The Historic Use of Computerized Tools for Marketing
and Market Research: A Brief Survey

by Jeff Lindsay, James Schuh, Walter Reade, Karin Peterson, and Christopher McKinney

Kimberly-Clark Corporation, 2100 Winchester Road, Neenah, WI 54956


Electronic tools have been used for many years to pursue traditional marketing techniques. While this is well known among marketers and most consumer products companies, there has apparently been some confusion about the historic availability of market research tools. Some tools that have been long used and are widely known among market research professionals have been "reinvented" in attempts to gain patent protection for technologies that actually are well known. To foster better appreciation of past efforts, we will cover a small but diverse sampling of topics in this paper. A comprehensive compilation of the prior art for any topic touched upon herein is outside the scope of this article.

In addition to discussing prior practices, we will also discuss a number of directions that we expect to be adopted by the industry in the future, if not already in practice at the moment.

Computerized Tools

Computers have long been an important part of marketing and of marketing research. Some early forms of computer-assisted marketing research included research interviews conducted in shopping malls or by telephone with a computerized tool. Initially called cathode ray tube (CRT) interviews, computer-assisted personal interviews (CAPI), or computer-assisted telephone interviews (CATI), the concept involved obtaining information from a user based on questions generated on a cathode ray tube screen by a computer, and entering the response into the computer by a keyboard or other means. In this manner, questions could be posed based on the input to previous questions, allowing for accurate completion of a survey and accurate entry of answers. An interviewer initially conducted such surveys, but later systems provided self-administered surveys.

Examples of such computer-assisted marketing research systems are described by M. Crask, R.J. Fox, and R.G. Stout, Marketing Research, Englewood Cliffs, New Jersey: Prentice Hall, 1994, p. 161:

"One example of a CAPI system for self-administered surveys is MAX. MAX is a microcomputer-based software package developed by POPULUS, a Connecticut research firm. The company has used MAX extensively for mall-intercept interviews.

"The J.C. Penney Company also used CAPI in combination with direct broadcast technology. Consumers view merchandise items on a television monitor and indicate on a CAPI system how likely they are to purchase each item. The information collected by Penney's aids their merchandise buyers in determining what products target consumers are most likely to buy."

Thus, prior to 1994, J.C. Penney and perhaps others used electronic displays of merchandise items coupled with a computer-assisted personal interview (CAPI) system to obtain and record a consumer response to the viewed item, allowing improved marketing decisions to be made.

An improvement over early CAPI systems is the Interactive Opinion Network™ (ION) of Market Facts, Inc. (Arlington Heights, Illinois), partially described at (as archived at ION is a multimedia-based computerized interviewing system placed in regional shopping malls throughout the United States using customized hardware and software to improve the interview process. Computer screens and speakers provide high-quality images, video, and sound to view products or services. The viewer is prompted to provide input with a touch-sensitive screen regarding the products or services portrayed. Data are acquired over a network and processed electronically to yield useful market research information.

The use of various electronic tools, including the Internet, electronic databases, e-mail, digital printing, computer-assisted interviews, and other computer-based techniques are further described in publications such as the book of Frederick Newell, Customer Relationship Management in the New Era of Internet Marketing, New York: McGraw-Hill, 2000.

The role of the computer-assisted interactive systems in marketing and market research was also assessed from a 1996 perspective by John Deighton in "The Future of Interactive Marketing," Harvard Business Review, Nov.-Dec. 1996, Vol. 74, No. 6, pp. 151-162. Deighton wrote:

"As the marketing faculty at the Harvard Business School thought about the evolving technology landscape, it seemed to us that the main thrust of the transformation in marketing practice could be reduced to this: a shift from broadcast marketing to interactive marketing. Mass-marketing concepts and practices are taking advantage of new ways to become more customized, more responsive to the individual.

"The term interactive, as we interpret it, points to two features of communication: the ability to address an individual and the ability to gather and remember the response of that individual. Those two features make possible a third: the ability to address the individual once more in a way that takes into account his or her unique response....

"Interactivity has already made major inroads into marketing budgets in the past decade in the form of direct mail, catalog retailing, telemarketing, and the incorporation of response devices into broadcast advertising. Developments in data storage and transmission, however, hold out the promise of new and better interactive tools to manage relations with customers and to link the networked corporation to its channels and its collaborators. Although the World Wide Web may be the ultimate interactive medium, there is still much that can be done with a pastiche of less exotic interactive technologies. For example, when a broadcast advertisement elicits a response such as a toll-free call -- which is then stored in a computer database and which triggers a personalized direct mailing -- that sequence represents a form of low-tech interactivity.... The Web, however, promises high-tech interactivity. When a consumer visits a Web site, many cycles of messages can be exchanged in a short time. When the consumer visits some time later, the dialogue can resume just where it left off."

Such interactive dialogs with the manufacturers or suppliers of products can be used for marketing research. The role of interactive shopping as a tool for capturing comprehensive customer-specific data has long been recognized. For example, see J. Abba et al., "Interactive Home Shopping: Consumer, Retailer, and Manufacturer Incentives to Participate in Electronic Marketplaces," Journal of Marketing, Vol. 61, July 1997, pp. 38-53 (see especially p. 46).

E-commerce has led to many innovations in marketing and market research over the Internet. Personalizing product offerings based upon known attitudes or purchasing habits of an individual is one successful strategy. Tailoring displayed ads or product promotions to search terms used by the user or by other information gleaned from surfing habits of the user has been widely used and discussed. Tools have been developed to assist consumers in the search for products, including interactive tools. For example, Häubl and Trifts discuss two-stage tools that first provide a subset of recommended product possibilities based on information provided by the consumer, followed by a comparison matrix to compare detailed features of the selected or recommended subset of alternatives to help the user make an informed choice based on an in-depth analysis of a limited number of possibilities (see G. Häubl and V. Trifts, "Consumer Decision Making in Online Shopping Environments: The Effects of Interactive Decision Aids," Marketing Science, Vol. 19, No. 1, Winter 2000, pp. 4-21). Numerous other interactive tools have been developed to assist in shopping, including the tools of the and Web sites.

