept fit a specific end use, and that the fit to the end use differs by the specific end

use. Three segments emerge from the study, based upon the pattern of the responses, with one of the three

segments exceptionally responsive to the fragrance descriptions. A second recurring segment emerged that

disliked fragrance messaging. A third segment emerged that liked fragrances, but liked other aspects of the

concept far more. The paper shows the nature of these segments, and the business opportunity to be gained

by creating fragrance concepts that appeal to the segments, rather than to the general population.

Introduction - Fragrance as a key factor in many products It is clear from the products sold in stores that fragrance is a key factor in marketing. One need

only go into the more prestigious department stores, and often the first people you encounter are the

fragrance salespeople offering a complimentary sniff and spray of a new fragrance, or a new cosmetic line

including a fragrance. Furthermore into the store are the health and beauty aids products, many of which

come in different fragrances. The fragrances of the shampoos, gels, and other health and beauty aids

constitute an integral part of the product, and are often used as differentiators to signal other functionalities

in the product (e.g., different fragrances used for different types of hair). Finally, for the household

environment (e.g., deodorizers), fragrance is key. By the judicious use of fragrance one can make the local

environment both more pleasing and more individualized, albeit temporarily. Household fragrances come

in a variety of different smells, and the marketing emphasis is often on the choice available to the buyer.

The importance of concepts as guides to development and marketing

Given the important of fragrance both for personal and household use, how can the researcher

identify what to present in a strong fragrance concept? In the academic literature the work with concepts

deals with the nature of how we process information. Advertising research recognizes the difference

between information and emotion in concepts (Golden & Johnson, 1982; Lautman & Percy, 1983). At the

practical, business level, however, there is not a published, integrated corpus of knowledge about fragrance

information in concepts as one might want. Companies do a great deal of fragrance research, and have

reams of data about what fragrance concepts win and what lose. Fragrance suppliers, as well, have such

volumes of data. Little has appeared in the literature, and there does not seem to have been a systematic,

publicly presented exploration of concepts in the way a fragrance-oriented business might find useful.

Business issues in fragrance development and marketing

The development of concepts for fragrances is handled by many sectors in a business, including

marketing, marketing research, fragrance suppliers, advertising agencies, and creative specialists. All too

often the concept development phase is left to the so-called ‘creative’ in the agency, who scans the

environment to identify trends, and combines the insights with observations to come up with the fragrance

concept. Sometimes, concept development is pushed to the ingredient supplier as well, and incorporated

into their services as yet another aspect of their offering. Market researchers are all too often left out of the

creative process and asked only to run studies that evaluate the goodness of these concepts. Even more

distressing, the market researchers may be bypassed entirely because the task is considered “creative”

rather than “evaluative”.

Market researchers can provide powerful insights beyond simply evaluating concepts. To do so

requires a base of knowledge about what features or statements in a concept are most appropriate for

specific end uses. These end uses can vary from the emotional (for a fragrance that will be worn in a

romantic venue) to strictly functional (for a fragrance that is best for dishwashing detergent).

A database with concept elements that fit these end uses provides an invaluable resource for the

marketer because the database shows the fit of end use statements. The database becomes a lexicon, with

four key uses:

1. End use: To the degree that the elements remain the same but the end uses change, the researcher

can demonstrate how changes in end use along a continuum of ‘fine’ to ‘functional’, or ‘emotional’

to ‘pragmatic’ clearly change the elements that fit.

2. Key subgroups: To the degree that this database can be queried to understand the reaction of key

subgroups (e.g. male/female; age, income, market, self profiled images) the database will provide

knowledge about how concept elements drive appropriateness, and how appropriateness varies with

the different types of individuals that the population at large comprises.

3. Springboard to creativity: To the degree that marketers, agencies, and creatives use the database as a

springboard for their work, the database will provide a strong contribution from market research in

the area of development, rather than simply in the area of evaluation.

4. Fragrance-Concept fit: To the degree that researchers can identify fit of actual fragrances to these

language elements, there will be increased learning about the interplay of end use, language, and the

actual fragrance stimulus itself.

Key acceptance drivers in concepts -- application of conjoint analysis When developing concepts the marketer and product developer need to learn what specific

statements should be put into concepts. Many research efforts require the respondent is instructed to check

the general areas of product, service, or use that are relevant to their choice of what they buy. For HBA’s

(health and beauty aids) this might entail a choice of or rating of such general categories as fragrance,

brand, price, ease of use, etc. This type of general information is very helpful to the researcher because it

shows whether the general feature (e.g., fragrance) is relevant. This type of information is less helpful,

however, when creating a specific concept because even though the researcher can say that the issue of

fragrance is relevant, the data do not show what to say about fragrance.

Writing concepts about a product entails more profound knowledge than is often the case with

conventional research methods. Writing concepts requires that the writer have an idea of what specifically

to say. Consumers respond to specifics even more than to generalities. Any research that can help the

concept writer probe more deeply into the consumer’s mind will be helpful. Experimental design applied to

concept research helps because it allows the researcher to investigate the independent variables to discover

which particular variables (viz., statements) drive consumer interest (Box, Hunter & Hunter, 1978).

Almost forty years ago, mathematical psychologists developed a tool that would later be widely

used by marketers to probe into the consumer’s mind. This tool or method is called conjoint analysis. At

first the method required the consumer respondent to compare two sets of features, and choose the one that

they wanted. Later, the conjoint analysis was expanded to comprise vignettes or descriptions of products.

This so-called ‘profile’ technique presented the consumer with a variety of such systematically varied

vignettes or concepts, with known components. The consumer merely had to state whether he was

interested in the concept, or perhaps rate the concept. By knowing the response to the concepts, and by

knowing the composition of each concept (easy to do, because the researcher created the composition), the

researcher could use statistical methods to determine what particular elements in the concept drove

acceptance (Agarwal & Green, 1991; Chrzan & Grisaffe,1992)..

Conjoint analysis has a long and respected history in marketing (Cattin & Wittink, 1982; Green &

Srinivasan, 1978). It is used to identify the specific features of products. In much of its earlier history the

conjoint method was used for the more important, ‘high profile’ projects that would achieve corporate

visibility and that would entail large expenditures when the results of the researcher were implemented.

Paul Green and his associates at the Wharton School of Business are most noted for their major

contributions in this area, and for the use of conjoint measurement in areas that are of major importance to

companies, such as hotel development (Wind, Green, Shifflet, & Scarbrough, 1989).

Expanding the application of conjoint analysis to business issues in the fragrance industry

The input of conjoint analysis is a set of stimuli called elements or attributes/levels. The output is

the weight of these elements in driving a response. Through the use of conjoint measurement the researcher

identifies the driving impact or power of each of the elements. To the degree that the researcher can make

data acquisition easy, and/or evaluate many elements with consumers, the researcher will have taken a

high-powered knowledge-development technology and expanded the scope of its application to the more

routine issues of concept development for a specific use.

There are two enhancements of conjoint measurement of direct interest here to knowledge workers

in the fragrance industry: self-authoring conjoint measurement and Internet-based approaches. The two

enhancements allow the fragrance supplier house, the manufacturer, and the advertising agency to construct

concepts based upon a more profound understanding of consumer needs and wants. The enhancements are

structural in nature, pertaining to the mechanism by which conjoint analysis studies are constructed and

implemented.

