Editor's note: Ed Cohen is president of Survey Perspectives, Inc., Baldwin, New York.

So what is TURF analysis? No! Nothing to do with either lawn care or horse racing! TURF is an acronym for "Total Unduplicated Reach and Frequency."

TURF analysis is essentially a statistical model, which had its start in media research. Media planners, seeking to maximize "reach" while minimizing media costs, needed information that would indicate the extent to which various media vehicles have overlapping (i.e., duplicated) - and, conversely, "mutually exclusive" (i.e., unduplicated) audiences.

For example, if Magazine X delivered an audience of 1.8 million readers, how much would the actual total audience increase if Magazine Y, with an audience of 1.4 million were added to the media mix? Or what would be the incremental audience gain if Magazine Z, with an audience of 1 million were added to both Magazine X and Magazine Y?

If we were to simply add up the three audience figures - totaling 4.2 million - and hinge our estimate on that number it would be misleading.

  • Adding the three numbers would be misleading because there could be (and probably are) people who read two of the three or even all three magazines. These "duplicated" people would then be double- or triple-counted.

Further, assume the media budget only allowed for a buy in two of the three magazines. The question is which two of the three would maximize the media budget effectiveness, demographics and advertising rates aside for the moment?

Should we sum each of the various paired combinations (X+Y, X+Z and Y+Z) to determine which pair of magazines would yield maximum reach? The obvious answer, as indicated earlier, is no, because we still would not have an estimate of audience overlap and unduplicated audience.

Of course, media planners/buyers are continually faced with a far more complex array of media options than the three noted above. TURF analysis is an indispensable tool in the decision-making, media selection process. While the TURF model was developed to measure the "total unduplicated reach and frequency" given various combinations of advertising vehicles, TURF analysis has been quite successfully adapted to marketing research.

What follows is a simplified exposition of how TURF analysis, with its impressively versatile range of applications, may be used to optimize a line of products, fragrances, flavors, colors or sizes.

TURF analysis for marketing research

Key objectives
As noted, TURF analysis is employed in marketing research primarily to optimize product line candidates, fragrance, flavor, color and/or sizing offerings.

  • One objective of TURF analysis is to identify the mix that will attract the largest number of consumers with the fewest number of entries or varieties.
  • A second objective is to calculate the incremental value to the full line of adding each additional possible product or variant.

Example #1

For illustrative purposes, let us assume that a manufacturer is interested in developing two flavors of ice cream out of three candidates - Flavors A, B, and C - that have been suggested. Which of the two flavor combinations is likely to sell ice cream to the most people - AB, AC, or BC?

In fact, the number of candidates included in this kind of TURF study typically ranges between seven and 35 to optimize a group of two to 10 different product variants.

This hypothetical market research study involves 200 target market respondents screened in mall locations in five geographically dispersed markets. Each respondent is exposed to the three flavors - randomly presented to minimize order bias.

Using a standard five-point purchase intent scale, respondents indicate their level of purchase commitment to each of the three flavors.

  •  The criterion scores may be either "top box" or "top two box" purchase intent scores, depending on their relative levels.

The TURF algorithm first examines the response pattern of each respondent, logging the level of purchase interest expressed in each of the three candidate flavors. That is, it determines which flavors are chosen by individual respondents.

  • For purposes of this discussion, a flavor is "chosen" by virtue of generating a "top two" purchase box flavor interest: "definitely/probably would buy." Any single respondent may, of course, "choose" all three flavors, two of the three, a single flavor or none of the flavors.

This set of responses comprises the TURF database. The model then calculates the various choice combinations, i.e., how many respondents chose none of the flavors, how many chose only Flavor A, how many chose both Flavors A and B, and then Flavors A and C. Similarly, the process continues for Flavor B and then for Flavor C, determining the various choice combinations in continuous pairwise iterations.
For this hypothetical ice cream flavor illustration, assume the research yields the following distribution of "top two box" (i.e., "definitely/probably would buy") scores:

TABLE I
PERCENTAGE CHOOSING FLAVORS

DEFINITELY/PROBABLY WOULD BUY
   
BASE (200)
  %
   
Flavor A 50
Flavor B 45
Flavor C 25

The customary, obvious, conclusion drawn from the data in Table I would be that the client should produce Flavors A and B, which generated the two highest levels of purchase intent.

But the TURF approach would lead to a different conclusion, revealing the following:

  • Flavor A was chosen by 50 percent.
  • Flavor B was chosen by 45 percent.

    - 30 percent also chose A.
    - 15 percent chose B exclusively.
  • Flavor C was chosen by 25 percent.

    - 5 percent also chose A.
    - 15 percent also chose B.
    - 5 percent chose C exclusively.

Thus, based on interest levels expressed for each flavor and calculated in a pairwise fashion for each of the three candidate flavors, the TURF data for the 200 consumers surveyed would indicate the following:

  • 65 percent would definitely/probably buy A or B.