The Internet has become an increasingly used tool for market research. By the mid-1990s, the Internet was widely being used to display graphic images of products or proposed products to observe consumer reaction and obtain feedback in various forms, such as Web-based forms associated with the graphic images displayed on a Web page. With the advent of "cookie" technology, thousands of individual users and their shopping habits can be tracked, allowing a marketer to understand cross-relationships between various products and how offering certain groups of products can enhance overall sales. Numerous Internet tools have played a role in marketing and marketing research, including discussion groups such as USENET news, moderated and unmoderated mail lists, Internet Relay Chat (IRC) and other chat services, e-mail, the World Wide Web, and other information access and retrieval systems such as Gopher or Archie. In 1996, Hoffman and Noval discussed the marketing implications of what they term "hypermedia computer-mediated environments (CMEs)" (D.L. Hoffman and T.P. Novak, "Marketing in Hypermedia Computer-Mediated Environments: Conceptual Foundations," Journal of Marketing, Vol. 60, No. 3, July 1996, pp. 50-68). Such CMEs can include the Web as well as other developments in electronic commerce such as EDI (electronic data interchange) systems, kiosks, electronic classified ads, and on-line services such as CompuServe and Minitel, the French videotex system (see, for example, Mindy McAdams, "The Sad Story of Videotex," 1995, available at

"Ninety-five percent of direct marketers responding report using the Internet/World Wide Web for sales or marketing applications, up from 83% one year ago. More than half (52%) make use of online services, up from 43% in 1998; and 51% use EDI, up from 45% in 1998. In addition, 46% use CD-ROM's, up from 40% in 1998; and 39% are using e-commerce technology in marketing their products and services."

Prior to the widespread convenience of the Internet, one marketing strategy in the late 1980s was to send free floppy disks to selected consumers containing a program or graphics files for displaying an electronic representation of a product, or for displaying a movie or animated graphics showing the product. Consumers could then respond to the product by telephone, mail, or e-mail, either to provide feedback or to place an order.

Several concepts involving electronic marketing customized according to a consumer profile are disclosed in US Pat. No. 6,216,129, issued April 10, 2001 to Eldering, based on applications filed March 12, 1999 and Dec. 3, 1998. This discloses how advertisements and other product information can be displayed to a potential consumer via a computer, cable TV, video, or other means, based on a consumer profile stored on a profile server, in cooperation with an ad server and/or content server. The price paid for each ad displayed can be proportional to the match between the consumer profile and the specifications given by the advertiser for the intended audience. The correlation can be expressed as a scalar dot product between a consumer characterization vector and an ad characterization vector.

Further background information on electronic marketing and especially Internet marketing can be found in many books and publications, including:

The John Wiley and Sons publication, the Journal of Interactive Marketing, now in its fifteenth year (volume 15 in 2001), is another excellent source to understand past interactive marketing methods and interactive market research tools. Other references on Internet marketing are listed at

Virtual Shopping Technologies

One of the leaders in computer-assisted marketing research is Allison Research Technologies (hereafter ART) of Atlanta, Georgia, which owns the trademark, Virtual Shopping® and has been conducting computer-assisted market research for about 10 years. This form of market research has long involved displaying a computer image of a product or suite of products to a consumer and obtaining a response input from the consumer that is requested, received, and stored via a computer interface.

The Virtual Shopping® methodology of ART includes creating a detailed virtual environment that subjects can interact with in much the same way that they would with the real objects being simulated. Computers are used to create images of products arrayed as in an actual store. The consumer can then interact with the display and provide indications regarding what products are appealing or what the consumer might purchase. Thus, computer-generated virtual environments of products are displayed and inputs from consumers are obtained via a computer interface. The response of the subjects are archived and analyzed to help retailers and manufacturers improve their marketing of goods. Some aspects of the Virtual Shopping® technology are described at the ART Web site at For example, to study how shoppers might seek information about a product being considered for purchase, ART can query each subject to determine what tools the subject might employ, then create a virtual representation of those tools. ART has been employing computers to assist in virtual shopping research for about 10 years. The term Virtual Shopping® was registered as a trademark (Reg. No. 1,881,580) on Feb. 28, 1995 by the Allison Hollander Corporation of Atlanta, now Allison Research Technologies. Thus, computer-assisted virtual technologies for marketing research date back at least to the early 1990s, though less sophisticated marketing research tools involving programs on floppy disks or electronic bulletin boards may have been in use years earlier.

An early online shopping system was the Virtual Emporium, launched in the Third Street Emporium of Santa Monica, California on Nov. 7, 1996. The Virtual Emporium offered consumers the merchandise selection of a two-million-square-foot mall in a 2,500-square-foot neighborhood store. Tuck Edwards, CEO of the Virtual Emporium, discussed the potential for studying consumer behavior and measuring use patterns via online research in his presentation, "Virtual Emporium - A New Shopping Experience," at the Annual Conference of the Association for Consumer Research, Oct. 16-19, 1997, Denver, Colorado (see Advances in Consumer Research, Vol. 25, ed. J.W. Alba and H.W. Hutchinson, Provo, Utah: Association for Consumer Research, 1998, pp. 60-61). Today, numerous models for online shopping have been developed.

The potential of virtual shopping technologies received attention in a 1996 publication, "Virtual Shopping: Breakthrough in Marketing Research" by Raymond R. Burke, Harvard Business Review, Vol. 74, No. 2, March-April 1996, pp. 120-31, abstracted online at (archived).. In this article, Burke discussed the potential of computer-created virtual environments as inexpensive research tools for marketers. Selected excerpts follow:

"Recent advances in computer graphics and three-dimensional modeling promise to bring simulated test marketing to a much broader range of companies, products, and applications. How? By allowing the marketer to re-create -- quickly and inexpensively -- the atmosphere of an actual retail store on a computer screen. . . . For example, in the Harvard Business School's Marketing Simulation Lab, a consumer can view shelves stocked with any kind of product. The shopper consumer 'picks up' a package from the shelf by touching its image on the monitor. In response, the product moves to the center of the screen, where the shopper can use a three-axis trackball to turn the package so that it can be examined from all sides. To 'purchase' the product, the consumer touches an image of a shopping cart, and the product moves to the cart. . . . During the shopping process, the computer unobtrusively records the amount of time the consumer spends in each product category, the time the consumer spends examining each side of a package, the quantity of product the consumer purchases, and the order in which items are purchases. . . .