1. Self-authoring systems. Self-authoring refers to the ability of the researcher to set up his own

conjoint based study. Where in previous years one required an expert to set up the study, the

current trend is towards systems that embed the expertise within the program, and require

relatively little technical expertise from the user. Thus today there are writing programs (Word®

for Windows), and presentation programs (Powerpoint® for Windows), which require little

technical expertise. One needs to know the instructions for these self-authoring programs, but with

a modest knowledge one can produce professional looking articles or presentations. Conjoint

analysis research need be no different. With self- authoring systems for conjoint measurement the

researcher need only know the basics of the system in order to design a study, launch it, run

respondents, and then analyze the data (Moskowitz, Gofman, Katz, Itty, Manchaiah, & Ma, 2001).

In the fast moving, cost-competitive world of fragrance, anything that allows the reduction of time

and cost is a welcome addition to knowledge building. This self-authoring technology is further

described below as the method used for the current fragrance concept study.

2. Internet-based studies for conjoint analysis. Over the past decade the growth and popularity of

Internet-based research has been observed and enjoyed by researchers and marketers alike

(Mizuno, 1997). In various trade publications there is a continuing mention of advances in Internet

based research. Previously the issues surrounding the Internet were ones of representation of the

respondents, but with increasing penetration of the Internet worldwide these issues are going

away. To the degree that the researcher can use the Internet to execute studies, the researcher can

accelerate the acquisition of knowledge and significantly reduce the cost. In the world of fragrance

development, the cost reduction is key because the available monies to do research are often small.

The speed up of knowledge acquisition is key because often fragrance suppliers work on briefs

from manufacturers, with a short time available. The Internet allows the fragrance suppliers and

fragrance researchers in general to do more concept research, without having to sacrifice the

power of quantitative research in light of cost constraints. The validity of conjoint measurement on

the Internet has been established by showing that respondent ratings ‘track’ the known variations

in the concept elements, according to rigid statistical criteria (Moskowitz, Beckley, Mascuch,

Adams, Sendros, & Keeling, 2002).

Mega studies across categories and meta analysis to reveal patterns

Internet-enabled conjoint measurement provides the researcher with the capability of conducting

larger scale studies than are ordinarily the case with more conventional methods. In a novel application of

conjoint measurement, Beckley & Moskowitz (2002) introduced the mega study, which comprises a set of

smaller, linked conjoint study. Each study deals with a specific topic, such as food preference or insurance

preferences. The smaller studies comprise the same or similar sets of concept elements (usually 36

altogether per study), combined with a single classification questionnaire common to all the studies. The

respondent is invited to participate in any of the studies he wishes, by means of an invitation that leads to a

‘wall’. The wall presents the available studies. The respondent completes the conjoint evaluation for a topic

of interest, and the subsequent classification questionnaire.

By linking the studies together with common elements, the researcher can determine the degree to

which a specific end use drives interest in a concept element. For fragrance, this means that the researcher

can present the same concept elements for different end uses (e.g., a fragrance for different seasons, etc.)

and determine how these end uses affect the impact of the same concept element. The common

classification questionnaire further links together the different studies.

For fragrance concepts there are at least two different types – fine fragrance and functional

fragrances. Functional fragrances, in turn, can be divided into fragrances used for health and beauty aids

(HBA) of a more personal nature, fragrances used to deodorize and odorize the environment, and

fragrances used to effect some type of action on a product such as a detergent. Each of these sub-categories

of fragrance can use the same set of concept elements. The results of a mega-study should then reveal

which particular concept elements work in each category and which do not. The results might further help

the marketer to create new and potentially better concepts for fragrance-oriented products. Finally, the

application of segmentation to the results of the mega study should reveal the existence of segments, and

help to ascertain whether the same segments appear in different fragrance-related product categories.

Objectives of this mega- study

This paper represents a first step in a systematic exploration of fragrance-oriented concepts for

both fine fragrances and functional fragrances (health & beauty aids, household). The objective of the paper

is to delineate the impact of fragrance and other elements in concepts, when those concepts are developed

to fit specific end uses.

Study design

We use conjoint analysis on the Internet as the basic tool for design and data acquisition. The

conjoint analysis method was set up to allow respondents to rate the fit of test concepts to end uses. The

conjoint methodology allowed respondents to choose a study about an end use in which to participate.

Unbeknownst to the respondent, the elements were the same for certain studies, and only the end use

differed. The 30 studies comprised nine for fragrance, six for HBA’s, five for household use (non-

deodorizer), and finally ten for environmental use (deodorizing and fragrancing an area).

Elements – the raw materials of the concepts

The database comprised two sets of elements; each set comprising 36 elements arrayed into four

silo or groups of elements. The elements comprised single, declarative statements. The elements were

created to be similar to the types of elements one sees in concepts – viz., elements designed to describe and

to ‘promote’ in a real-world fashion, rather than in a so-called ‘clinical’ fashion. One set of 36 elements

(Database #1) was appropriate for fine fragrances and to some extent health and beauty aids

(fine/functional). The other set of 36 elements (Database #2) was appropriate for health and beauty aids and

purely functional fragrances. There were element overlaps between the two sets because elements could fit

into both. Table 1 presents the element lists for the two sets of studies.

Table 1: Elements used for the conjoint analysis. (Mod = modified for different end uses) Set #1: Fine Fragrances & Health/Beauty Aids (Examples) Mod

E01 A tangy blend of fruit and flowers with a touch of spice…inspired by flutters of first love

E02 A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves, cedarwood and cardamom

E03 A sophisticated oriental blend that mixes spicy notes of cranberry with intense florals

E04 A delightful blend of violet, vanilla and rose

E05 A mix of lush white roses and fresh orchids…experience the magic

E06 A sexy floral oriental mix of orange flower, iris, orchid and sandalwood

E07 A silky blend of soy protein, pomegranate extract, and green tea

E08

Pure tranquility for your senses with relaxing oils and aromatic extracts of Rose, Iris, Nutmeg, Sandalwood

and Incense

E09 Energize your senses with a sparkling blend of aromatic extracts of Sweet Lime, Green Mâte, Osmanthus

E10 An ultra-fresh unisex scent that's 100% pure

E11 Instantly boost your energy when you put it on

E12 Rejuvenate...soothes and nourishes at the same time

E13 Enriched with shea butter, vitamin E, and aloe vera

E14 Leaves skin softer, smoother and silkier

E15 Contains moisturizing agents that hydrate and nurture your skin

E16 With an exclusive complex of natural ingredients and Beta-Carotene

E17 A unique scent…mysterious, modern, intimate and sensual

E18 With pro vitamin B5...leaves you feeling soft, smooth and amazing

E19 Just a whiff...will lift your spirits immediately

E20 You'll feel refreshed as it glides on your skin

E21 Feel confident knowing you smell good on any occasion

E22 Add it to your morning routine and start each day feeling completely fresh

E23 Gives you a deep sensation of total relaxation and well-being

E24 Feel the tension melt away

E25 Feel perfectly fit, from head to toe

E26 Magnetic, warm, natural and sexy

E27 Romantic, tender, memorable

E28 From a well known designer

E29 From Lancôme

E30 From Giorgio Beverley Hills

E31 From Ralph Lauren

E32 From CHANEL

E33 From Estee Lauder

E34 From Calvin Klein

E35 From Givenchy

E36 From Donna Karan

Set #2 – Household & Environment Functional Fragrances (Examples)

E01

Now you can enjoy four new fresh scents that remind you of the seasons: Spring Garden, Summer Citrus,

