James H. Myers

Benefit Structure Analysis:A New Tool for Product Planning

T HE 196O's saw the introduction of severaltechnologies/models/processes for providing

structure for consumer markets. Most of these in-volved the application of multivariate statisticaltechniques to problems of product positioning,market segmentation, the new product develop-ment process, and determining the order andstructure underlying sets of marketing variables.A recent article by Green reviews and critiquesmany of these efforts.'

This article presents another approach to struc-turing consumer markets: Benefit Structure Anal-ysis. It differs from other methods in severalimportant respects:

1. Il was developed especially for finding newproduct opportunities in very broad pixjd-uct/service categories, such as entirely newtypes of food or beverages, new forms ofbanking services {e.g., the bank credit card),or new multi-purpose tools for home repairor redecoration.

2. It determines consumer reactions to a largenumber (75-100) of relatively specific benefitsdesired from a type of product/service and tomany (50-75) features or physical characteris-tics of the product/service.

3. These reactions are in terms of both desirefor and perceived deficiencies in each benefitand characteristic.

• ABOUT THE AUTHORJames H. Myers is DeBell professor of businessadministration in the School of Business Adminis-tration of UCLA

4. The technique also provides relatively com-plete information as to ambient conditionssurrounding the use of the product (time ofday, other persons present, use or task, etc.).

A benefit structure study also provides a com-plete cross-sectional view of current usage pat-tems (i.e., what products/services are now usedfor each purpose or objective) within the broadproduct/service category selected. For most prod-ucts/services this information is not availableelsewhere, yet it can be obtained easily from datacollected for a Benefit Structure Analysis study.

To illustrate this technique a study of house-hold cleaning products is presented. This study ishypothetical, but the data shown are actual re-sults from a composite of three separate commer-cial studies conducted by the author in broadproduct areas involving foods, beverages, and rec-reational vehicles.

Benefit Structure Study for HouseholdCleaning Products

Household cleaning products were defined asany product used to clean surfaces within thehome other than ixigs, drapes, furniture, or to beused for dusting. Examples of products used forthese purposes are Ajax, Lysol, 409, Ammonia, GloCoat, Windex, etc.

Qualitative Phase

The study began with 25-50 in-depth interviews(for some studies focus groups would work aswell). The interviewer asks the respondent to re-call all occasions when she cleaned any interior

23

24 Journal of Marketing, October 1976

surfaces during the day prior to the interview. Foreach of these occasions she is asked:

1. What was the cleaning chore? (e.g., sink, floor,walls, bathtub, toilet bowl, shower stall,appliances, cabinet facings)

2. What product(s) were used in this operation?(type and brand)

3. What benefits were sought, or what were theobjectives of this cleaning? (see Exhibit 1 forexamples of benefits)

4. What were the physical characteristics or attri-butes of the product(s) used? (see Exhibit 1 forexamples of product characteristics)

5. What applicator (if any) was used? (e.g., mop,brush, sponge, rag)

6. What time of day was the work done; wereother family members involved, etc.?

In this study it was also necessary to ask aboutother cleaning tasks done during the past week.Since many items are cleaned only on an infre-quent basis by housekeepers (e.g., windows, ap-pliances, broiler pans, etc.), they might be missedin a discussion of cleaning chores undertaken onlythe day before. In studies of food or beverageproducts consumed frequently, it is usually notnecessary to go back beyond the previous day.

It may even be possible to omit the qualitativephase of the study entirely if a company alreadyhas extensive information about the product/service area from previous research.

The various cleaning occasions, products, bene-fits, attributes, etc. derived from the interviewswere used to design a large scale study to deter-mine the structural relationships among theseelements in quantitative terms.

Quantitative Phase—Methodoiogy

The quantitative phase of the study determinedthe degree to which each benefit and each productcharacteristic was desired by the consumer foreach usage (cleaning occasion) and the extent towhich they were or were not being received.

A nation-wide survey of 500 housekeepers wasconducted in 12 metropolitan areas. The homeinterview began by questioning the woman aboutall the cleaning occasions she encountered duringthe previous day. Going through the entire day,she told what cleaning chores she did and whatproducts were used for each occasion.