"Once images of the product are scanned into the computer, the researcher can make changes in the assortment of brands, product packaging, pricing, promotions, and shelf space within minutes. Data collection is fast and error-free because the information generated by the purchase is automatically tabulated by the computer." (p.123)

"An increasing number of companies utilize electronic media to plan and control different aspects of retailing. Point-of-sale terminals track the unit inventory and prices of most existing consumer products. Store floor plans contain information on the dimensions and locations of products on the shelf. In some cases, they also contain scanned images of products and packaging. Industry associations and marketing research companies have assembled this information into comprehensive databases for the grocery, drug, and general-merchandise trades. . . . MarketWare Corporation manages one such database. It is continually updated as new products are introduced. The retailers' floor plans, shelf-layout files, dimension databases, prices, and product images can be used to construct three-dimensional models of the retails store for consumer research.

"In addition, researchers must generate images to represent the marketing ideas they intend to test. In a world where desktop publishing, computer-aided design, and digital production have become commonplace, that process is also becoming easier. More and more marketing elements are being developed on the computer, which means that new advertisements, promotional material, merchandising information, and product and packaging designs are often available in electronic form. These materials can be incorporated directly into a shopping-simulation program. Marketers can use three-dimensional-modeling software to prototype product concepts and create many variations on the theme for testing purposes." (p. 124)

"[O]ne can photograph the storefronts, walls, and shelf displays of physical retail stores and 'wrap' the images onto the corresponding architectural models to create photo-realistic, three-dimensional walkthroughs. Another technology is surround video, in which digitized panoramic photographs of a store can be used as a backdrop for computer generated products and displays.

"Of course, the physical store is not the only place where consumers learn about new products. Shoppers also scan advertisements, review articles, and talk with friends or salespeople before making a purchase. . . . These experiences can be simulated on the computer by supplying shoppers with scanned images of magazine pages, newspaper articles, and product brochures. Television commercials and staged interactions with salespeople can be made available using digital video technology. Researchers at MIT's Sloan School of Management have used this approach to study consumers' reactions to several product concepts, including electric cars and instant cameras. . . .

"At the Harvard Business School's Marketing Simulation Lab, consumer goods are often the subject of investigation. Participants . . . [are] asked to take a series of trips through a simulated store and to shop the same way they would in a conventional store. The trips can be simulated on the computer monitor or through a head-mounted display (HMD) and head-tracking device. An HMD allows for total immersion. . . . (p. 125)

"Goodyear conducted a study of nearly 1,000 people. . . . Each respondent took a trip through a number of different virtual tire stores stocked with a variety of brands and models. . . . Goodyear found the results of the test valuable on several fronts. First, the research revealed the extent to which shoppers in different market segments valued the Goodyear brand over competing brands. Second, the test suggested strategies for repricing the product line. (p. 125)

"To learn about patterns of loyalty and substitution, [a] company created a virtual vending machine stocked with a broad selection of snack foods. Four hundred respondents, recruited from six shopping malls across the United States, purchased items from the vending machines on repeated occasions. In some instances, the respondents' preferred item is made 'temporarily out of stock.' That action allowed the company to measure the percentage of customers who switched between brands and between snack categories when a specific snack was not available. It also allowed the company to assess the demand for each kind of snack overall and by consumer segment."

Thus, consumer input can be solicited, obtained, stored, and used for subsequent analysis by marketers. The consumer input can be in the form of indicating purchase intent via a graphical interface or by entry of text.

Earlier information on the use of virtual shopping was published by Burke et al. in 1992 regarding a study comparing virtual shopping results with real shopping experiences, further supplemented with a text-based computer interfaced to simulate shopping decisions. The publication by Raymond R. Burke, Bari Harlam, Barbara Kahn, and Leonard Lodish, "Comparing Dynamic Consumer Choice in Real and Computer-Simulated Environments," Journal of Consumer Research, 19 (1), 71-82 (June 1992), found that the visual cues of the virtual shopping experience were important in accurately predicting brand market shares and consumer price sensitivity observed in the supermarket. (See also MSI Technical Report, No. 91-116, pp. 1-31.)

Burke (1996, p. 129) suggests that virtual shopping can be made even more realistic by supplementing the visual and auditory aspects of the simulated environment with additional modes of sensory perception such as touch, taste, and smell, though these can be harder to simulate. He mentions force-feedback systems that have been developed to provide a sense of touch in virtual-reality simulations, but warned that they are (were) "primitive and expensive." However, recent gains in virtual reality simulations may open many opportunities for future multisensory shopping environments. Aroma technology is already available to rapidly simulate a plethora of odors. Tactile simulations have also improved substantially since 1996.

Computer-Assisted Customization or Selection of Products

The use of computers, microprocessors, or the Internet to select a product is well known, as are systems of obtaining information from a consumer to customize products or offer a tailored selection of products to meet consumes needs. For example, adult consumers suffering from incontinence could interact with a Web site, a computer-aided kiosk, a personal digital assistant with a wireless connection to a data port, an electronic bulletin board, an e-mail server, or an electronic voice-recognition system via telephone, for example, to provide personal information (e.g., details of the incontinence problem, body size, gender, clothing preferences, other desired performance attributes, etc.) that can be interpreted and processed electronically to identify a suitable product or combination of products, after which the consumers could be provided with information regarding the correct products to buy at a retailer, or with options to receive the product directly via shipment or delivered mechanically from a kiosk or other dispensing device in response to the customer's input.