Sparkling Apple, and Invigorating Breeze

E02 A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves, cedarwood and cardamom

E03 A light, breezy, outdoor scent adds freshness

E04

A sparkling blend of citrus tones including lemon, grapefruit and orange enhanced with a water floral accord

and sheer musk

E05 A mix of lush white roses and fresh orchids…experience the magic

E06

The essence of mint, lavender and spices blend with citrus notes to create an intense and distinctive

fragrance…sparkling fresh and clean smell

E07

The aroma of freshly picked wild raspberries puts an exhilarating effect in the air... tantalize the senses with

this refreshing luscious aroma

E08

Pure tranquility for your senses with relaxing oils and aromatic extracts of Rose, Iris, Nutmeg, Sandalwood

and Incense

E09 Energize your senses with a sparkling blend of aromatic extracts of Sweet Lime, Green Mâte, Osmanthus

E10 Uniquely formulated to go deeper and eliminate odors caused by germs, mold and mildew

E11 Instantly releases more dirt, dust and road grime x

E12 Leaves a long lasting showroom shine x

E13 A multipurpose cleanser that gently cleans tough dirt

E14 The easiest way to make your car look new x

E15 Now you can get the look of new…without going to extremes

E16 Kills 99.9% of all germs on contact

E17 The quick and easy way to disinfect anytime, anywhere

E18 Enjoy a remarkably fresh and clean scent…every time you use it

E19 Just a whiff...will lift your spirits immediately

E20 Give your car that great smell of clean

E21 The power to clean and deodorize for everyday, thorough cleaning

E22 A gel-based formula...stops odors and leaves a clean, fresh and natural fragrance

E23 Gives you a deep sensation of total relaxation and well-being

E24 Feel the tension melt away

E25 Gentle enough for everyday use

E26 For a fresh look, feeling and smell

E27 Tender, memorable smell

E28 From Armor All X

E29 From Febreze X

E30 From Wizard X

E31 From Castrol Super Clean X

E32 From Glade X

E33 Available in a spray, foam or freshener form X

E34 From Little Tree X

E35 From your favorite brand

E36 From your store brand

Experimentally designed concepts

The concept elements were divided into four silos comprising nine elements each ( E01-E09, E10-

E18, etc.). The elements in a silo were related to each other. The division of elements into silos constitutes

a bookkeeping system to ensure that a concept would not contain two elements of the same type that

communicated different and perhaps opposite messages. The experimental design comprised a total of 60

combinations, with each element appearing 3x in the 60 combinations, and absent from the other 57

combinations. The individual concepts comprised 2-4 elements, respectively. This approach allowed the

data from an individual to generate a complete model that could be analyzed by ordinary least-squares

procedures (Systat, 1997).

Figure 1 gives an example of the concept, as the respondents would see it on the Internet. For this

particular experimental design (36 elements in 60 combinations) a total of 200 experimental designs were

created. Each experimental design comprised the same 36 elements, but in a set of 60 unique combinations.

This approach ensured that no single design would bias the results because of an unexpectedly powerful or

poor combination.

Figure 1: Example of a test concept as a respondent would see it

Identical classification questionnaire across all of the 30 studies

The classification questionnaire instructed the respondent to profile himself on the following nine

key characteristics:

1. Gender

2. Age

3. Frequency of using the specific product

4. Where the product is purchased

5. Factors affecting preference/choice of the product (choose three of 12)

6. Type of residential area (e.g., city; suburb, etc.)

7. Geographical market

8. Income

9. Self profiling on three questions designed to elicit lifestyle (traditional, excitement, the full

experience)

Field execution The respondents were recruited by e-mail invitation, invited to participate in a set of studies, led to

a ‘wall’ comprising the 30 studies, and instructed to select a study in which they wished to participate. The

respondents were ‘general consumers’, who had previously agreed to participate in these types of studies.

The respondents were not, however, members of a standard market research panel of the type one might

develop on the Internet.

The respondents completed the interview, comprising an introduction (specifying the end use),

rating each of the 60 concepts (36 elements systematically varied) and the extensive classification

questionnaire. The interview lasted approximately 15-20 minutes. The reward was a sweepstakes with a

first prize and several smaller prizes. Figure 2 shows a screen shot of the ‘wall’ to which respondents were

invited. The researcher was able to monitor completions of the study, and could make a study disappear

from the wall temporarily. This strategy ensured that the less popular studies would fill up because

respondents coming into the study and reaching the wall would see only the studies that had not yet filled

their quota. Figure 3 shows the orientation page for a study in which the respondent decided to participate.

Figure 2: Example of the ‘wall’, which allowed the respondent to select a study in which to

participate.

Figure 3: Example of an orientation page, which was the first screen a respondent encountered.

Results

The self-defined importance of fragrance in the respondent’s opinion The classification questionnaire instructed the respondent to select three of 12 factors that they

believed were important in driving selection of the product. This type of self-profiling can show the relative

frequency with which fragrance is selected as a key driver. Table 2 shows the pre-eminent position

fragrance, followed by price, brand and mood. The numbers in Table 2 are the percent of respondents who

selected each driver as being important. Fragrance, however, is certainly extremely high on the list, perhaps

because the invitation did mention fragrance. As one might expect, fragrance is not as important in product

categories dealing with cleaning, but it is important. It is important to stress that the percent values and the

use of the word ‘important’ apply to the general importance as shown by the frequency with which

‘fragrance’ is selected. One does not know from percent values the intensity of important, but only the

frequency with which people say that fragrance (or another factor) enters into their choice of a product.

Table 2: Percent of respondents choosing each of the nine main factors (e.g., fragrance, price) as

being important for the particular product. The products are ranked from high to low within each

general product category. The factors (columns) are also ranked from high to low.