The interviewer then selected a single cleaningoccasion from all that were mentioned by the re-spondent, using a rotational sampling pattern.This one occasion became the focus of a muchmore intensive probe about product(s) used, ben-efits/characteristics desired and received, etc.

Note that this random cross-section of one clean-ing occasion should represent all U.S. householdcleaning occasions (within usual survey lim-itations) in proportion to the frequency with whichthey occur. (Care must be taken in survey timing,since pattems may differ over different days ofthe week and for the time of year—e.g.. Springcleaning.)

For this one occasion respondents were askedthe degree to which each of the 75-100 benefitswas received; then she was asked whether shewanted that benefit any more or any less than sheactually got it. For these measurements a four-point scale was used: 4 = a whole lot, 3 = prettymuch, 2 = somewhat, 1 = not at all. The verbaldescriptors for this scale were selected based on apre-test using the Thurstone Equal-Interval tech-nique. For example, a respondent might reportthat when cleaning her linoleum floor yesterday,she wanted the benefit, "cleans without scrub-bing," "a whole lot" (scale value = 4), but that shereceived it only "somewhat" (scale value = 2),resulting in a benefit deficiency of 4 minus 2, or^-

Similarly, respondents were asked how muchthey wanted and how much they got of each ofthe 50-75 product characteristics. Finally, varioustypes of supporting information were collected,such as: (1) what cleaning product(s) were used (ifany) and brand; (2) applicator used; (3) at whattime of day; (4) other persons present; (5) ac-tivities both before and after the cleaning process;(6) how often this item is normally cleaned.

The resulting data comprised a multi-dimen-sional matrix: Benefits Wanted x Benefits ReceivedX Product Characteristics Wanted x ProductCharacteristics Received x Cleaning Chores xProduct(s) Used x Brands x Applicators Used xVarious Supporting Data. Each dimension hadfrom perhaps five to as many as 100 categories.This n-dimensional matrix is called the CompleteBenefit Matrix. It comprises the data bank for sev-eral simple and cross-tabulations, as well as for awide variety of multivariate analyses. Only a fewof the more useful types of analysis can be shownbelow; space limitations preclude showing thefull spectrum of possible analytic approaches.

Primary Analytic Paradigm

All analyses were done on an aggregate basis,with responses pooled across all survey respon-dents or across some selected sub-group of re-spondents. Exhibit 2 shows schematically theprincipal types of analyses and relationships thatare normally most useful for product planningpurposes. Some of these are presented below, to

Benefit Structure Analysis 25

EXHIBIT 1

EXAMPLES OF BENEFITS

BleachesRemoves stainsRemoves greaseRemoves built-up dirtCleans tub ringLess elbow greaseCan see it workCleans cracks (grout) betterDoesn't leave residueNo rinsing necessaryDoesn't damage surfacesKills mildewDisinfectsRemoves discolorationRemoves soap scum

Chrome sparklesDoesn't dullDoesn't hurt handsDissolves greaseDoesn't remove gloss from paintBoosts detergentsStrips waxLess build-upLets color come throughStands up to damp moppingSeals porous floorsDoesn't yellowNo streakingDoes two jobs at onceLeaves it "squeaky clean"

EXAMPLES OF PRODUCT CHARACTERISTICS

Strong smellAbrasive/scratchyThin liquidLow sudsQuick dryingCan wipe onDark colorCausticContains waxContains ammoniaThick liquidLight colorContains antiseptic

illustrate procedures and suggested applications.Each analysis presented is indicated by the smallnumbers in Exhibit 2. Several different tech-niques can be utilized to establish relationships,depending upon objectives; they include cross-tabulations, graphic plots, simple correlations,factor analysis, discriminant analysis, canonicalanalysis, and simultaneous row-eolumn cluster-ing. I

Products Used By Cleaning Task

One of the most basic analyses is the Product-by-Use matrix, with product types as rows andcleaning chores as columns. (See relationship #1in Exhibit 2; space Hmitations preclude showingthe matrix itself.) This is quite similar to theStefflre Item-Use matrix;^ it differs in that re-spondents were not asked which products theywould "consider appropriate" for a particularcleaning use, but rather which product(s) they ac-tually used for the particular cleaning occasion en-countered during the previous day. Since the ma-trix for this study is summed over all 500 respon-dents to reveal a representative cross-section ofproducts and uses on a joint basis, it puiports torepresent the total U.S. market for householdcleaning products.