Similar scenarios can be constructed for the purchase of diapers and baby products, combinations of skin care and other cosmetic products, nutritional products, pet care products, toiletries, tissue products, clothing, automotive goods, computer software and hardware, lawn care products, and the like.

Customer input obtained via electronic or interactive means can also be stored and processed as a marketing research resource. Such data could be especially helpful if the input from the consumer can be linked to additional digital information about the consumer. For example, the consumer may use a loyalty card or smart card in using an electronic kiosk to receive a discount. The card then allows the consumer to be identified, along with relevant historical information about purchases. In another example, the consumer is an Internet user whose shopping history is partly revealed by information associated with cookies from past Web purchases. In another example, identifying information entered by the customer over a computer system or telephone line can be used to access a database containing information about purchases made by the consumer. Alternatively, without requiring identification of individual users, the demographic class or geographic location of the user can be assessed and used to build a database relating consumer needs to demographic or other variables.

For example, a virtual shopping environment can be used to display products of potential value to the consumer, based on product attributes the consumer seeks and based on other personal information entered. The consumer can then virtually explore a variety of products that may be suitable, and make an informed selection. The products can then be immediately delivered to the consumer, shipped, or made ready for pick up at a retail environment.


One instructive example of interactive computer-assisted marketing to help a consumer select a customized product or assortment of related products was described by Kim Ann Zimmermann in "Fashion Trip Combines CD, Web Access," in DNR (a men's fashion magazine of Fairchild Publication, New York), Vol. 28, No. 101, Aug. 26, 1998, p. 40. Zimmerman described how dozens of apparel, cosmetics, and footwear manufacturers and retailers will use the "Fashion Trip CD" system for "virtual shopping" combining Internet access with an interactive compact computer disk. Shoppers use the interactive environment to view and "try on" apparel and make-up using digital mannequins based on their body type and skin tone. If they want to purchase merchandise, they can go to an individual vendor's Web site to place an order or look for the nearest retail outlet. Although the new CD is aimed at young women, other demographic groups may be chosen for similar projects.

FashionTrip uses 3-dimensional technology originally from ModaCAD Inc. (formerly of Culver City, CA , the company is now of Chicago, IL) to create the feeling that users get as they walk through a shopping mall. Shoppers can "visit" individual stores and select clothes and cosmetics to try on computerized models. Participants in the project include Bongo, The Sak, Wet Seal, Guess Footwear, Nicole Miller, Almay, and Clinique. Users can access a Web site with the CD to obtain fashion news and advice and information about music and entertainment.

"The Web access via the CD will also enable users to view items simultaneously with other users, enabling them to try on apparel on side-by-side computerized models.

"The Web site, which can also be accessed from the CD, will provide fashion news and advice, and music and entertainment information. From the site, users will be able to link directly to specific areas of the Web site geared toward the same demographic group. "It really has a lot of play value, which is an ideal way to reach our target demographic of young women," said Reed. "The idea that they can call up a friend and they can both be trying on clothes on a model suited to their body types that they can both see is very appealing," she said . . . .

"Joyce Freedman, chief executive for ModaCAD, noted that the variety of designers participating is key to reaching the target audience. "The ability to mix and match their favorite brands is key," she said. The CD incorporates a search engine where items are sorted by brand, color or product category. Users can view all of the white shirts, for example, or all of the skirts from a particular designer . . . .

"ModaCAD said the virtual stores will receive updates as retailers and apparel makers change their offerings and users will be able to receive updates via the Internet.

"Freedman said ModaCAD is developing similar projects aimed at other demographic groups as well as different areas of retail."

Another related example is the interactive marketing of, described in "Global Retailing in the Connected Economy: Part 1 of 2," Chain Store Age Executive, Vol. 75, No. 12, Dec. 1999, pp. 69-74:

"The market has to [be] relevant to the individual consumer. On one level, this means understanding the target customer and merchandising accordingly. The local store with the staff that knows its customers and stocks its stores to suit their tastes is the classic model of customer-driven merchandising.

"Global retailers can accomplish this by utilizing POS technology and customer-relationship software, and by staffing stores with people who know the local language and culture.

"E-tailers can duplicate this model, too, with the help of interactive technology and customization., for instance, has a "personal model," enabling customers to personalize the Web site with their measurements and personal preferences, and make use of a virtual dressing room and real-time personal shoppers.

" and are using collaborative filtering to learn customer preferences and filter information on their behalf, anticipate their interests and offer customized service."


Mining Data from Actual Purchases

Data regarding actual purchases made by consumers can be extremely valuable for marketing research. Such data can be provided directly from the point of sale using scanned UPC information. An example is ScanTrak™ point-of-sale data obtained from a variety of retailers. Such data is typically divorced from the identity of the individuals making the purchase, but do show what was purchased and what items were purchased together (by single consumers), what price was paid, and so forth. Further, retailers using loyalty cards or smart cards can obtain detailed information about the historical buying habits of individual consumers, which can be used to gain further insight into purchase habits.

Another class of data involving actual purchases is household panel data. Typically, a group of consumers is selected to be representative of the population as a whole or of a target population. The selected consumers, who agree to participate in the generation of household panel data and may be compensated for their role, are asked to scan each item they purchase using an electronic device in their home. Further, they may provide additional information such as where the purchase was made, what price was paid, and so forth. The data so acquired can then be transmitted to a central source. Data from many individuals can be combined to provide a useful database representative of purchase habits for a large sector of the population. Examples of commercially available household panel data services include the services of AC Nielsen and IRI. AC Nielsen, for example, uses about 55,000 households balanced according to the US census relative to income, age, education level, and so forth. The data can be analyzed to show consumer purchase patterns as a function of demographic and socioeconomic variables.

In addition to raw purchase data, marketers are often interested in attitudinal variables as well, such as the emotional reaction of a consumer to a product, the impressions that a product package or advertisement made, the reasons for a purchase, the degree of satisfaction with a product, and so forth. There are many ways to combine attitudinal data with household panel data.

