Consumer Research Methods
Market research is often needed to ensure that we produce what customers really want and not what we think they want.
Primary vs. secondary research methods. There are two main approaches to marketing. Secondary research involves using information that others have already put together. For example, if you are thinking about starting a business making clothes for tall people, you don’t need to question people about how tall they are to find out how many tall people exist—that information has already been published by the U.S. Government. Primary research, in contrast, is research that you design and conduct yourself. For example, you may need to find out whether consumers would prefer that your soft drinks be sweater or tarter.
Research will often help us reduce risks associated with a new product, but it cannot take the risk away entirely. It is also important to ascertain whether the research has been complete. For example, Coca Cola did a great deal of research prior to releasing the New Coke, and consumers seemed to prefer the taste. However, consumers were not prepared to have this drink replace traditional Coke.
Secondary Methods. For more information about secondary market research tools and issues, please see my handout.
Primary Methods. Several tools are available to the market researcher—e.g., mail questionnaires, phone surveys, observation, and focus groups. Please see my handout for advantages and disadvantages of each. The below flow chart suggests appropriate choices for different types of information needs.
Surveys are useful for getting a great deal of specific information. Surveys can contain open-ended questions (e.g., “In which city and state were you born? ____________”) or closed-ended, where the respondent is asked to select answers from a brief list (e.g., “__Male ___ Female.” Open ended questions have the advantage that the respondent is not limited to the options listed, and that the respondent is not being influenced by seeing a list of responses. However, open-ended questions are often skipped by respondents, and coding them can be quite a challenge. In general, for surveys to yield meaningful responses, sample sizes of over 100 are usually required because precision is essential. For example, if a market share of twenty percent would result in a loss while thirty percent would be profitable, a confidence interval of 20-35% is too wide to be useful.
Surveys come in several different forms. Mail surveys are relatively inexpensive, but response rates are typically quite low—typically from 5-20%. Phone-surveys get somewhat higher response rates, but not many questions can be asked because many answer options have to be repeated and few people are willing to stay on the phone for more than five minutes. Mall intercepts are a convenient way to reach consumers, but respondents may be reluctant to discuss anything sensitive face-to-face with an interviewer.
Surveys, as any kind of research, are vulnerable to bias. The wording of a question can influence the outcome a great deal. For example, more people answered no to the question “Should speeches against democracy be allowed?” than answered yes to “Should speeches against democracy be forbidden?” For face-to-face interviews, interviewer bias is a danger, too. Interviewer bias occurs when the interviewer influences the way the respondent answers. For example, unconsciously an interviewer that works for the firm manufacturing the product in question may smile a little when something good is being said about the product and frown a little when something negative is being said. The respondent may catch on and say something more positive than his or her real opinion. Finally, a response bias may occur—if only part of the sample responds to a survey, the respondents’ answers may not be representative of the population.
Focus groups are useful when the marketer wants to launch a new product or modify an existing one. A focus group usually involves having some 8-12 people come together in a room to discuss their consumption preferences and experiences. The group is usually led by a moderator, who will start out talking broadly about topics related broadly to the product without mentioning the product itself. For example, a focus group aimed at sugar-free cookies might first address consumers’ snacking preferences, only gradually moving toward the specific product of sugar-free cookies. By not mentioning the product up front, we avoid biasing the participants into thinking only in terms of the specific product brought out. Thus, instead of having consumers think primarily in terms of what might be good or bad about the product, we can ask them to discuss more broadly the ultimate benefits they really seek. For example, instead of having consumers merely discuss what they think about some sugar-free cookies that we are considering releasing to the market, we can have consumers speak about their motivations for using snacks and what general kinds of benefits they seek. Such a discussion might reveal a concern about healthfulness and a desire for wholesome foods. Probing on the meaning of wholesomeness, consumers might indicate a desire to avoid artificial ingredients. This would be an important concern in the marketing of sugar-free cookies, but might not have come up if consumers were asked to comment directly on the product where the use of artificial ingredients is, by virtue of the nature of the product, necessary.