BiodegradableConcentratedSelf-polishingCan spray onAttractive colorContains deodorantEconomicalPine smellPerfumed smellLemon smellStainsLittle odorNo deterioration when stored

The Product-by-Use matrix shows one form ofmarket structure: what products are used for eachcleaning chore, and what cleaning chores are seenas appropriate for each product type. These rela-tionships show the market planner how productscurrently on the market are being used; theyshow what products are used as substitutes or ascomplements to one another; they suggest howsingle-use products might be altered to makethem into multi-use products; they reveal clean-ing tasks where now only a single product is con-sidered appropriate, so that competitive entriesmight be considered. Perhaps most important,they show the relative frequency of the variouscleaning tasks; if "amount used" is requested,then volumetric analysis is also possible.

Benefits Wanted vs. Benefits Received

At the heart of a Benefit Structure Analysisstudy is an examination of how much each of the75-100 benefits is wanted and received. This isdone by summing the ratings across all respon-dents and across all cleaning occasions. Exhibit 3shows a Benefit Deficiency matrix for one specificbenefit: "Removes Grease." Marginal totals showthe extent to which removing grease was both

26 Joumal of Marketing, October 1976

EXHIBIT 2MOST USEFUL LINKAGE ANALYSES

BenefitsWanted

BenefitDeficiencies

/N

\ /

ProductCharacteristics

Wanted

IProducts

BenefitsReceived

Cleaning Chores

t \J

ProductCharacteristic

Deficiencies \

ProductCharacteristics

Received

wanted and received by housewives^ on the four-point scale (see relationship #2 in Exhibit 2).

Exhibit 3 shows that 196 of the 500 respondentssaid they wanted to remove grease "a whole lot";143 said "not at all"; frequencies for benefits re-ceived are similar (147 reported being able to re-move grease "a whole lot" while 187 said "not atall"). However, simple comparisons of marginalfrequencies obscure the extent to which specificbenefits are wanted but not received.

For example, 34 respondents said they wantedto remove grease "a whole lot" but the cleaningproduct they used did so "not at all", yielding arather impressive "benefit deficiency" score of 3 (4minus I). Another 26 wanted to remove grease "awhole lot" but did so only "somewhat" (4 minus 2= 2); 23 wanted this "pretty much" but got it"not at all" (3 minus 1 - 2), for a total of 49respondents with a "benefit deficiency" score of 2.An additional 79 respondents had a "benefit de-

ficiency" score of 1, making a total of 162 of the500 respondents with some perceived deficiency inthe desired benefit, "removal of grease"—almostexactly one-third of the sample. While some of the79 deficiency scores of I should be discounted be-cause of the basic unreliability of any such surveydata, it is still apparent that there exists a rathersizeable group of respondents who were not satis-fied with the grease removal properties of thecleaning compound they used.

"Deficiency Indexes" can be calculated fromthese data in a number of ways:

1. Average Deficiency: .335This is a simple weighted average of bothpositive (wanted but not received) and nega-tive (received but not wanted) deficiencies.

2. Average Positive Deficiency: .566Since most benefit statements are worded inthe affirmative (something potentially desir-able), "positive deficiencies" would normally

Benefit Structure Analysis 27

be much more important than negative. AnAverage Positive Deficiency Index can becalculated by ignoring negative deficiencies(setting them equal to zero) and recalculat-ing the weighted average deficiency, whichmore accurately reflects the extent to whichhousewives wanted a particular benefit butdid not get it.

3. Proportion With Deficiency: .329The above weighted averages are useful inlater plots but they do not communicatevery meaningful information to manage-ment. The proportion of the sample withsome specified deficiency (1 or more, 2 ormore, etc.) is much clearer and is calculatedeasily.