Purchase patterns for a particular household obtained from market panel data can be combined with results of a survey taken by individuals in the household regarding their views on issues such as the relationship between national brands and product quality, the importance of coupons or temporary price reductions in making a purchase, the willingness to switch brands in certain product categories, and so forth. By relating consumer attitude to purchase patterns, much information of value to the marketer can be obtained. Computers and electronic databases are standard tools in such methods, and sophisticated forms of data analysis are needed to determine the most important trends. Fuzzy logic tools and neural networks may be of value in mining such data.

Data from multiple sources can be better unified when the household panel members use loyalty cards or smart cards that allow identification of the consumer during shopping, such that point-of-sale data is automatically combined with historical purchase data and other data for the household or consumer in question (providing that a network is established to do this), including past data scanned as part of the household panel study, or data available on a smart card, all of which can be combined with available attitudinal data for the consumer.

For example, a manufacturer can track changes in purchase patterns during a promotion of a particular product and relate the apparent effect of the promotion to consumer attitudes about promotions and purchase decisions to understand how future promotions might be more effectively tailored for a targeted demographic group.

Internet surveys or other computer-assisted survey techniques can be a useful tool in providing the attitudinal data or other data that can be combined with household product data, loyalty card data, smart card data, or general point-of-sale data from a region or demographic group being studied.

Surveys of targeted groups can be obtained in several ways. In most cases, the chances of receiving a completed survey increase if there is an incentive to the customer to participate. For example, loyalty card users may be offered an instant discount to complete an in-store survey, such as an interactive electronic survey provided on a touch-sensitive electronic screen while the customer is checking out or before or after check out. Loyalty card users or others may also be offered "points" that can be accumulated and redeemed for products or prizes. Surveys may also be offered to visitors of a Web site, perhaps with an opportunity to enter a contest or an offer of a discount coupon or other economic incentive for completion of the survey. Surveys may also be solicited by mailings or by e-mail. For example, a database of e-mail addresses for a targeted group such as purchasers of books or new parents may be used to solicit completion of a survey from those in the database, optionally with an economic incentive offered for participation.

In addition to scanners used by household panelists, hand-held scanners and other electronic devices (including wireless communications devices) have been used by consumers to help generate useful data for marketing research. For example, Safeway once gave some customers a hand-held "Easi-Order" device that allowed customers to download personalized grocery lists, compiled from their loyalty card information, before placing an order (see "Internet Retailing," Euromonitor, January 2000).

In-Store Observation for Marketing Research

While computer-generated virtual environments can inexpensively simulate actions of consumers in retail environments, computerized tools can also be deployed in the retail environment to measure and analyze the response of consumers to products in a real setting. Computer vision systems are a particularly promising tool for automating such research. For example, hidden cameras can observe the actions of a user in front of a display panel, an advertisement, or a shelf of products. Computer vision can track the motion of eyes, head, and hands, for example, to determine what features draw attention or elicit actions such as picking up an object, reading its label, or putting it in a shopping cart.

Several uses of such technology are mentioned in the following excerpt from the Innovation section of Technology Review, May 2001, p. 32:

"Engineers at IBM's Almaden Research Center in San Jose, CA, report that a number of large retailers have implemented surveillance systems that record and interpret customer movements, using software from Almaden's BlueEyes research project. BlueEyes is developing ways for computers to anticipate users' wants by gathering video data on eye movement and facial expression. . . .

"BlueEyes software makes sense of what the cameras see to answer key questions for retailers, including, How many shoppers ignored a promotion? How many stopped? How long did they stay? Did their faces register boredom or delight? How many reached for the item and put it in their shopping carts? BlueEyes works by tracking pupil, eyebrow and mouth movement. When monitoring pupils, the system uses a camera and two infrared light sources placed inside the product display. One light source is aligned with the camera's focus; the other is slightly off axis. When the eye looks into the camera-aligned light, the pupil appears bright to the sensor, and the software registers the customer's attention."

A similar system was mentioned in "Big Brother Logs On," Technology Review, September 2001, p. 61:

"From video signals, the Carnegie Mellon system detects and tracks both invariant aspects of a face, such as the distance between the eyes, and transient ones, like skin furrows and smile wrinkles. This raw data is then reclassified as representing elemental actions of the face. Finally, a neural network correlates combinations of these measurable units to actual expressions. While this falls short of robotic detection of human intentions, many facial expressions reflect human emotions, such as fear, happiness or rage, which, in turn, often serve as visible signs of intentions."

If the consumer can be identified, such as through the use of a loyalty card, voluntary entry of user identification (perhaps motivated by the chance to win a prize or receive a discount), biometrics, or image recognition, the subsequent data obtained from observation of the consumer can be related to historical information regarding the purchasing behavior of the individual or household of the individual. Household panel members, for example, may identify themselves using a swipe card or ID number when in a store with interactive displays for market research. In one scenario, a shopping card has a transmitter that can be used to track the motion of an individual around a store. Physical motion through aisles -- including high dwell time in regions of interest -- can be recorded and related to other information available regarding the individual or the items placed in the card and purchased. Such studies can also be conducted in a virtual environment.

Fuzzy logic systems and/or neural networks may be employed to sift through the data generated and identify meaningful trends of predictive tools.

Electronic Auctions for Promotions

Catalina Marketing Corporation (St. Petersburg, Florida) and other corporations have developed many useful tools for providing targeted coupons to consumers. For example, based on the scanned items purchased by a consumer, a computer may detect that the consumer has purchased toys suited for an infant. In response, the sales receipt may be printed with a coupon for related products such as baby food on the back side of the receipt, or a separate coupon may be printed in addition to the receipt.

Looking forward, we propose that electronic vendor auctions will become an important part of marketing efforts. Electronic auctions refer to the use of computer or microprocessor to determine which of one or more competing products will be promoted to the consumer, wherein the vendors of the competing products offer the retailer or other third party an incentive such as cash for the privilege of excluding the competitor in making a promotion to the consumer. Electronic auctions between mutliple advertisers, for example, could readily be built into the advertisement selection systems proposed by Eldering in US Pat. No. 6,216,129, issued April 10, 2001.

