Editor’s note: Michael DeHart is research sciences director, Socratic Technologies, Inc., Chicago.

NCURA (non-cannibalized unduplicated reach analysis) is a new technique that provides TURF-like results but which better facilitates the modeling of optimal product mix strategies. This is accomplished by evaluating consumers’ purchase decisions for possible cannibalistic substitutability among options. NCURA provides modeling solutions to help decide what combination of products to “shelf” together and the impact of trading out one variety for another. Conversely, it also indicates which products are highly cannibalistic of each other and therefore not necessary to stock in order to obtain maximum unduplicated reach.

TURF (total unduplicated reach and frequency) has become a widely used market research tool over the past several decades. Not only has TURF gained in popularity, but extensions and variations on the theme have been made, dependent upon the specific research agenda being addressed.

While TURF is successful at what it proclaims to do, algorithms have been created in order to improve the standard TURF model by adding factors intended to mimic real-world market activity, while simultaneously providing more information than originally possible with TURF. NCURA provides a similar basic output as does TURF, but several other critical factors are taken into account when producing this output. Additionally, NCURA provides other critical results that TURF cannot provide, thereby offering an important marketing advantage.

To illustrate this technique, we will be analyzing 20 different cookie varieties and the market reach associated with different combinations. Our database contains data on all 20 cookie varieties, as evaluated by 500 respondents. All cookie varieties were rated on a 5-point scale with regard to the following question:

Additionally, the cookie manufacturer is looking to establish two different marketing schemes: one for large supermarkets, where they have a much larger shelf space available, and a second for smaller corner-stores, where they may only be able to stock only three to five types of cookies. An additional concern the manufacturer wishes to address is, what happens if the favored cookie flavor is temporarily out of stock? What should it be replaced with so as to lose as few sales as possible?

Comparing NCURA and TURF

While NCURA provides output similar to TURF, it seeks to incorporate a more real-world scenario and further provides directly valuable analysis that is impossible to accomplish with the traditional TURF model alone.

One concern many researchers have had with the standard TURF analysis is that it fails to accurately identify what should happen if the top-rated product is not available without completely rerunning the model. Perhaps a shipment is late, the manufacturer is out of stock, or a senior v.p. shoots down a top-rated product for some reason. NCURA provides a “possible substitution” alternative showing, in effect, what is the best marketing recommendation if one wishes to substitute one cookie variety for another.

Additionally, NCURA not only tells you what products to include in your sales mix, but also provides important insight into what units not to include. Recommendations on excluded units are not based simply on low consumer scores but rather on common statistical techniques that provide the client with a comprehensive list of what varieties are so similar in terms of market reach that there is no significant increase in terms of market share if they are included on the shelf. This is especially useful for the situations where shelf space or production capacity is limited.

An NCURA example

First, we will examine the simple top box purchase intent score proportions of all cookie varieties (i.e., the percentage of the sample which stated that they would “definitely” purchase the variety of cookie). The Top 10 cookie varieties (in terms of top box scores) are shown in Figure 1.

If we were to stop our analysis here and draw our conclusions, we would want to stock chocolate chip, double fudge, mint chip and sugar cookies, and if space provided, the other six highest top box frequencies as well. However, this is not taking into account any possible demand duplication. For example, a significant proportion of people who rank chocolate chip as a 4 or 5 also rank double fudge as a 4 or 5, which may indicate that the two are very close substitutes for one another. This means that if chocolate chip were not available, it might be that customers would be just as happy with double fudge. Therefore you wouldn’t need to stock both flavors in order to make a sale.

If we are trying to gain optimal market share with the fewest number of varieties, we need to isolate cases and remove close substitutes which do not add marginal sales. The ultimate goal is to provide a cookie display that garners the maximum share of the market, not just to display the highest-rated cookie varieties.

Clearly, TURF would provide this information. However, TURF would not provide possible substitution recommendations, nor would it tell us what cookie varieties are relatively unimportant to stock (assuming limited shelf space once again) even though they score highly with regard to top box frequency.

After running the NCURA analysis, we find the optimal combination of cookie varieties to display in order to obtain the largest proportion of market share, as shown in the left side of Figure 2. What we have found is to be expected: as we add more and more cookie varieties, it becomes a game of diminishing returns with regard to proportion of market share captured. On the surface, the results in Figure 1 look very much like standard TURF results. However, comparing our results with standard TURF output reveals the differences (Figure 2).