4. Average Negative Deficiency: .231Setting all positive deficiencies equal to zeroand calculating a weighted average negativedeficiency yields the Average Negative De-ficiency.

5. Proportion With Negative Deficiency: .150Proportion of all respondents having a nega-tive deficiency of 1 or more (or 2 or more, ifdesired), indicating they got the benefit morethan they wanted it.

Similar deficiencies can be calculated for eachof the product characteristics; however, the situa-tion here is quite different than for benefits. Sincepixiduct characteristics are neutral (i.e., purelydescriptive), whereas most benefits are positive,negative deficiency indexes for the former shouldbe more important than the negatives in the caseof benefits, since they reflect the extent to which aparticular characteristic was received althoughnot wanted (e.g., strong aroma, grittiness). A

Product Characteristic Deficiency matrix may beprepared for each of the 50-75 characteristicsusing the same format as for benefits.

Plots of Benefits Wanted vs. Deficiencies

While the above information provides usefuldiagnostics for each benefit and product charac-teristic, it needs to be organized and summarizedfor the market planner. Plotting the AverageWanted Benefit scores (as columns) vs. AverageDeficiency scores (as rows) for each benefit/charac-teristic is a simple and vei"y useful first step.Using benefits to illustrate several possible plotswould be meaningful: Average Wanted Benefitscore vs. Average Positive Deficiency score foreach benefit; Average Wanted Benefit score vs.Proportion of Sample with Some Deficiency;Average Wanted Benefit score vs. ProportionWanting the Benefit "pretty much" or "a wholelot" that felt some deficiency.

The plots enable rapid identification of "oppor-tunity points" as a basis for focusing further diag-nostic and planning efforts. Opportunity pointsare benefits/characteristics that are wanted butnot received, or product characteristics that arereceived although not wanted. In the case ofcharacteristics, the planner can instruct the de-velopment labs to add or remove certain productfeatures (e.g., color, odor, antiseptic), assumingthis is possible.

A single plot can portray the entire cleaningproducts market. But even more useful informa-tion may be gained from plots for each of themajor product types as well as plots for each ofthe major cleaning uses. Such plots often showtwo or three products types (or u,ses) which areparticularly vulnerable because they are not ade-

TOTAL: 187

EXHIBIT 3

BENEFIT DEFICIENCY MATRIX

Benefit #49: "Removes greaseReceived:

Wanted

"A whole lot""Pretty much""Somewhat""Not at all"

"Not at all"

342322

108

"Somewhat"

26301513

"Pretty much"

27251310

"A Whole lot"

109206

12

Marginal

Wanted

1969856

143

Sums

Got

1477584

187

84 75 147Averages: Wanted: 2.70 Received: 2.36

Want minus Received - - 3No. of Respondents: 12(Please note that these totals include different combinations of "Want minus Got").

- 216

- 146

0257

179

249

334

28 Journal of Marketiiig, October 1976

BENEFIT GROUPING

EXHIBIT 4COMPARATIVE SIZE OF MAJOR BENEFIT WANTED

SEGMENTS & BENEFIT DEFICIENCIES

% of SampleWanting Beneht

Pretty Much or More

% of SampleHaving

Deficiency

A. Removes dust/dirt/film

B. Removes porcelainstains

C. Leaves no film/residue/scratches

D. Removes grease/wax/stains

E. Convenient to use/store

F. No unpleasant odorduring/after use

G. Gentle on hands/skin

V////A

V/TT\

/////I

V / //A

V / / /kquately meeting some needs that are highly de-sired or even moderately desired. Carefully de-signed new product entries could be expected tocompete effectively in these product areas becausethey are specifically targeted at benefits which peoplesay they want a great deal but are not getting now.

This type of analysis was used for two clients inthe food and beverage industry and clearly re-vealed the existence of such gaps for severalmajor product types; for example, a "pick-me-up"beverage that failed to "pick up"; a type of saucethat was considered far too fattening and thatstained clothing and table coverings.