For example, when a targeted coupon is to be printed out for a consumer, several vendors may have products that could be the basis for coupons to be given to the consumer. If the consumer has purchased baby food, two or more vendors of diapers may wish to provide the consumer with a coupon for a diaper product. A computer program under control of the vendor may then electronically receive an offer from the competitors and select the most lucrative one to determine which competing product is selected for the issuance of a coupon. The amount offered by the vendors may be programmed to be a function of the present purchase and, if available, data on past purchases by the consumer. For example, if a vendor's computer system recognizes that a consumer is a loyal user of a competitor's product, a higher-than-normal incentive to the retailer may be offered for the privilege of issuing a coupon, in hopes of switching the consumer away from the competitor's products.

In other cases, the vendor may allow for competing coupons to be printed unless a bonus is offered to achieve exclusivity. The retailer may also wish to limit the printed coupons to a fixed number, or to a limited number of product categories, and can offer manufacturers of a wide variety of products the opportunity to auction for the available slots, whether the products represented are competing products or not.

Generally, the auction is conducted automatically by one or more computers. Data may be uploaded to a single computer by multiple vendors to define an auction strategy, or two or more vendor computers may interact with a retailer's computer to issue offers in an auction.

Targeted Coupons and Other Promotions

Improved targeted coupons with custom graphics can be printed at the point of sale or elsewhere in a retail environment in response to a consumer being identified via a smart card, loyalty card, or other means. The product featured, the level of discount and the graphics printed thereon can all be tailored to consumer data in a database. Selected coupons can be customized according to consumer profile information to increase the appeal of the coupon. Details such as the background color, border designs, fonts, text color, images on the coupon, and so forth can be selected according to categorical profiles pertaining to the consumer. For example, such details can be varied according to general preferences linked to gender, age, children, product preferences, payment method, zip code, etc. Consumer database information is accessed once the customer has been identified through use of a loyalty card, encoded coupon with a digital identifier, or other identifying means. For example, if the consumer is known to have purchased diapers and dog food, a diaper coupon may be printed with a picture of a baby playing with a puppy.

A targeted coupon, whether delivered to a consumer at the point-of-sale, mailed to the consumers home, or delivered by other means, may be printed with a unique code that can be tracked upon redemption of the coupon to provide marketing information about the effectiveness of the coupon for particular users or users classified into particular categories, such as upper income dog-owners, suburban dwellers with SUVs, etc.

Advances with Shopping Carts and Other Mobile Devices

Many innovations have been made in recent years regarding interactive shopping carts and instrumented carts for retail shopping. Some of these advances can be extended to any mobile device relevant to shopping, such as a handheld personal data assistant, an automobile, or other mobile device. They can also be extended to include targeted promotions to the user to provide for a more interactive "live" shopping experience. The instrumented carts or other devices can also provide information to a host computer about shopping habits and rapid feedback about consumer response to offered promotions or displays to provide a tool for further market research. Such research can involve patterns of motion in a retail environment as a function of user demographics and promotions, or can involve acquiring data on the purchase response or apparent interest level of consumers to various in-store promotions. Interest level can be gauged indirectly by purchase activity or dwell time in an area, or by input entered by the consumer on an interactive device, or by video camera monitoring of the consumer.

In one example, a shopping cart (or, alternatively, an automobile) can be equipped with electronic means to provide targeted promotions to a consumer who has been classified or identified through the use of a loyalty card or other means. The targeted promotions such as multimedia advertisements, special discounts, or electronic coupons that are provided to the user can be a function of the location of the consumer. For example, when the consumer is near the produce section of a grocery store, an interactive screen may display promotions for fruits or vegetables, such as a temporary price reduction on broccoli or an instant coupon for an apple dip product. Third-party incentives could also be offered, such as frequent flyer miles or free long-distance telephone time if a featured product is purchased.

Targeted promotions offered via an interactive display or other means to the consumer can also be a function of historical data from the particular consumer (e.g., data for the consumer from a retail database accessed through use of a loyalty card). Thus, if it is known that the consumer has made recent purchases of diapers, a display device on the shopping cart may offer an instant discount on a specific brand of diapers. The offered promotion may also be determined by an electronic auction in which competing vendors automatically bid for the right to provide an exclusive promotion to the consumer. Thus, based on the past purchase history of the consumer, a program may select automatically-generated offers from two or more competing manufacturers to select the promotion most profitable to the retailer or most likely to entice a purchase from the consumer. Thus, personalized, targeted, dynamically- displayed promotions can be directed toward a consumer at locations in the store or in other environments where the promotion will most likely result in a purchase.

The location of the consumer (or the cart or automobile or other device) can be obtained by triangulation, such as with a GPS device attached to the cart or an in-store triangulation device or other position-sensing device based upon sensing a signal emitted from the cart (radio signal, infrared beams, etc.). The location of the cart can also be sensed by photodetectors or miniature cameras on the cart that read markings or colors on the floor or baseboards or ceiling, for example. The location can also be detected by sensors installed in various locations of the store that receive and interpret a steady or periodic signal emitted from the cart. Further, an electronic scanning device associated with the cart or consumer can be used to scan goods during shopping, providing information that can be interpreted to specify a location. Once the location is known, targeted promotions reflecting that location can be provided.

Catalina Marketing Corporation offers several coupon related technologies, for example. Once promotions are selected, they can be provided to the consumer through electronic devices attached to a shopping cart, such as a flat color screen or audio speakers or both. The screen can be a touch screen to permit interaction with the consumer, wherein the consumer may make selections, provide input, play games, cancel display of unwanted promotions or select desired promotions (discounts, free gifts, points, or other bonuses). Recipes and gift ideas may also be displayed, as well as news, weather information, Internet sites, and the like.

The selection of promotions may occur on a central server in communication with the cart or other device, or by a processor attached to the cart that responds to location signals and to classification information about the consumer.