Frag-

rance Price Brand Mood

Associ-

ations

Memor-

ies

Bottle

design Color

Adverti-

sing

Exterior

package

design

Average 92% 70% 42% 28% 16% 13% 7% 5% 5% 4%

Fine Fragrance

Summer 99% 56% 33% 44% 14% 22% 11% 2% 3% 2%

Spring 99% 60% 30% 38% 19% 26% 15% 7% 3% 3%

Winter 99% 51% 36% 38% 25% 22% 15% 1% 3% 2%

Job 99% 61% 31% 42% 14% 16% 6% 1% 2% 5%

Romantic 97% 55% 31% 40% 23% 21% 12% 1% 4% 2%

Party 97% 52% 40% 42% 15% 22% 10% 5% 1% 3%

Young Adult 96% 60% 35% 38% 14% 23% 16% 1% 2% 3%

Fall 95% 61% 27% 39% 20% 21% 14% 4% 2% 3%

After Sport 93% 57% 29% 41% 19% 15% 9% 3% 2% 5%

HBA

Aroma Therapy 99% 65% 14% 53% 24% 13% 6% 12% 4% 6%

Shower Gel 98% 74% 33% 35% 12% 10% 6% 13% 2% 2%

Fragranced Soap 98% 64% 31% 38% 16% 9% 4% 17% 6% 6%

Bath Therapy 96% 71% 26% 48% 16% 9% 9% 9% 7% 4%

Body Lotion 93% 71% 42% 27% 16% 9% 6% 2% 8% 5%

Shampoo 83% 78% 55% 21% 12% 3% 6% 7% 7% 6%

Environment &

Deodorizers

Air Freshener 98% 80% 38% 25% 10% 15% 6% 5% 1% 5%

Romantic Setting 98% 61% 22% 43% 18% 15% 7% 24% 3% 6%

Closest 95% 81% 38% 23% 18% 11% 2% 4% 4% 7%

Carpet 94% 83% 52% 21% 9% 9% 1% 1% 9% 2%

Office 93% 72% 41% 15% 16% 12% 8% 5% 4% 3%

Living Room 93% 81% 46% 19% 17% 5% 7% 4% 4% 1%

Car 93% 75% 42% 26% 16% 6% 3% 8% 5% 11%

Child Room 91% 81% 47% 16% 14% 11% 4% 3% 5% 8%

Hotel Room 88% 66% 34% 19% 22% 15% 5% 4% 5% 3%

Pet Area 84% 81% 56% 13% 18% 6% 1% 1% 7% 2%

Cleaners

Bathroom

Cleanser 81% 82% 67% 7% 11% 6% 6% 3% 13% 3%

Laundry 80% 87% 77% 6% 10% 8% 3% 2% 7% 2%

Kitchen Cleanser 78% 85% 70% 7% 8% 8% 4% 9% 11% 4%

Dish Detergent 77% 84% 70% 10% 11% 4% 5% 2% 7% 4%

Fabric Stain

Remover 68% 81% 75% 6% 12% 3% 7% 5% 14% 7%

The interest “utility” model

Market researchers typically deal with incidence statistics, showing the number or proportion of

respondents who, having seen a concept, rate it to be interesting. In conventional market research this is

known as the ‘top two box’ when the rating scale is the five-point purchase intent scale. The top two boxes

are ‘definitely would purchase’ and ‘probably would purchase’. In this study the respondents rated each

concept on the anchored 1-9 scale. These ratings were converted to a binary value, 0 to denote not

interested (rating 1-6), or 100 to denote interested (rating 7-9). Note that the assignment of the two values is

arbitrary, but follows the convention of the market research community.

The ratings thus comprised a binary matrix of 0’s and 1’s. For the sake of computational

simplicity and understanding, the ratings for each respondent were analyzed by ordinary least squares

regression to generate an equation of the form:

Rating = k0 + k1(Element 1) + k2(Element 2) … k36(Element 36) (1)

The additive constant, k0, a computed parameter, shows the estimated conditional probability of a

person saying “interested” in the concept (viz., rating of 7-9) if there are no elements present. Clearly this

additive constant is a computed parameter because all the rated concepts comprised elements. The

coefficients, k1…k36, show the conditional probability that a person will be interested in the concept if the

specific element (element 01 to element 36, respectively) is introduced into the concept. The model is

additive so that one can put in 2-4 concept elements, add together the additive constant and the coefficient,

and thus arrive at an estimate of interest. When the data from a group of respondents is aggregated the

result shows the estimated proportion of respondents who would be interested in the concept.

Basic interest level in fragrance-oriented concepts shown by the additive constant The additive constant, k0, can be interpreted as showing the basic level of interest in the fragrance-

oriented concept. The higher the value of additive constant the more the basic interest will be, even without

concept elements. Furthermore, when the additive constant is high then the elements do not have to do

much work in convincing the respondent that the idea is good. A high total (corresponding to high

acceptance) can be achieved by incorporating only modestly performing elements. With a high additive

constant much of the work is already done and to achieve a given concept score the elements themselves

need not do well.

A sense of the range of additive constants in different categories comes from previously reported

values. For instance, credit cards have constants around 10-25, meaning that without elements to increase

interest only 10% to 25% of the respondents are interested in the credit card without any additional

information (Rabino & Moskowitz, 1999). Essentially, this additive constant means that the idea of a credit

card without additional information is not persuasive. Concepts about computers show additive constants

around 30 (Ewald & Moskowitz, 2001).

Table 3 shows the additive constant for the 30 different products. In the classification

questionnaire respondents checked of the key drivers that they thought to be relevant for choosing a

product in the category. Table 3 shows the additive constant for both the total panel and for those

individuals who said that they were driven by fragrance, by brand, or by price, respectively. There were

nine other factors, but these represent three rather different drivers of purchase.

A key result emerging from Table 3 is the very large difference in basic interest in concepts from

different studies, even of the same type. The additive constants differ within a general category, such as

fine fragrance or HBA. The same elements can generate far higher base interest when positioned as a

young adult fragrance (constant = 45) than when positioned as a fragrance to be worn during the fall, the

winter, at a party or on the job (constants of 36,36 and 31, respectively). Similarly, a concept positioned

for a body lotion is far more interesting than the same concept positioned for a shower gel (constants of 43

vs. 33, respectively). The same differences in the basic pattern occur for the other two sets of functional

fragrances as well.

Another question underlying the analysis in Table 3 is whether predisposition to fragrance would

generate higher concept ratings, which in turn might manifest itself in a substantially higher basic interest

(additive constant). This proved not to be the case. Whether a respondent said he/she was driven by

fragrance, by brand, or by price the additive constant was quite similar. The standard deviation across the

three drivers, fragrance, brand, price, respectively, is relatively small. The only major differences across

drivers are for the aroma therapy and bath therapy products. For those two end uses the brand is very

important, perhaps because when the word ‘therapy’ is added the consumer looks for brand as a

reassurance. Table 3: Value of the additive constant for the different studies, by total panel and by respondents

self-defining themselves as driven by fragrance, brand or price, respectively. Total Fragrance Brand Price StDev

Fine Fragrance

Young Adult Fragrance 45 45 52 46 4

Spring Fragrance 40 40 40 41 1

Summer Fragrance 40 40 47 40 4

Romantic Fragrance 38 39 39 41 1

After Sport 37 39 50 39 6

Job Fragrance 36 36 33 34 1

Fall Fragrance 36 36 34 34 1

Party Fragrance 36 35 35 38 2

Winter Fragrance 31 32 34 30 2

HBA

Body Lotion 43 43 38 39 3

Aroma Therapy 42 43 63 39 13

Bath Therapy 40 40 53 40 8

Fragranced Soap 39 40 44 35 5

Shampoo 35 35 38 32 3

Shower Gel 33 32 40 29 6

Environment & Deodorizers

Hotel Room Deodorizer 51 51 54 51 2

Child Room Deodorizer 50 52 55 49 3

Air Freshener 48 49 46 46 2

Office Deodorizer 47 48 48 47 1

Closest Deodorizer 46 46 43 47 2

Pet Area Deodorizer 46 47 44 45 2

Carpet Deodorizer 46 44 46 47 1

Romantic Setting 45 45 30 45 8

Car Deodorizer 42 42 41 42 1

Living Room Deodorizer 40 40 41 37 2

Cleaners

Fabric Stain Remover 52 54 53 58 2

Laundry Fragrance 49 53 48 49 2

Bathroom Cleanser 49 52 48 46 3

Kitchen Cleanser 45 44 45 44 1

Dish Detergent 44 43 44 46 2

The role of the fragrance description elements – how strongly do they perform in concepts?

By specific design, the first nine concept elements in each study comprised statements about

fragrance (see Table 1). These elements, denoted as E01-E09, were taken from current advertising for

different products and applied to the same product and to other, similar products. Since the fragrance

elements were the same within a class (e.g., all fine fragrances had the same nine elements), one can

measure comparatively the impact of fragrance descriptions versus other concept elements.