Major Benefit and Deficiency Structure

The next analytic step is to construct a Benefitand Deficiency Structure graph for the householdcleaning products market on an aggregate basis.This analysis shows that the primary benefitswhich people want from cleaning products can begrouped into seven major categories, labeled Athrough G, as shown in Exhibit 4. For example,Benefit Group A includes benefits that are similarin terms of removal of surface dust, dirt, film;Segment D includes the removal of heavy grease,wax, or stains; Segment F involves odor andaroma.

76%

63

61

59

58

54

35

32%

28%

26%

41%

34%

62%

56%

These benefit segments are produced by group-ing together specific benefits that are perceived assimilar across all respondents and all cleaning oc-casions. What is needed is a technique that simul-taneously groups benefits and weights them inrank-order of frequency desired. If one or morebenefits that are wanted "a great deal," are, infact, "isolates" (i.e., are not perceived as similar toany other benefits) then the technique used mustidentify "isolates" as well as groupings. Presuma-bly the desired technique would be sequential: ex-tracting first the benefit or benefit cluster desiredmost, then the one desired next, and so on untilall of the most important benefits are accountedfor.

Unfortunately, no such technique appears tohave been developed. The best that can be done isto judgmentally combine the information from afactor analysis of all 75-100 benefits (from "bene-fits wanted" ratings) plus the average "benefitwanted" scores for key "marker" benefits whichmost clearly define a cluster (i.e., have the highestfactor loadings). These groupings can then be dis-played in bar graph form, in rank-order of desiredintensity. Deficiencies for each of these benefitstructure dimensions may be developed and jux-taposed on the same graph.

Benefit Structure Analysis 29

Exhibit 4 shows at a glance the "benefit anddeficiency structure" of the household cleaningproducts market. Note that it shows each benefitgrouping separately; combinations of two or moredesired benefit groupings that are wanted can bedeveloped also. The exact same analysis should bedone for product characteristics, to develop amarket structure in comparable terms. Large de-sired benefit linkages with large deficiencies canbe quickly spotted, as well as smaller desiredbenefit linkages with large deficiencies. The latteroften signal market segments that offer real prom-ise for some new product, or for the repositioningof some existing product. For example, analysis ofthese linkages showed that not many respondentswanted to remove grease/wax/stains with a prod-uct that had no unpleasant odor, but among thosewho did, a sizeable proportion perceived that theproducts they were, using were not delivering thiscombination of benefits.

Relationships of Benefits and Characteristics

Once it is determined what benefits and combi-nations of benefits are wanted, how are thesebenefits achieved? A Benefit Structure Analysisstudy shows what types of product characteristicsare seen by respondents as being most closely asso-ciated with any single benefit or any group of relatedbenefits. This provides a linkage between benefitsand those product characteristics which producethe desired benefits (see relationship #3 in Ex-hibit 2).

While this can be done in several ways, thesimplest is to correlate benefit received scoreswith product characteristics received scores,across all respondents and cleaning occasions.Since both are measured on the same four-pointinterval scale, the size of the correlation should bequite meaningful.

The sizes of the correlations are important. Sel-dom are any correlations in excess of .40 found.Most benefits show highest correlations in the .30range. This makes the job of the new productplanner more difficult, in that relationships arenot high enough to clearly indicate which productcharacteristics should be used to produce a givenbenefit. Still, comparisons of those relating high-est with those relating lowest usually indicaterather clear differences in the general types ofproduct characteristics that are and are not per-ceived to be related to a particular benefit. If noneof the relationships is high (perhaps .30 or more),it would suggest that the planner is free to useany product characteristics he chooses, since nospecific characteristic is related strongly to the

particular benefit under study, when viewedacross all products and respondents.

Reiationships of Benefits and ProductsIn a manner parallel to linking benefits with

product characteristics, benefits can be linked toproducts to show what types of products are con-sidered best at delivering each of the benefits. Oneuseful approach is to rank-order all product cate-gories in terms of their average benefit receivedscores for each benefit, with summation across allrespondents who reported on a particular prod-uct. The resultant rankings have been found to bequite useful in indicating the general types ofproducts that are most (and least) likely to pro-duce a specific benefit (see relationship #4 inExhibit 2).