Thus, for example, an interactive shopping cart may provide promotional displays on a color screen or promotional messages from speakers or headphones, wherein nearby goods are being promoted. For example, as the shopping cart enters an aisle featuring pet supplies, the promotions presented to the consumer may feature offers for discounts on dog food, if the consumer database for the consumer indicates that the consumer owns a dog. Alternatively, a consumer driving a car may be provided with an audio description of discounts for a local fast-food business when the automobile is within a certain distance of the business. Game-like features including contests and sweepstakes may be incorporated into an interactive display to promote goods or to motivate the user to visit specific store locations.

Classifying the consumer is a process in which the consumer is identified in some way to permit subsequent targeting to be meaningful. For example, a consumer may be identified by using a loyalty card (frequent shopper card) upon entering the store. Use of such a card may be required to activate the display device on the shopping cart or to access the "premium" cart at all. Use of the loyalty card provides identifying information that can be used to access data concerning the consumer in a consumer database. The card also may be a "smart card" having consumer information stored thereon. The consumer may provide other forms of identification through card scans (e.g., a credit card), manual data entry (e.g., username and PIN), etc. If desired, the consumer can be identified automatically (at least in part) by means of biometrics such as voice matching, iris scanning, thumb print identification, and so forth (see "Biometrics Not Just Fantasy Anymore" by Doug Brown, [email protected] Week, Sept. 25, 2000, p. 104). Classification can also occur by obtaining input from the consumer in response to questions or alternatives offered to the consumer, such as a small survey form offered on a touch screen on the cart to be completed before using the cart.

The cart also may include a scanning device to permit self-serving check out and payment for the items purchased. The scanning device could provide location information to a processor in the cart or, by wireless signal, to a central processor for tracking of cart position. Selected goods would also be immediately detected and that information could be used as a basis for other targeted promotions and offers.

Embodiments using an automobile generally would not need the step of classifying the computer. The automobile may be uniquely identified and tracked by a signal it emits, or by an identification code issued when the driver requests promotional input.

Continuous Data Acquisition

Acquisition of data from consumers in many venues is expected to occur, continuing present trends in which computers and microchips are increasingly relied upon for marketing research. In addition to current uses of data obtained at the point of sale (e.g., from loyalty cards) or from household panels, where products are scanned regularly to provide product data to organizations such as AC Nielsen or IRI, other computer-assisted systems are evolving or are under consideration to enable manufacturers and marketers to better understand consumer needs, consumer patterns of product use, purchase behavior, and consumer attitudes towards products. Some of these concepts are disclosed in the book The Websolution of Shopping to 2010 by Frank Feather, Toronto, Ontario: Warwick Publishing, 2000. Feather forecasts increased use of the Internet to continuously connect people with their homes, automobiles, their appliances, their work, and so forth. Wireless communication is expected to be increasingly important.

Marketing research in such a heavily interconnected future can pose many opportunities and challenges, particularly in the vast quantities of data that can be acquired and processed. The data may be provided to marketers or other third parties directly from transmissions by microchips contained in the products or associated with the products, typically with permission from the user. For example, instrumented appliances or other articles can be used, wherein a microchip or other signal generation device in the article senses one or more parameters relating to product use and wherein means are provided for sending a signal to a third party containing data about the sensed one or more parameters. For example, a refrigerator may have a microchip and associated electronics that detect when a door is opened, how long it is opened, what the temperature in the refrigerator is, what the power usage is, and so forth. Such data can be transmitted to the manufacturer or vendor to increase knowledge of consumer usage patterns, and to better understand features that are important for successfully meeting consumer needs. Consumer attitudes and other information can be solicited by electronic means and transmitted with the measured usage data (e.g., using an Internet form or electronic panel associated with the article) or via separate means (e.g., a survey form or telephone interview).

Users of such instrumented articles may be compensated for agreeing to make product use and performance data available. For example, the article may be sold at a reduced price for those agreeing to use an instrumented article, or a free upgrade or reduced upgrade price may be offered for a future improved product. Information obtained from instrumented articles can be used for direct marketing of related products to the user in addition to providing market research data and product performance information.

Continuous feedback from purchased products regarding consumer use patterns and, optionally, consumer response to the product can be regularly mined for market information.

Dynamic Price Studies

The effect of price on purchasing can be studied more rapidly by allowing the effective price of an article to vary dynamically in a retail environment, within a range acceptable to the retailer, and optionally with guaranteed profit levels or returns for the participating retailer. The price may be displayed electronically (e.g., via an LCD screen or LED display) for a variety of goods, with the vendor or retailer being able to automatically change the displayed price and price entered in the retailer's product database that may be accessed at the time of purchase. This provides rapid information about price effects. This can be supplemented with virtual shopping, but the ability to track real purchases as a function of price may offer improved information in a real setting.

Fuzzy Logic in Marketing

New methods of data representation and analysis have dramatically increased the ability of computers to process information intelligently. Computers have traditionally excelled in tasks that required a large number of small, repetitive tasks. A group of methods that fall under the rubric of "soft" computing have enabled computers to model the human reasoning process and allow computers to make or assist with complex problems such as pattern recognition, strategic planning, decision support, etc.

Of the various tools employed in "soft" computing, fuzzy logic and artificial neural networks have been at the forefront of increased machine intelligence (see, L. A. Zadeh, "Fuzzy Logic, Neural Networks, and Soft Computing," Comm. ACM, vol. 37, no. 3, pp. 77-84, Mar. 1994).

Fuzzy logic is a way to quantify and add functionality to the vagueness, imprecision, and uncertainty. Fuzzy logic blurs the boundaries between groups and classifications, and allows an element to partially belong to more than one set. One salient aspect of fuzzy logic is that it allows computers to process abstract or subjective concepts that are represented with linguistic variables (e.g., "somewhat hot", "very expensive").

Fuzzy logic has also found application in approximation modeling. As systems become more complex -- such as with increasing parameter space or nonlinearity -- it becomes more difficult and costly to model them, with diminishing increases in utility with the increased precision. One of the more powerful features of a fuzzy system is that it can smoothly and universally model a system without the need to know the underlying governing equations. As an illustration, thousands of truck drivers back up a semi to a loading dock every day; they can express in non-mathematical language their method of accomplishing the task. ("If the truck turns too far to the left, turn the steering wheel a little to the right.") Fuzzy systems have been able to encode this linguistic knowledge and actually perform the task of backing up a truck without the use of mathematical equations.