A previous paper introduced the measure of ‘relative importance’ of a category of concept

elements (Moskowitz, Krieger & Rabino, 2002). This measured was operationally defined as follows:

Relative Importance = (Sum Of Squares For A Category) / (Total Sum Of Squares) (2)

The total sum of squares is, of course, a function of the number of elements. For this analysis the number of

elements is the same for the total study (36), and for the fragrance descriptive elements (9).Thus the

formula (1) does not have to be adjusted for different studies with varying numbers of element.

Every study comprises 36 elements, each of which has a utility value. The utility value can be

positive, showing that the element adds to the total interest, or negative, showing that the element detracts

from the total interest. Whether positive or negative, large utility values mean that the concept element

plays a key role, whereas small utility values mean that the concept element is essentially irrelevant. An

idea of relative importance can be obtained by first estimating the total sum of squares (squaring each of

the 36 elements and adding the 36 squares together). The proportion of that total sum of squares

attributable to the first nine elements, E01-E09, measures the contribution of the fragrance elements.

Table 4 shows clearly that fragrance descriptions play different roles in the various end uses. For instance,

for summer fragrance the descriptions generate 86% of the total variation in the concept element scores,

whereas for a fragrance to be worn at the job the same descriptions generate only 60% of the total

variation.. Similar differences in the impact of the same nine elements occur for each of the four general

categories of end use, suggesting that the role of fragrance description is quite different. Table 4: Relative importance of fragrance elements as defined by the percent of total variability due

to the nine fragrance elements (E01-E09) Fragrance SS Fragrance SS

Fine Fragrance Environment Deodorizers

Summer Fragrance 86% Air Freshener 66%

Spring Fragrance 86% Pet Area Deodorizer 65%

Romantic Fragrance 80% Child Room Deodorizer 64%

After Sport 77% Living Room Deodorizer 61%

Fall Fragrance 76% Hotel Room Deodorizer 55%

Young Adult Fragrance 71% Carpet Deodorizer 52%

Winter Fragrance 71% Romantic Setting 51%

Party Fragrance 70% Closest Deodorizer 47%

Job Fragrance 60% Car Deodorizer 47%

Office Deodorizer 41%

HBA Cleaners

Aroma Therapy 82% Laundry Fragrance 79%

Shampoo 80% Bathroom Cleanser 61%

Fragranced Soap 78% Fabric Stain Remover 54%

Body Lotion 78% Kitchen Cleanser 47%

Bath Therapy 73% Dish Detergent 47%

Shower Gel 47%

Winning versus losing fragrance elements

What specific fragrance elements do well and what do poorly, and in what context? This is a key

issue for the development of fragrance concepts. In previous studies with food it was clear that most of the

strong performing elements dealt with descriptions of the particular food (Beckley & Moskowitz, 2002).

Yet, not every food related description did well. Only those descriptions that painted a word picture of the

food performed strongly.

These data allow the researcher to do the same type of analysis, but with fragrance descriptors

(elements E01-E09) instead of food descriptions. The key difference is that the same descriptors appear in

all of the concepts for a particular class of products (e.g., fine fragrances, HBA, cleaners, or environmental

fragrances). Norms for these studies are as follows:

1. +15 or higher excellent

2. +10 to +15 very good, strong performer

3. +5 to +10 good performer, adds to the concept

4. zero to +5 irrelevant, but does not hurt

5. negative do not help the concept, actually detract from the concept

The 30 studies give 270 utility values for the same sets of elements. Table 5 shows the ranked order

performance of elements, with only those elements scoring either +14 or above or –14 and below

presented. Even from this list it is clear that several patterns play out in the data:

1. Large range. There is a very large range of utility values for pure fragrance descriptions, with a high of

+19 and a low of –21. This range appears across the different elements and end uses. Fragrance

description does make a difference.

2. No single element always does well in all end uses. Rose tends to do well most frequently, but even

rose cannot do well when it is combined (in concept form) with other features for a laundry product

(“Pure tranquility for your senses with relaxing oils and aromatic extracts of Rose, Iris, Nutmeg,

Sandalwood and Incense”)

3. Fragrance elements in fine fragrances do well. Fine fragrances rarely show negative utilities for

fragrance descriptions.

4. Some negativity. For cleaning products fragrance descriptions tend to be negative, and often strongly

negative.

5. Important to use clear language. Certain fragrance descriptions do poorly again and again, possibly

because they are hard to understand. For example the element “Energize your senses with a sparkling

blend of aromatic extracts of Sweet Lime, Green Mâte, Osmanthus” does poorly, because no one

understands the meaning of Green Mâte, Osmanthus; A woody, spicy fragrance with notes of

grapefruit, eucalyptus leaves, cedarwood and cardamom. Maybe no one can understand Green Mate,

Osmanthus or cardamom. They like simple phrases.

Table 5: Utility values for the highest and lowest performing elements involving fragrance

description End use Type Element Text Utility

Winning Elements

Spring Fragrance Fine A mix of lush white roses and fresh orchids…experience the magic 19

Spring Fragrance Fine A delightful blend of violet, vanilla and rose 17

Aroma Therapy HBA

Pure tranquility for your senses with relaxing oils and aromatic

extracts of Rose, Iris, Nutmeg, Sandalwood and Incense 16

Pet Area Deodorizer Environ.

Remove the toughest of stains and dirt…contains active enzymes to

help you clean after your pet 16

Pet Area Deodorizer Environ. Kills 99.9% of all germs on contact 16

Aroma Therapy HBA A mix of lush white roses and fresh orchids…experience the magic 16

Pet Area Deodorizer Environ.

Natural enzymes tackle tough stains and odors and remove them

permanently with no "cover-up" perfume odors 16

Fragranced Soap HBA A delightful blend of violet, vanilla and rose 15

Pet Area Deodorizer Environ.

Powerful cleaning...helps remove tough stains and dirt without the

damaging effects of chlorine bleach…safer for pets 15

Romantic Fragrance Fine A mix of lush white roses and fresh orchids…experience the magic 14

Party Fragrance Fine A delightful blend of violet, vanilla and rose 14

Aroma Therapy HBA A delightful blend of violet, vanilla and rose 14

Party Fragrance Fine A mix of lush white roses and fresh orchids…experience the magic 14

Romantic Fragrance Fine A delightful blend of violet, vanilla and rose 14

Summer Fragrance Fine A delightful blend of violet, vanilla and rose 14

Losing Elements

Carpet Deodorizer Environ.

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -14

Pet Area Deodorizer Environ.

A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves,

cedarwood and cardamom -14

Air Freshener Environ.

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -14

Romantic Setting Environ.

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -14

Shampoo HBA

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -14

Body Lotion HBA

A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves,

cedarwood and cardamom -15

Fabric Stain Remover Cleaner

A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves,

cedarwood and cardamom -15

Closest Deodorizer Environ.

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -15

Child Room Deodorizer Environ.

A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves,

cedarwood and cardamom -15

Romantic Fragrance Fine A silky blend of soy protein, pomegranate extract, and green tea -15

Laundry Fragrance Cleaner

A sparkling blend of citrus tones including lemon, grapefruit and

orange enhanced with a water floral accord and sheer musk -15

Dish Detergent Cleaner

A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves,

cedarwood and cardamom -16

Body Lotion HBA

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -16

Pet Area Deodorizer Environ.

Pure tranquility for your senses with relaxing oils and aromatic

extracts of Rose, Iris, Nutmeg, Sandalwood and Incense -16

Fabric Stain Remover Cleaner

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -16

Bathroom Cleanser Cleaner

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -16

Spring Fragrance Fine A silky blend of soy protein, pomegranate extract, and green tea -17

Living Room Deodorizer Environ.