Simultaneous Row/Column LinkagesUp to this point, all benefit linkages have been

done one-at-a-time, and the resultant data aremore diagnostic than structural. The same gen-eral types of information can be presented in aform that provides more over-all structure to themarket at the same time that it allows specificdiagnostic information although presented in adifferent format. This is done using a process thatgroups and displays two variables simulta-neously.

The simultaneous row/column clustering algo-rithm used in Benefit Structure Analysis was de-veloped by Dr. Edward W. Forgy^ and named"LARC" (large row/column simultaneous cluster-ing). Exhibit 5 shows the LARC analysis of clean-ing tasks (columns) vs. benefits wanted (rows).Cleaning tasks that are seen as similar in terms ofbenefits desired are grouped together, and bene-fits that are seen as similar in terms of cleaningtasks are grouped. Heavy lines have been drawnarbitrarily around blocks of the largest numbers;these blocks show the most clearly-defined link-ages between benefit groupings and cleaning taskgroupings. That is, lines show what groups ofsimilar benefits are desired by housewives for aparticular grouping oi similar cleaning tasks, andvice versa.

It is important to note that LARC or any othersimilar process will not define clusters of rows orcolumns as clearly and accurately as will the var-ious algorithms that have been designed solely forclustering, such as optimal partition clustering,hierarchical clustering, and even factor analysis.However, this weakness is compensated for by theability of a LARC-type analysis to both group anddisplay simultaneously the linkages between thetwo variables under study (see relationship #5 inExhibit 2).

30 Joumal of Marketing, October 1976

Several types of row/column analysis are useful:

1. Products by cleaning task2. Products by benefits wanted, received, and

the various benefit deficiencies3. Cleaning tasks by benefits wanted, received,

and the various benefit deficiencies4. Products by characteristics desired, received,

and their deficiencies5. Cleaning tasks by characteristics desired, re-

ceived and their deficiencies6. Benefits/characteristics wanted, received, and

their deficiencies vs. such other question-naire information as demographics, psycho-graphics, brand, usage volume, etc., as avail-able

Row/column clustering provides a highly usefulform of structure for the market planner. By look-ing at the linkages of cleaning tasks and benefits(indicated by the larger numbers that have beenenclosed within the heavy lines on Exhibit 5), hecan see at a glance which general types of cleaningtasks require which general types of benefits (orproduct characteristics). For example, the firstfour columns of Exhibit 5 show that cleaningtasks involving furniture (nos. 19, 4. 14, 20) re-quire products that remove dust, dirt, and film(benefit nos. 12, 38, 42, 14), and that leaves noresidue or scratches (benefit nos. 17, 36, 43). Themarket planner can check the array of existingproducts to see if they deliver each of the specificbenefits wanted in each grouping. And the group-

Exhibit 5Benefits Wanted by Cieaning Occasion

Cieaning TasksData ordered by row/column simiiaritles

Benefit

ID11121314151617181920212Z2324252627282930313233343536373839

to*!424344454647ta49bO51

22832t12502284737**930195139211336414410

7263334ta20179

4042294354523631151435253243816461827111

123842142260321857475940296149312346515420173643445830271950523953155533164125244535133448265628372111

Freq.

494454424406373357295250233235236372395458439632570592583668824

la 3 13 19

Ij 4 14 20

6 7 6

6 12 21

7 13 22

9 10 11 12 13 14 15 lo 17 16 19 £0 21

1 7 9 17 10 20 16 8 4 il S iS 14

2 ti 10 18 11 21 17 9 it id o 16 15

74 76b8 49 4966 41 454t 47

11 1918 129 1040 tl4 411 10

14 1412 1612 11U 1414 1719 ib17 1212 1416 la26 1729 lti25 17

33 3340 3648 4645 47

10 1717 17 19

3630 33 36

32 322b 27

4066 3b75 31

26 3a31 2<429 24 24

30 to32 3126 27

2014 1314 2016 1415 1413 2215 1211 14

24 2125 2424 2614 15 20 2Q

22 2422 2d

24 29la 2021 2219 2224 221114 1430 2527 29

4

.ii

7761to44

t<*3520a3

9942196b

33 a3 7lO -128 9t 66 1012 109 b7 U8 Iil7 16It 118 b3 i3 bb

15IU

2 7 24

23 2a24 2429 28

30 20

u75 b19 3213 15U 17It 1926 2036 1920 2417 166 b6 bil IU22 2218 1914 116 1313 1314 1325 1737 25JO 2o10