Focus groups are well suited for some purposes, but poorly suited for others. In general, focus groups are very good for getting breadth—i.e., finding out what kinds of issues are important for consumers in a given product category. Here, it is helpful that focus groups are completely “open-ended:” The consumer mentions his or her preferences and opinions, and the focus group moderator can ask the consumer to elaborate. In a questionnaire, if one did not think to ask about something, chances are that few consumers would take the time to write out an elaborate answer. Focus groups also have some drawbacks, for example:
- They represent small sample sizes. Because of the cost of running focus groups, only a few groups can be run. Suppose you run four focus groups with ten members each. This will result in an n of 4(10)=40, which is too small to generalize from. Therefore, focus groups cannot give us a good idea of:
- What proportion of the population is likely to buy the product.
- What price consumers are willing to pay.
- The groups are inherently social. This means that:
- Consumers will often say things that may make them look good (i.e., they watch public television rather than soap operas or cook fresh meals for their families daily) even if that is not true.
- Consumers may be reluctant to speak about embarrassing issues (e.g., weight control, birth control).
Personal interviews involve in-depth questioning of an individual about his or her interest in or experiences with a product. The benefit here is that we can get really into depth (when the respondent says something interesting, we can ask him or her to elaborate), but this method of research is costly and can be extremely vulnerable to interviewer bias.
To get a person to elaborate, it may help to try a common tool of psychologists and psychiatrists—simply repeating what the person said. He or she will often become uncomfortable with the silence that follows and will then tend to elaborate. This approach has the benefit that it minimizes the interference with the respondent’s own ideas and thoughts. He or she is not influenced by a new question but will, instead, go more in depth on what he or she was saying.
Personal interviews are highly susceptible to inadvertent “signaling” to the respondent. Although an interviewer is looking to get at the truth, he or she may have a significant interest in a positive consumer response. Unconsciously, then, he or she may inadvertently smile a little when something positive is said and frown a little when something negative is said. Consciously, this will often not be noticeable, and the respondent often will not consciously be aware that he or she is being “reinforced” and “punished” for saying positive or negative things, but at an unconscious level, the cumulative effect of several facial expressions are likely to be felt. Although this type of conditioning will not get a completely negative respondent to say all positive things, it may “swing” the balance a bit so that respondents are more likely to say positive thoughts and withhold, or limit the duration of, negative thoughts.
Projective techniques are used when a consumer may feel embarrassed to admit to certain opinions, feelings, or preferences. For example, many older executives may not be comfortable admitting to being intimidated by computers. It has been found that in such cases, people will tend to respond more openly about “someone else.” Thus, we may ask them to explain reasons why a friend has not yet bought a computer, or to tell a story about a person in a picture who is or is not using a product. The main problem with this method is that it is difficult to analyze responses.
Projective techniques are inherently inefficient to use. The elaborate context that has to be put into place takes time and energy away from the main question. There may also be real differences between the respondent and the third party. Saying or thinking about something that “hits too close to home” may also influence the respondent, who may or may not be able to see through the ruse.
Observation of consumers is often a powerful tool. Looking at how consumers select products may yield insights into how they make decisions and what they look for. For example, some American manufacturers were concerned about low sales of their products in Japan. Observing Japanese consumers, it was found that many of these Japanese consumers scrutinized packages looking for a name of a major manufacturer—the product specific-brands that are common in the U.S. (e.g., Tide) were not impressive to the Japanese, who wanted a name of a major firm like Mitsubishi or Proctor & Gamble. Observation may help us determine how much time consumers spend comparing prices, or whether nutritional labels are being consulted.
A question arises as to whether this type of “spying” inappropriately invades the privacy of consumers. Although there may be cause for some concern in that the particular individuals have not consented to be part of this research, it should be noted that there is no particular interest in what the individual customer being watched does. The question is what consumers—either as an entire group or as segments—do. Consumers benefit, for example, from stores that are designed effectively to promote efficient shopping. If it is found that women are more uncomfortable than men about others standing too close, the areas of the store heavily trafficked by women can be designed accordingly. What is being reported here, then, are averages and tendencies in response. The intent is not to find “juicy” observations specific to one customer.
The video clip with Paco Underhill that we saw in class demonstrated the application of observation research to the retail setting. By understanding the phenomena such as the tendency toward a right turn, the location of merchandise can be observed. It is also possible to identify problem areas where customers may be overly vulnerable to the “but brush,” or overly close encounter with others. This method can be used to identify problems that the customer experiences, such as difficulty finding a product, a mirror, a changing room, or a store employee for help.
Online research methods. The Internet now reaches the great majority of households in the U.S., and thus, online research provides new opportunity and has increased in use.