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -17

Laundry Fragrance Cleaner

Pure tranquility for your senses with relaxing oils and aromatic

extracts of Rose, Iris, Nutmeg, Sandalwood and Incense -17

Laundry Fragrance Cleaner

A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves,

cedarwood and cardamom -18

Pet Area Deodorizer Environ.

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -18

Pet Area Deodorizer Environ. A mix of lush white roses and fresh orchids…experience the magic -18

Child Room Deodorizer Environ.

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -19

Bathroom Cleanser Cleaner

A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves,

cedarwood and cardamom -20

Laundry Fragrance Environ.

Energize your senses with a sparkling blend of aromatic extracts of

Sweet Lime, Green Mâte, Osmanthus -21

Tracking a single fragrance description through 30 end uses An important benefit of a mega-study is the ability to track the performance of one or more

elements through many end uses. More than anything else, this rapid analysis reveals the fact that the same

language can perform differently depending upon its end use. Although that fact should not be surprising, it

is rare that the interaction of language and end use is empirically measured, especially through a large

number of categories.

The element “A tangy blend of fruit and flowers with a touch of spice…inspired by flutters of first

love “ appeared in all of the 30 studies, and thus is a candidate for this type of meta-analysis. Table 6 shows

quite clearly that this element performs well in some studies and poorly in others. What is more interesting,

however, is the fact that the strong performance is not in one particular area (e.g., deodorizer) and weak in

another area (e.g., HBA). Rather, performance must be looked at in a case-by-case basis, even within a

single area. What works for a spring fragrance, for example, may not work for young adult fragrance.

Another instance comes from the cleaning area. The element works well for dishwashing detergent, but not

quite as well for a kitchen cleanser. Furthermore, even for laundry products the fragrances may differ. The

same element differs dramatically when used for “fabric stain remover” (utility = +6) versus for “laundry

fragrance: (utility = -1).

Table 6: Performance of the element “A tangy blend of fruit and flowers with a touch of

spice…inspired by flutters of first love” Total Fragrance Brand Price

Fine Fragrance

Romantic Fragrance 7 6 7 5

Spring Fragrance 6 7 5 5

Party Fragrance 6 7 3 7

Winter Fragrance 5 5 3 3

Fall Fragrance 4 4 1 2

Summer Fragrance 2 1 -5 -4

Job Fragrance 1 1 6 -1

Young Adult Fragrance -1 -1 -7 2

After Sport -4 -4 -8 -3

HBA

Bath Therapy 5 5 -3 5

Shower Gel 4 4 -2 5

Shampoo 4 6 2 2

Aroma Therapy 4 4 9 3

Fragranced Soap 3 4 3 1

Body Lotion -3 -3 -6 -4

Environment

Living Room Deodorizer 9 10 4 13

Bathroom Cleanser 9 9 8 8

Carpet Deodorizer 8 10 10 7

Car Deodorizer 8 8 10 9

Hotel Room Deodorizer 6 8 -2 7

Air Freshener 5 5 5 4

Office Deodorizer 4 3 11 3

Child Room Deodorizer 4 4 -1 6

Closest Deodorizer 2 2 1 2

Pet Area Deodorizer -2 0 0 -1

Cleaner

Dish Detergent 10 13 8 8

Fabric Stain Remover 6 6 5 5

Kitchen Cleanser 5 6 4 5

Romantic Setting 4 5 16 4

Laundry Fragrance -1 2 3 -2

Concept response segmentation Segmentation of the respondents refers to the division of the respondents into groups with similar

properties governing the respondents within a group. There are many methods by which to segment

consumers, and a very extensive scientific and popular literature is emerging from the research efforts

(Green & Krieger, 1991; Mitchell, 1983; Moskowitz, 1996; Moskowitz, Jacobs & Lazar, 1985; Wells,

1975).

Segmentation in fragrance preferences has already been reported (Moskowitz, 1986), based upon

responses to actual fragrances. The segmentation approach can be expanded into an analysis of responses

to concepts (Rabino & Moskowitz, 1994). One could divide the respondents by any number of different

variables, such as age, income, self stated choice of what drives interest in the particular concept, etc.. The

latter, choice of key drivers, appears not to be a particularly relevant set of variables on which to divide the

respondents, as shown by the similarity of additive constants (Table 3) and utilities for a common element

(Table 6). Other studies dealing with individual differences suggest that although one can always divide

respondents by their responses in a classification questionnaire, the performance of concept elements is

very similar across the different segments (Moskowitz & Bernstein, 2000).

A more productive way to assess individual differences and discover segments uses the pattern of

the responses. The pattern of responses is obtained in the same format as equation 1 above, with the

exception that there is no binary re-coding of the ratings prior to the regression analysis. That is, the 36

coefficients, k1..k36, are computed by ordinary least squares, but the dependent variable is the original 9-

point rating. This slight modification, no re-coding, suffices to reveal the magnitude of impact or driving

force of each element on degree of acceptance, rather than on membership in an acceptor vs. non-acceptor

class. The same approach has been done before in a variety of studies, with good results (Beckley &

Moskowitz, 2002).

Once the 36 coefficients are computed for each person in a study, the next task is to cluster the

respondents together on the basis of their 36 coefficients, using the Pearson correlation as a measure of

dissimilarity (Systat, 1997). For every pair of respondents, each with 36 elements within a study, there is a

value for the Pearson R. The R statistic varies from a high of +1 (perfect linear relation), through an

intermediate value of 0 (no linear relation), down to a lowest value of –1 (perfect negative or inverse

relation). The measure of distance corresponding to this is (1-R). The distance measure varies from a high

of 2 (1 - - 1 = 2; corresponding to an R of -1, or perfect inverse relation), down to a low of 0 (1 – 1 = 0;

corresponding to an R of +1, or perfect linear relation).

Following this approach, the 30 studies can be subjected to clustering. Although the number of

clusters to be extracted is left to the researcher, one can settle on a specified number ahead of time (e.g.,

three), and see what emerges. For this analysis each of the 30 studies was subjected to the clustering, to

extract the three segments. The results were then analyzed by looking at the average utility values of the

nine fragrance-specific elements, E01-E09. A cluster was categorized as Pro-Frag (positive to fragrance

descriptions) if the average utility value for the nine elements was 10 or higher. A cluster was categorized

as Con-Fragrance (negative to fragrance descriptions) if the average utility value for the nine elements was

–10 or lower. It is important to keep in mind that on the average, most of the utility values were only

modest in magnitude.

Table 7 shows the percent of respondents falling into each of these two major fragrance-response

categories. The most fascinating thing about these results is that the majority of respondents fall into either

Pro-Frag segment or Con-Frag segment. Fragrance language is clearly polarizing. Furthermore, the percent

of respondents who react positively, on the average, to the fragrance language varies with the particular end

use. For instance, for a shower gel product, 42% of the respondents are Pro-Frag, whereas 26% of the

respondents are Con-Frag. In the creation of winning concepts, therefore, fragrance language can be a

source of strength or a fatal weakness.

Table 7: Percent of respondents in the different studies falling into one of two segments: Pro-Frag

(strongly positive towards the fragrance descriptions) or Con-Frag (strongly negative towards the

fragrance descriptions). The percent need not add to 100 because there is a third segment not shown

here.