A second tool that has been gaining use is the artificial neural network (ANN). An ANN is a collection of inter-connected neurons that map various input variables to output variables. ANNs can be trained to recognize and extract patterns from data without any a priori knowledge of what the patterns might look like. Because of their power and flexibility, neural networks have been used in data mining applications.

Because fuzzy logic and neural networks have had success in other fields such as engineering and finance, they have seen increasing use in marketing applications. Fuzzy logic and neural networks in a variety of forms can be applied to enhance current market research practices, including the processing of point-of-sale data, of household panel data, calculation of elasticities and cross-elasticities, prediction of sales lift caused by advertising or other promotions, inventory management, interpretation of household panel sales data in light of attitudinal surveys, selection of attitudinal variables to correlate with sales data, and the like. Fuzzy logic and neural networks can be especially helpful in interpreting virtual marketing data, and can even be adapted in real time to intelligently select scenarios presented to the participants of a virtual shopping study based on the response to prior scenarios to resolve ambiguities or to improve the usefulness of the data.

Use on the Internet

A number of implementations of fuzzy logic and neural networks have been implemented or proposed for use in Internet marketing.

Ronald Yager from Iona College proposed and outlined the use of fuzzy intelligent agents for targeted Web marketing (Ronald R. Yager, "Targeted E-commerce Marketing Using Fuzzy Intelligent Agents, IEEE Intelligent Systems and their Applications, vol. 15, no. 6, pp. 42-45, 2000). In the most basic form of Web advertising, an advertiser purchases a subscription that indicates how often an ad would appear on a Web page. Other advertisers are also able to purchase subscriptions, and which advertiser's ad is displayed is chosen at random in proportion to the size of each subscription.

Fuzzy intelligent agents would process information about a Web page visitor (e.g., age and income) and use fuzzy inference to determine which of a number of ads to display (i.e., if age is young and income is average, then . . .). Each advertiser's fuzzy agent would asses the available information and bid accordingly for the ad space. Once an advertiser won the ad space, the agent could also determine which, of several, ads to place.

In a separate publication, Yager outlined a fuzzy method of improving the quality of information available to on-line consumers, with the goal of gaining customer goodwill by providing unbiased and easily understandable product information (Ronald R. Yager and Gabriella Pasi, "Product Category Description for Web-Shopping in E-Commerce," International Journal of Intelligent Systems, vol. 16, pp. 1009-1021, 2001). In this method, products are clustered into fuzzy price categories (low end, moderate, high end) and linguistic descriptions of the relevant product features associated with each category are defined. Ultimately, the consumer is able to extract summaries with respect to a feature, such as, "Most TV's in the high end price range category provide extremely high resolution." With the availability of such information, it would be easier for the consumer to "understand the product line, see what they are getting for their money, and more easily and confidently locate products that are of particular value for the money."

Data Mining

Fuzzy logic has proved an effective way of extracting and representing information from large data sets. For example, Setnes and Kaynak demonstrated the use of fuzzy clustering in direct marketing target recognition. They used a 170-feature database from the campaigns of a large financial services provider to extract fuzzy rules governing which clients to target. Based on these rules, the fuzzy model approach improved target recognition by approximately 20% over a number of other statistical methods (M. Setnes and U. Kaymak, "Fuzzy Modeling of Client Preference in Data-Rich Marketing Environments," Research in Management, ERIM Report Series No. ERS-2000-49-LIS, Erasmus Research Institute of Management (ERIM), Rotterdam School of Management, Erasmus Universiteit, Rotterdam, Aug. 2000, pp. 1-10, available online at ; see also M. Setnes and U. Kaymak, "Fuzzy Modeling of Client Preference from Large Data Sets: An Application to Target Selection in Direct Marketing," IEEE Transactions on Fuzzy Systems, Vol.9, No.1 (Feb. 2001), pp.153-63; M. Setnes, U. Kaymak, U., and H.R. van Nauta Lemke, H.R., "Fuzzy Target Selection in Direct Marketing," Proceedings of the IEEE/IAFE/INFORMS 1998 Conference on Computational Intelligence for Financial Engineering (CIFEr), New York: IEEE (Cat. No.98TH8367), 1998, pp. 92-97; and M. Setnes, R. Babuska, U. Kaymak, and H. R. van Nauta Lemke, "Similarity Measures in Fuzzy Rule Base Simplification," IEEE Transactions on Systems, Man and Cybernetics -- Part B: Cybernetics, vol. 28, no. 3, pp. 376-86, 1998).

Preference Modeling

Fuzzy logic has also been applied to consumer preference modeling, where product features may interact nonlinearly in determining preference. In one example, a fuzzy preference model was constructed to predict the consumer rating of chocolate chip cookies based on dough lightness, size, and the amount of visible chips. A neural network extracted the fuzzy membership functions from consumer data, and the resulting model was shown to integrate product features similarly to how consumers make judgments. Ultimately, the model was combined with an automated inspection system during the manufacturing process where a digital image of the cookie was used to extract features and predict consumer preference online, allowing only cookies deemed "acceptable" and higher to be sold (Valerie J. Davidson, Joanne Ryks, and Terrence Chu, "Fuzzy Models to Predict Consumer Ratings for Biscuits Based on Digital Image Features," IEEE Transactions of Fuzzy Systems, vol. 9, no. 1, pp. 62-67).


The techniques of market research and marketing have not changed dramatically in recent years, but the electronic tools available have allowed many of these techniques to be done faster, less expensively, or more accurately. Though this review is far from comprehensive, we hope it provides some insight regarding the status of interactive marketing research, and ways it can be or has been enhanced with modern technology.

Curator: Jeff Lindsay,   Contact:
Last Updated: Dec. 27, 2001

URL: ""