What we see is that the top three cookie varieties are the same using both TURF and NCURA (note that this area is denoted by the yellow shaded region). The first difference we see is that TURF provides a reach of 55.4 percent, while NCURA provides a reach calculation of 57.8 percent. This is due to the fact that NCURA takes into account more complex substitutability interactions.

This tends to better mimic real-world market activity. How many times have we all gone to the corner store to pick up a certain variety of cookie (or flavor of ice cream) only to find that our first choice isn’t available? Did we leave empty-handed or was there another variety that we chose instead? For many categories, there will be something in the mix that will still satisfy our craving.

Note also that looking past the top three cookie varieties, we begin to see completely different patterns emerge. NCURA advocates for replacing vanilla almond cookies (from the standard TURF results) with praline crisps. This is due to the aforementioned deeper modeling of inter-substitutability.

As mentioned previously, NCURA also provides insight into intelligent, actionable alternatives if, for some reason, chocolate chip cookies are unavailable. The NCURA model indicates that if we wish to find an acceptable substitution for chocolate chip, we should consider double fudge cookies. In Figure 3 we see a comparison of the optimal NCURA results and results obtained from substituting double fudge for chocolate chip.

Statistical note: The reader will notice that in the optimized NCURA results, we find that peanut butter fudge swirl is a strong correlate to double fudge. When we substitute double fudge for the chocolate chip, however, some of the flavor correlates change. This is because when we identify a highly desirable flavor (that would likely result in a sale), all those people with a satisfied demand are removed from the subsequent steps of modeling. When flavors are substituted for those in the optimal mix, different people are removed, so the remaining flavor substitutability structure is changed.

Note also the columns labeled “highly correlated flavors.” These are flavors that are so similar in respondent-level scores to the main cookie flavor that we advocate NOT including them on the shelf (unless there are no restrictions on the amount of shelf space available, in which case all varieties should be stocked). Further, respondents score “highly correlated flavors” so similarly to the parent flavor that, for example, even though 55.4 percent of respondents indicate they definitely would purchase chocolate chip, store owners are likely to see that 57.8 percent of respondents will choose chocolate chip if mint chip and espresso chip are not available. In this case, we recommend not stocking mint chip and espresso chip cookie varieties due to the high correlation of respondent attitudes amongst these cookie varieties.

NCURA simulator: additional insights and the online environment

NCURA has been used to produce an online simulator which allows researchers to directly probe NCURA results for any possible combination of variables entered (cookie varieties, advertising messages, sports car features, etc.). This is a critical tool for meeting the demand-based needs of smaller stores, or any other situations where shelf space is limited. The client can choose to see the top three, four, or n number of variables displayed along with the associated output scores. Additionally, the user can input the variables of their choice (say apple walnut, candied pecan, and oatmeal raisin) and obtain the overall unduplicated market reach, possible substitutions, and highly correlated variables for the combination of their choosing. This allows for virtually unlimited analysis (limited only by the number of variables, or cookie varieties, in the database). Additionally, the simulation tool allows respondents to look at any possible subset of the data (such as how men and women compare, different age groups and their associated preferences, regionally based preferences, etc.). One particularly nice feature of the online simulation tool is that it allows users to upload databases in real-time, thereby allowing for the monitoring of time series-based trending of changing market behavior over the course of any desired time period. Clearly, consumer (or business-to-business) markets are not static environments, and the ability to track changes over the time allows for up-to-date, intelligent marketing changes.

NCURA is a tool not only for cookie varieties or ice cream flavors, but also for messaging strategies. For example, if an automobile manufacturer wants to market a new car through direct mail and television commercials, two very different marketing strategies emerge. When using direct mail, space is less limited, and potential consumers may view the messages as many times as desired. This is equivalent to stocking cookie varieties in a major supermarket where shelf space is abundant. When the automobile manufacturer wants to consider marketing strategies for a television commercial, other concerns emerge. When using a television commercial, if too many messages are conveyed consumers tend to forget those messages shown towards the beginning of the commercial in favor of messages conveyed later in the commercial. NCURA provides a tool whereby a savvy automobile manufacturer may consider different combinations of three to five messages, thereby creating as much interest as possible given the constraints of television commercials.  | Q