Pro-Frag

Con-

Frag

Pro-

Frag

Con-

Frag

Fine Fragrance Environment

Romantic Fragrance 44% 35% Romantic Setting 54% 30%

Winter Fragrance 42% 31% Pet Area Deodorizer 34% 61%

Job Fragrance 40% 60% Living Room Deodorizer 31% 69%

Summer Fragrance 34% 66% Hotel Room Deodorizer 29% 62%

Young Adult Fragrance 34% 47% Carpet Deodorizer 28% 62%

Party Fragrance 30% 0% Closest Deodorizer 26% 58%

Fall Fragrance 30% 20% Air Freshener 26% 63%

Spring Fragrance 23% 21% Office Deodorizer 24% 54%

After Sport Fragrance 23% 41% Child Room Deodorizer 22% 63%

Car Deodorizer 0 0%

Cleaner HBA

Dish Detergent 30% 63% Shower Gel 42% 26%

Kitchen Cleanser 20% 68% Bath Therapy 39% 17%

Fabric Stain Remover 19% 64% Aroma Therapy 34% 27%

Bathroom Cleanser 19% 72% Shampoo 32% 61%

Laundry Fragrance 17% 78% Body Lotion 25% 38%

Fragranced Soap 12% 51%

Creating an optimum concept for a fragrance in light of the segmentation

One of the key uses of conjoint measurement is to create new concepts that have a greater chance

of generating good products and eventually greater market success. The existence of segmented results

militates against finding one good concept by using “total panel data”, and indeed a simplistic strategy may

disguise some very important business opportunities. As an example, consider the detailed results for a

‘summer fragrance’ shown in Table 8, for total panel, and for the three segments. Segment S2 is the

Fragrance-Pro, and likes many of the elements dealing with fragrance. Segment S3 is the Fragrance-Con,

and dislikes many of these same elements. On the average most of the elements appear to be weak,

whereas they are actually quite strong performers for some of the respondents, and strong negative

performers for the other respondents. There are two consequences of this segmentation:

1. Choosing modest performers, because ‘that’s all there are’. Without knowledge about the

segmentation one might look at the data and choose one or two modest-performing elements. The

concept would be weak. This is clearly the case for the ‘total panel’ in Table 8.

2. Loss of opportunity. The marketer, not looking for segments, might miss an excellent opportunity

to create a segmented winning idea, and following that, a winning entry. This lost opportunity, the

consequence of missing knowledge, might have significant implications for market success and

bottom-line profitability. Table 8: Performance of all 36 concept elements, for a ‘summer fragrance’, by total panel and by the

three concept-response segments. Summer Fragrance Total S1 S2 S3

Base Size 150 72 51 27

Constant 40 44 27 53

E04 A delightful blend of violet, vanilla and rose 14 13 35 -26

E05 A mix of lush white roses and fresh orchids…experience the magic 11 16 31 -39

E23 Gives you a deep sensation of total relaxation and well-being 3 2 5 2

E24 Feel the tension melt away 3 4 -1 6

E29 From Lancôme 3 2 1 9

E11 Instantly boost your energy when you put it on 2 3 -1 8

E33 From Estee Lauder 2 4 2 -1

E34 From Calvin Klein 2 5 2 -3

E31 From Ralph Lauren 2 3 4 -3

E15 Contains moisturizing agents that hydrate and nurture your skin 2 2 -2 10

E21 Feel confident knowing you smell good on any occasion 2 6 -3 2

E14 Leaves skin softer, smoother and silkier 2 5 -4 5

E20 You'll feel refreshed as it glides on your skin 2 3 0 0

E28 From a well known designer 2 5 0 -3

E01

A tangy blend of fruit and flowers with a touch of spice…inspired by flutters of first

love 2 -7 23 -16

E26 Magnetic, warm, natural and sexy 2 3 3 -3

E36 From Donna Karan 1 0 4 0

E22 Add it to your morning routine and start each day feeling completely fresh 1 2 1 0

E06 A sexy floral oriental mix of orange flower, iris, orchid and sandalwood 1 -11 32 -28

E27 Romantic, tender, memorable 1 2 2 -5

E19 Just a whiff...will lift your spirits immediately 1 3 -1 -2

E30 From Giorgio Beverley Hills 1 -2 3 3

E25 Feel perfectly fit, from head to toe 1 2 -4 4

E12 Rejuvenate...soothes and nourishes at the same time 0 2 -7 10

E35 From Givenchy 0 0 1 -1

E17 A unique scent…mysterious, modern, intimate and sensual 0 -1 -3 8

E13 Enriched with shea butter, vitamin E, and aloe vera 0 -1 -2 4

E32 From CHANEL -1 -1 -1 -2

E08

Pure tranquility for your senses with relaxing oils and aromatic extracts of Rose, Iris,

Nutmeg, Sandalwood and Incense -2 -15 31 -30

E10 An ultra-fresh unisex scent that's 100% pure -3 -6 1 -1

E16 With an exclusive complex of natural ingredients and Beta-Carotene -3 -6 -3 4

E18 With pro vitamin B5...leaves you feeling soft, smooth and amazing -5 -3 -11 1

E09

Energize your senses with a sparkling blend of aromatic extracts of Sweet Lime,

Green Mâte, Osmanthus -7 -17 11 -15

E07 A silky blend of soy protein, pomegranate extract, and green tea -9 -16 1 -7

E03 A sophisticated oriental blend that mixes spicy notes of cranberry with intense florals -11 -18 13 -37

E02

A woody, spicy fragrance with notes of grapefruit, eucalyptus leaves, cedarwood and

cardamom -11 -32 17 -11

A sense of the loss of power in for the fragrance concept can be obtained from Table 9. The table

shows the optimum combination of elements for the summer fragrance concept, developed on the basis of

winning elements for ‘total panel’, and segments 1-3, respectively. Segment 2 is the Fragrance-Pro group.

The concept score (% top 3 box) for Segment 2 can be substantially increased from 67 to 82, simply by

focusing on the segment. This winning concept comes at the expense of not focusing on the other two

segments. A similar increase in concept acceptance, from 47 to 71, can be obtained for Segment 3, the

Fragrance-Con segment by concentrating only on that segment to the exclusion of the other two segments.

The elements for Segment 3 do not deal at all with fragrance, and the summer fragrance product might not

be appropriate for this group at all.

Table 9: Optimum concepts - summer fragrance - for total panel and the three segments Tot S1 S2 S3

Additive Constant 40 44 27 53

Total Panel (Basically interested) : Sum utilities = 62 64 67 46

E04 A delightful blend of violet, vanilla and rose 14 13 35 -26

E11 Instantly boost your energy when you put it on 2 3 -1 8

E23 Gives you a deep sensation of total relaxation and well-being 3 2 5 2

E29 From Lancôme 3 2 1 9

Segment 1: Sum utilities = 60 73 57 31

E04 A delightful blend of violet, vanilla and rose 14 13 35 -26

E14 Leaves skin softer, smoother and silkier 2 5 -4 5

E21 Feel confident knowing you smell good on any occasion 2 6 -3 2

E34 From Calvin Klein 2 5 2 -3

Segment 2 (Fragrance Pro): Sum utilities = 51 43 82 14

E04 A delightful blend of violet, vanilla and rose 14 13 35 -26

E09

Energize your senses with a sparkling blend of aromatic

extracts of Sweet Lime, Green Mâte, Osmanthus -7 -17 11 -15

E23 Gives you a deep sensation of total relaxation and well-being 3 2 5 2

E36 From Donna Karan 1 0 4 0

Segment 3 (Fragrance Con): Sum utilities = 37 36 21 71

E07

A silky blend of soy protein, pomegranate extract, and green

tea -9 -16 1 -7

E12 Rejuvenate...soothes and nourishes at the same time 0 2 -7 10

E24 Feel the tension melt away 3 4 -1 6

E29 From Lancôme 3 2 1 9

Discussion

The value of experimentally designed concepts in traditional image driven categories

All too often there is the quietly yet forcefully expressed feeling that image-related products such

as fragrance, cannot be investigated by the conventional research procedures. Those who make these

statements feel that fragrance concepts, like fragrances themselves, must be created by purely artistic

principles, rather than be subjected to scrutiny by market research tools. This may be the case, but the

current data suggest that one can learn a great deal about the ‘algebra of the consumer mind’ through

scientifically designed experiments.