15 2129 2715 1412 1515 1917 27

11 13272713131317

i4352426

28

4blb20192731

351413212535

17It171930

15141924

lb4319253036to

i!0331720243038

2130?02936405b

Jl272430413647

272933334148

c!62923273*t<:4>)

;j42b222935414?

•i.i2ai.kit,:>!tob7

Ji3351655351Ii.

4i354660475777

&b345d6a4b

4149

Benefit Structure Analysis 31

ings themselves may suggest other similar or re-lated benefits that were not covered specifically inthe study that should be considered.

With this information the planner can:

• See how many general types of productsshould be on the market.

• Qjnsider possible additional uses for existingproducts (i.e., a product designed for one taskin a particular cluster should also be suitablefor other tasks in that cluster—or could bemodified to do so).

• Conceive new products that deliver the exactcombinations of benefits wanted for a singlecleaning task (or for a group of related tasks).

• Reposition existing products by stressing ad-ditional benefits within a benefit cluster thatare already offered by a company's own

. .|>roducts, etc. . ,. .. „_., .,•' • • ' " . * • " • ' • " • • • y''^'V''tr^.

Additional Anaiyses . ' '•'•'•-.'•'.'•'' *'.

There are many additional analyses that mightbe useful fw the market planner. For example,those respondents who most wanted a particularcleaning benefit or characteristic can be sortedout and compared to the total sample of respon-dents in terms of demographics, at what time ofday, and other questionnaire data not included inprevious analyses. One firm asked for mental andphysical "feeling tone," plus limited psychograph-ics.

Another useful analysis consists of isolatingthose respondents who most wanted a particularbenefit (or benefit cluster, obtained from factorscores) and determining the following averagesfor each product characteristic: (I) averagewanted score; (2) average received score; (3) aver-age positive deficiency score; (4) average negativedeficiency score. Comparisons among these aver-age scores show which characteristics are particu-larly crucial for a given benefit or benefit cluster.(This also can be done by correlation analysis, asreported earlier.)

The limit of possible useful analyses is basedlargely upon the imagination of the researchanalyst or market planner.

Summary

Benefit Structure Analysis (BSA) was developedout of a confluence of ideas from several earliernew product development models. Its greatestlegacy is from the Market Structure Study pro-cess proposed in the early 196O's by Stefflre.** Inparticular, his Item-Use matrix showed how anarray of existing products could be systematically

related to the various uses for which each productwas seen as being suitable. This provided "struc-ture" for a particular type of consumer product orservice market. The household cleaning productsstudy utilizes this basic idea in relating productsto uses, benefits wanted/received and deficienciesto products and to uses (see Exhibit 5), productcharacteristics desired/received and deficienciesto products and uses, plus many additional rela-tionships within the complete benefit matrix.

BSA also draws on the work of Haley, who wasnot the first to be concemed with the benefitswhich people wanted from products/services butwho best showed how markets could and shouldhe segmented on the basis of benefits, since "thebenefits which people are seeking in consuming agiven product are the basic reasons for the exis-tence of true market segmentation."'

Finally,-BSA is related to the various productpositioning models proposed by Green.*' Johnson/and others. These efforts utilize quantitative tech-niques to identify the principal types of productcharacteristics/benefits that are used by consum-ers in perceiving or ex'aluating competing productswithin a specified product type. This is similar tothe Benefit and Deficiency Structure graph shownin Exhibit 4. (In the household cleaning productsstudy products were plotted on the first threebenefit segment dimensions; this 3-dimensionalconfiguration was not shown due to space lim-itations.)