One potential benefit of online surveys is the use of “conditional branching.” In conventional paper and pencil surveys, one question might ask if the respondent has shopped for a new car during the last eight months. If the respondent answers “no,” he or she will be asked to skip ahead several questions—e.g., going straight to question 17 instead of proceeding to number 9. If the respondent answered “yes,” he or she would be instructed to go to the next question which, along with the next several ones, would address issues related to this shopping experience. Conditional branching allows the computer to skip directly to the appropriate question. If a respondent is asked which brands he or she considered, it is also possible to customize brand comparison questions to those listed. Suppose, for example, that the respondent considered Ford, Toyota, and Hyundai, it would be possible to ask the subject questions about his or her view of the relative quality of each respective pair—in this case, Ford vs. Toyota, Ford vs. Hyundai, and Toyota vs. Hyundai.
There are certain drawbacks to online surveys. Some consumers may be more comfortable with online activities than others—and not all households will have access. Today, however, this type of response bias is probably not significantly greater than that associated with other types of research methods. A more serious problem is that it has consistently been found in online research that it is very difficult—if not impossible—to get respondents to carefully read instructions and other information online—there is a tendency to move quickly. This makes it difficult to perform research that depends on the respondent’s reading of a situation or product description.
Online search data and page visit logs provides valuable ground for analysis. It is possible to see how frequently various terms are used by those who use a firm’s web site search feature or to see the route taken by most consumers to get to the page with the information they ultimately want. If consumers use a certain term frequently that is not used by the firm in its product descriptions, the need to include this term in online content can be seen in search logs. If consumers take a long, “torturous” route to information frequently accessed, it may be appropriate to redesign the menu structure and/or insert hyperlinks in “intermediate” pages that are found in many users’ routes.
Scanner data. Many consumers are members of supermarket “clubs.” In return for signing p for a card and presenting this when making purchases, consumers are often eligible for considerable discounts on selected products.
Researchers use a more elaborate version of this type of program in some communities. Here, a number of consumers receive small payments and/or other incentives to sign up to be part of a research panel. They then receive a card that they are asked to present any time they go shopping. Nearly all retailers in the area usually cooperate. It is now possible to track what the consumer bought in all stores and to have a historical record.
The consumer’s shopping record is usually combined with demographic information (e.g., income, educational level of adults in the household, occupations of adults, ages of children, and whether the family owns and rents) and the family’s television watching habits. (Electronic equipment run by firms such as A. C. Nielsen will actually recognize the face of each family member when he or she sits down to watch).
It is now possible to assess the relative impact of a number of factors on the consumer’s choice—e.g.,
- What brand in a given product category was bought during the last, or a series of past, purchase occasions;
- Whether, and if so, how many times a consumer has seen an ad for the brand in question or a competing one;
- Whether the target brand (and/or a competing one) is on sale during the store visit;
- Whether any brand had preferential display space;
- The impact of income and/or family size on purchase patterns; and
- Whether a coupon was used for the purchase and, if so, its value.
A “split cable” technology allows the researchers to randomly select half the panel members in a given community to receive one advertising treatment and the other half another. The selection is truly random since each household, as opposed to neighborhood, is selected to get one treatment or the other. Thus, observed differences should, allowing for sampling error, the be result of advertising exposure since there are no other systematic differences between groups.
Interestingly, it has been found that consumers tend to be more influenced by commercials that they “zap” through while channel surfing even if they only see part of the commercial. This most likely results from the reality that one must pay greater attention while channel surfing than when watching a commercial in order to determine which program is worth watching.
Scanner data is, at the present time, only available for certain grocery item product categories—e.g., food items, beverages, cleaning items, laundry detergent, paper towels, and toilet paper. It is not available for most non-grocery product items. Scanner data analysis is most useful for frequently purchased items (e.g., drinks, food items, snacks, and toilet paper) since a series of purchases in the same product category yield more information with greater precision than would a record of one purchase at one point in time. Even if scanner data were available for electronic products such as printers, computers, and MP3 players, for example, these products would be purchased quite infrequently. A single purchase, then, would not be as effective in effectively distinguishing the effects of different factors—e.g., advertising, shelf space, pricing of the product and competitors, and availability of a coupon—since we have at most one purchase instance during a long period of time during which several of these factors would apply at the same time. In the case of items that are purchased frequently, the consumer has the opportunity to buy a product, buy a competing product, or buy nothing at all depending on the status of the brand of interest and competing brands. In the case of the purchase of an MP3 player, in contrast, there may be promotions associated with several brands going on at the same time, and each may advertise. It may also be that the purchase was motivated by the breakdown of an existing product or dissatisfaction or a desire to add more capabilities.