A large mega-study provides a unique opportunity to understand the performance of concept

elements across different end-uses. Most of the time researchers concentrate on a narrow product area, with

a limited set of stimuli. By using a larger number of test stimuli (i.e., concept elements) and many end uses,

the researcher can identify emergent patterns that might have been missed. These emergent patterns can

form the basis of rules, such as the nature of elements that work across many end uses versus those that

work within a few end uses, or the performance of brands versus other elements in different end uses

When the same set of concept elements are used to assess reactions to a variety of end-uses,

principles emerge that help the researcher better understand the consumer. The types of information that

emerge from these experimentally designed concepts are the following:

1. Basic interest in the idea. Does this basic interest vary by end use in a common category, or by

user group?

2. Performance of the same concept element across different end uses. The data in this study clearly

show that end use plays a strong role in influencing the performance of the concept element. All

end uses are not the same, even if they appear to fall into the same general category. What works

for shampoo may not work for bath gel.

3. The existence and the nature of segments. Segmentation is critical, for without segmentation there

are lost opportunities for the business. The data clearly reveal the existence of segments with

rather diverse and often opposite reaction patterns. These segments are not the typical ways by

which the marketer or the product developer divide people. Rather, the segments represent

different mind-sets of consumers, who in many other ways would be classified as being very

similar.

Business applications – writing better concepts for products with fragrance

The results provide a unique way to develop research-driven insights to support concept

development. Furthermore, the approach provides researchers with a framework, method, and illustrative

data by which they can expand the function of research beyond evaluation to augmenting the creative

process. Where traditionally the creative process has been separate from research, this approach puts

research solidly in the camp of supporting the creative process with actionable information. Furthermore,

the approach is feasible, rapid, and iterative where desired, and amenable to data basing for corporate and

scientific learning (Greene & Moskowitz, 2000).

References

Agarwal, M.K., & Green, P.E. (1991). Adaptive conjoint analysis versus self explicated models: Some

empirical results. International Journal of Marketing, 8, 141-146.

Beckley, J., & Moskowitz, H.R. (2002). Dabasing the consumer mind: The Crave It!, Drink It!, Buy It! &

Healthy You! Databases. Paper 74-5, Institute Of Food Technologists, Anaheim.

Box, G.E.P., Hunter, J. & Hunter, S. (1978). Statistics for Experimenters. New York: John Wiley.

Catkin P., & Wittink, D.R. (1982). Commercial use of conjoint analysis: A survey. Journal of Marketing,

46, 44-53.

Chazzan, K. & Grisaffe, D.B. (1992). A comparison of telephone conjoint analysis with full profile

conjoint analyses and adaptive conjoint analysis. In: M. Metering (Ed.), 1992 Sawtooth Software

Conference,. 225-242. Sun Valley, ID: Sawtooth Software Inc.

Ewald, J., & Moskowitz, H.R. (2001). “Always On” – Bringing market research down to the development

engineer, closer to the customer, and into the vortex of product development. Proceedings of the 54th

ESOMAR Congress, Rome.

Golden, L.L., & Johnson, K.A. (1982). The impact of sensory preference and thinking versus feeling

appeals on advertising effectiveness. In: Advances in Consumer Research, 10, (ed. R.P. Bagozzi & A.M.

Tybout) Proceeding of the Association for Consumer Research, 203-208.

Green, P.E. and Krieger, A.. (1991). Segmenting markets with conjoint analysis. Journal Of Marketing. 55,

20-31. Green, P. E. and Srinivasan, V. (1978). A general approach to product design optimization via conjoint

analysis. Journal of Marketing, 45, 17-37.

Greene, J. & Moskowitz, H.R. (2000). Creative segmentation: A new approach for direct marketing. Paper

presented at the National Center For Direct Marketing, Conference, Orlando.

Lautman, M.R., & Percy, L. (1983). Cognitive and affective responses in attribute-based versus end-benefit

oriented advertising., In: Advances in Consumer Research, 11 (ed. T.C. Kinnear), Proceedings of the

Association for Consumer Research, 11-17.

Mitchell, A. (1983). The Nine American Lifestyles, New York, MacMillan.

Mizuno, M. (1997). Conjoint analysis of consumer preferences in cyber space. Advances in Human Factors

Ergonomics, 21 B, 475-478.

Moskowitz, H.R. (1986). Sensory segmentation of fragrance preferences. Journal Of The Society Of

Cosmetic Chemistry, 37, 233-247

Moskowitz, H.R. (1996). Segmenting consumers worldwide: An application of multiple media conjoint

methods. 49th

ESOMAR Congress, Istanbul, 535-552.

Moskowitz, H.R., Beckley, J., Mascuch, T., Adams, J. & Sendros. A. & Keeling, C. (2002). Establishing

data validity in online conjoint: Experience with internet-based conjoint. Journal Of Online Research, 1, 1-

11.

Moskowitz, H.R., & Bernstein, R. (2000). Variability in hedonics: Indications of worldwide sensory &

cognitive preference segmentation. Journal Of Sensory Studies, 15, 263-284.

Moskowitz, H.R., Gofman, A., Katz. R., Itty, B., Manchaiah, M., & Ma, Z. (2001). Rapid, inexpensive,

actionable concept generation & optimization – The use and promise of self-authoring conjoint analysis for

the foodservice industry. Foodservice Technology, 1, 149-168

.

Moskowitz, H.R., Jacobs, B.E., & Lazar, N. (1985). Product response segmentation and the analysis of

individual differences in liking. Journal Of Food Quality, 8, 168-191

Moskowitz, H.R., Krieger, B., & Rabino, S. (2002). Element category importance in conjoint analysis:

Evidence for segment differences. Journal of Targeting, Measurement and Analysis for Marketing, 10,

366-384

Rabino, S.,& Moskowitz, H.R. (1994). Sensory segmentation: An organizing principle for international

product concept generation. Journal Of Global Marketing, 8, 73-93.

Rabino, S., & Moskowitz, H.R. (1999) Segmentation, optimization and database mining. The Journal of

Database Marketing, 7, 31-45

Systat. (1997). Systat, the system for statistics. Systat division of SPSS Inc., Evanston.

Wells, W, D. (1975). Psychographics, A critical review. Journal of Marketing Research, 12, 196-213.

Wind, J., Green, P.E., Shifflet, D., & Scarbrough, M. (1989). Courtyard by Marriott: Designing a Hotel

Facility with Consumer-Based Marketing Models. Interfaces. 19, 5-47.