Managerial Considerations

Benefit Structure Analysis does not directlycompete with any of the above models. It wasdeveloped especially for locating new product op-portunitities in very broad product or service cat-egories, such as an entirely new type of food (e.g..Instant Breakfast) or banking service (e.g., bankcredit card) or alcoholic beverage (e.g.. MalcolmHereford's 30-proof Cows). In such cases the mar-ket planner cannot be at all sure where his bestopportunities might lie and needs a technique toanalyze many different types of loosely relatedproducts/services in a single study. BSA enableshim to spot weaknesses and opportunities in eachgeneral type of product or type of use, oppor-tunities that might well be overlooked on an apriori basis.

For example, it was pointed out earlier that oneBSA study identified a type of beverage productthe was used primarily for a "pick-up" but thatfailed to deliver this benefit sufficiently. In thisparticular case the client company: (1) decided itwould not develop and introduce a product of thissame type that provided more pick-up; (2) spent

32 Journal of Marketing, October 1976

many hours of discussion with planning, labora-tory, and management personnel considering newtypes of pick-up beverages that would be entirelydifferent from anything on the market; (3) decidedto abandon the project entirely as being incom-patible with the basic product technology thecompany had had its primary experience with. Inthe case of the sauce product that was consideredtoo fattening and staining, that particular com-pany also decided not to develop and introduce aproduct of this type since it did not have re-sources to take on the existing well-establishedbrands; however, it did utilize this information inreviewing its own sauce products and in planningentirely new types of sauces.

On the other hand, if the market planner isasked to work within a rather sp)ecific productcategory, such as white liquor beverages, softdrinks, small cars, or be«r, he is probably betteradvised to utilize one or more of the models orresearch processes proposed by Green, Johnson,and others discussed above. In such cases therange of benefits and/or product characteristicsthat have primary relevance is much narrower.Most investigators, including the present author,have found that the two or three major benefits

{i.e., groupings of similar individual benefits/characteristics) will often account for upwards of70% of the variance in overall evaluations ofproducts or frequency of use for a specific type ofproduct.

The rather simple principal line of questioningthat is at the heart of Benefit Structure Analysis(benefits/characteristics wanted/received and theirdeficiencies) can easily be included as part ofmany consumer surveys that have some otherprimary objective. This is especially appropriatewhen a firm has conducted previous research on aparticular product line and feels confident that itunderstands the benefits/characteristics that areimportant for its product. Sometimes only ten totwenty benefits or characteristics need to be cov-ered.

No single study can provide all types of marketinformation needed by the planner. The techniquepresented here provides direction as to generalareas of greatest opportunity for new product/service development plus specific diagnostic in-formation useful for either modifying or reposi-tioning existing products. Through broad applica-tion it may prove to be a useful tool for marketanalysis and planning.

ENDNOTES

1. Paul E, Green, "Marketing Applications of MDS: As-sessment and Out\ook." Joumal of Marketing, Vol. 39 No. 1(January 1975), pp. 24-31.

2. Volney Stefflre, "New Products and New Enterprises,"Market Structure Studies, Inc., 1968.

3. Dr. Edward W. Forgy is a statistical consultant in LosAngeles, CA. His program is proprietary, but a somewhatsimilar procedure is described in "Some Eliciting and Com-putational Procedures for Descriptive Semantics," V. Stef-flre, P. Reich, and M. Wendell, in Explorations in Mathemat-ical Anthropology, P. Kay, ed. (Cambridge: M.I.T. Press,1971).

4. Stefflre, same as reference 2 above.5. Russell I. Haley, "Benefit Segmentation: A Decision

Oriented Research Tool," Joumal of Marketing, Vol. 32 No. 3(July 1968), pg. 30.

6. Paul E. Green and Yuram Wind. Mulliatlrihute Deci-sions m Marketing: A Measurement Approach (Hinsdale, Il-linois: The Dryden Press, 1973).

7. Richard Johnson, "Market Segmentation: A StrategicManagement Tool." Joumat of Marketing Research. Vol. VIII(February 1971). pp. 13-18.

8. Benefit Structure Analysis was not actually used forany of these products, however.

MARKETING MEMO

What is a weed? A plant whose virtues have not yet been discovered.—Ralph Waldo Emerson, Fortune of

the Republic