    - 50 percent A and 15 percent who chose B, exclusively.
  • 70 percent would definitely/probably buy A or C.

    - 50 percent A and 20 percent who chose C but not A.
  • 55 percent would definitely/probably buy B or C.

    - 45 percent B and 10 percent who chose C but not B.

The TURF-based conclusion, then, would differ from the earlier one. TURF analysis indicates that the optimal pair of flavors is A+C, rather than the A+B suggested by the previous, more traditional analysis.

TURF analysis may also be used to determine the incremental value of adding successive products, flavors, et al: i.e., going from one to two, from two to three, three to four, n to n+1, etc.

For example, in this hypothetical case, adding Flavor B to the A+C combination would "definitely/probably" bring an additional 15 percent of the consumers to the client's flavors, thereby increasing flavor "reach" from 70 percent to 85 percent. Were more flavors brought to the mix, the 85 percent would very likely increase to some extent, but that growth may not cost-justify the additional flavors.

Put somewhat differently, if a line currently consists of two (or N) products, how much additional purchase interest is a third, fourth, fifth, sixth, etc. product (n additional products) likely to elicit? Where is the point of diminishing returns? This leads to the following TURF-based example.

Example #2

A manufacturer of air fresheners was in the process of evaluating its existing array of seven fragrances, four of which were doing quite well and three others not so well. R&D developed five new fragrances for consumer testing. Based on practical production, marketing and distribution considerations, a lower limit of six fragrances and an upper limit of eight fragrances had been set for the line.

Further, since four of the seven fragrances were effective market performers, there was no intent to eliminate them from the product line. They did, however, need to be included in the study for a comprehensive TURF analysis.

The questions that needed answering were:

  • What is the appeal of each new R&D fragrance relative to the four good market performers and the three weaker ones?
  • Of the 12 fragrances, and building incrementally on the four successful fragrances, which two to four of the remaining eight fragrances would be most likely to yield the optimal line configuration?

TABLE II
PERCENTAGE CHOOSING FRAGRANCES

DEFINITELY/PROBABLY WOULD BUY
 
BASE (500)
   %
EXISTING FRAGRANCES
   
Better Market Performers  
A 42
B 40
C 39
D 36
 
Weaker Market Performers  
E 21
F 18
G 11
 
R&D CANDIDATES  
V 38
W 35
X 17
Y 15
Z 10
Table II summarizes the purchase interest scores for the various fragrances. It is clear that, as a group, the current "better marketing performers" elicited relatively high choice levels. TURF results, shown in Table III, determined that these four fragrances were chosen by an aggregate of 73 percent, clearly justifying and supporting their retention as a fragrance grouping.

While two R&D candidates (Fragrances V and W) each individually outpaced their counterparts and the "weaker performers" - being chosen by 38 percent and 35 percent, respectively (see Table II) - TURF analysis disclosed that new candidate V offered the greatest incremental potential (+8 percent). New candidate V increases the potential "reach" by five flavors to 81 percent from the 73 percent noted for the existing fragrances A-D.

New candidate W, while eliciting high choice scores on a par with new candidate V (35 percent, 38 percent respectively), was found not to significantly enhance the breadth of the fragrance line appeal. In fact, new candidate X, although chosen by only 17 percent, drove the net gain up an additional seven percentage points to a cumulative 88 percent. This type of finding, although infrequent, clearly suggests a strong niche appeal for this air freshener fragrance.

TABLE III
MAJOR TURF CHOICES
(Partial Data)

EXISTING FRAGRANCES Incremental
  Potential
Cumulative
   Choices
     
Better market performers (A-D) 73% 73%
     
Plus (Cumulative Appeal)    
New Candidate V 8% 81%
     
New Candidate X 7% 88%
     
Weaker Existing
Fragrance E
2% 90%
     
New Candidate W 1% 91%

TURF indicates that retaining weaker existing fragrance E and/or introducing new candidate W would not materially enhance interest in the overall line. Their inclusion, based on their small potential contribution, could not be justified on a cost/return basis.

The conclusion to be drawn from this TURF analysis is that with a line of six products (the four existing better market performers, A-D, and two of the five new candidates, V and X) the brand includes a line of fragrances with the potential of substantial interest to 88 percent of the air freshener consumers.

It should be pointed out that these data are based on total category users. Depending on a brand's position in the marketplace and specific competitive considerations, additional supplementary TURF analyses may also be warranted to examine line configurations among select user subgroups, e.g., heavier vs. lighter category users and/or different brand user subgroups, in addition to total sample analysis.

Conclusion

TURF analysis is a versatile technique that should be considered:

  • To plan optimization of product lines.
  • To provide guidance for possible line extensions.

It offers reliable, cost effective guidance to research and marketing decisions. TURF analysis studies may be conducted as stand-alone projects or integrated into more comprehensive concept/product studies.