Physiological measures are occasionally used to examine consumer response. For example, advertisers may want to measure a consumer’s level of arousal during various parts of an advertisement. This can be used to assess possible discomfort on the negative side and level of attention on the positive side.
By attaching a tiny camera to plain eye glasses worn by the subject while watching an advertisement, it is possible to determine where on screen or other ad display the subject focuses at any one time. If the focus remains fixed throughout an ad sequence where the interesting and active part area changes, we can track whether the respondent is following the sequence intended. If he or she is not, he or she is likely either not to be paying as much attention as desired or to be confused by an overly complex sequence. In situations where the subject’s eyes do move, we can assess whether this movement is going in the intended direction.
Mind-reading would clearly not be ethical and is, at the present time, not possible in any event. However, it is possible to measure brain waves by attaching electrodes. These readings will not reveal what the subject actually thinks, but it is possible to distinguish between beta waves—indicating active thought and analysis—and alpha waves, indicating lower levels of attention.
An important feature of physiological measures is that we can often track performance over time. A subject may, for example, be demonstrating good characteristics—such as appropriate level of arousal and eye movement—during some of the ad sequence and not during other parts. This, then, gives some guidance as to which parts of the ad are effective and which ones need to be reworked.
In a variation of direct physiological measures, a subject may be asked, at various points during an advertisement, to indicate his or her level of interest, liking, comfort, and approval by moving a lever or some instrument (much like one would adjust the volume on a radio or MP3 player). Republican strategist used this technique during the impeachment and trial of Bill Clinton in the late 1990s. By watching approval during various phases of a speech by the former President, it was found that viewers tended to respond negatively when he referred to “speaking truthfully” but favorably when the President referred to the issues in controversy as part of his “private life.” The Republican researchers were able to separate average results from Democrats, Independents, and Republicans, effectively looking at different segments to make sure that differences between each did not cancel out effects of the different segments. (For example, if at one point Democrats reacted positively and Republicans responded negatively with the same intensity, the average result of apparent indifference would have been very misleading).
Research sequence. In general, if more than one type of research is to be used, the more flexible and less precise method—such as focus groups and/or individual interviews—should generally be used before the less flexible but more precise methods (e.g., surveys and scanner data) are used. Focus groups and interviews are flexible and allow the researcher to follow up on interesting issues raised by participants who can be probed. However, because the sample sizes are small and because participants in a focus group are influenced by each other, few data points are collected. If we run five focus groups with eight people each, for example, we would have a total of forty responses. Even if we assume that these are independent, a sample size of forty would give very imprecise results. We might conclude, for example, that somewhere between 5% and 40% of the target market would be interested in the product we have to offer. This is usually no more precise than what we already reasonably new. Questionnaires, in contrast, are highly inflexible. It is not possible to ask follow-up questions. Therefore, we can use our insights from focus groups and interviews to develop questionnaires that contain specific questions that can be asked to a larger number of people. There will still be some sampling error, but with a sample size of 1,000+ responses, we may be able to narrow the 95% confidence interval for the percentage of the target market that is seriously interested in our product to, say, 17-21%, a range that is much more meaningful.
Cautions. Some cautions should be heeded in marketing research. First, in general, research should only be commissioned when it is worth the cost. Thus, research should normally be useful in making specific decisions (what size should the product be? Should the product be launched? Should we charge $1.75 or $2.25?)
Secondly, marketing research can be, and often is, abused. Managers frequently have their own “agendas” (e.g., they either would like a product to be launched or would prefer that it not be launched so that the firm will have more resources left over to tackle their favorite products). Often, a way to get your way is to demonstrate through “objective” research that your opinions make economic sense. One example of misleading research, which was reported nationwide in the media, involved the case of “The Pentagon Declares War on Rush Limbaugh.” The Pentagon, within a year of the election of Democrat Bill Clinton, reported that only 4.2% of soldiers listening to the Armed Forces Network wanted to hear Rush Limbaugh. However, although this finding was reported without question in the media, it was later found that the conclusion was based on the question “What single thing can we do to improve programming?” If you did not write in something like “Carry Rush Limbaugh,” you were counted as not wanting to hear him.