You need a creative approach
Editor's note: The authors are employees in Sensory Evaluation at Avon Products, Inc. Maisie Wong is senior research analyst, Ayn Gelinas is manager, Sensory Evaluation, and Phyllis Rocha is a category manager, fragrance.
Penny, an interviewer for a data collection agency, approached an Hispanic-looking couple at the mall and asked the woman, in English, if she would like to participate in a marketing research survey. The couple quickly walked away. A group of teenage girls overheard Penny and asked her if they could participate. They were Hispanic, they spoke English, and they would like to answer Penny's questions. Penny was ecstatic! She had filled a good portion of her quota with these helpful girls!
What's wrong with this picture?
The answer -as we encountered it- is: The "normal" rules for conducting a marketing research survey do not apply when the target sample is an Hispanic population.
Fortunately, we discovered this in the planning stages of our sensory evaluation formulation guidance study. As we researched the topic of testing among Hispanics, we found that most of our experience in testing among Avon's customary target population simply did not apply. We acknowledged that fact and accepted the challenge of starting from ground zero. During each phase of the study, we encountered idiosyncrasies inherent to the Hispanic population. Further research gave us a better understanding of the fragmented Hispanic cultures. Through careful planning and a pilot study conducted prior to implementation of the actual study, we gained invaluable insights into ways to overcome some of the cultural barriers and ensure that the research was of the highest quality. We'd like to share our experiences and relay some anecdotes that may be helpful to researchers working with this population.
Recruiting
While the number of Spanish/English bilinguals is growing, the Hispanic Policy Development Projects of New York City and Washington, D.C. report that 6,800,000 Hispanics speak only Spanish or are bilingual and prefer Spanish. This equates to approximately one-third of the Hispanic population. Therefore, bilingual interviewers are a must!
The interviewer should speak to the respondent in the language with which the respondent is most comfortable. In our study, to determine the potential respondent's language preference, the interviewer approached the respondent and asked her, in Spanish, if she would like the interviewer to speak to her in Spanish or in English. By speaking to potential respondents in Spanish first, we had a higher chance of reaching a larger proportion of this population on the first attempt. After the potential respondent replied, the interviewer continued the screening in the respondent's language of choice. This also determined the language of all subsequent printed material-instructions, questionnaires, reminder sheets-received by the respondent.
To be successful at this stage, we found that we had to go to the respondents, wherever that might be. Malls are commonly used sites for intercept screening; however, even carefully selected malls may not generate enough Hispanic traffic to fulfill the requirements of a study. This occurred at one of the sites selected for our study. The traffic in the mall was slow, and the number of Hispanic women passing through was low.
Given these circumstances, we considered alternatives within the local Hispanic communities. We discovered a laundromat and beauty salon located on the neighborhood's main street, which, in this instance, proved to be entirely appropriate and fertile grounds for the recruiting of female Hispanic respondents. We asked the owner of the beauty salon to participate in the study. She was not a qualified respondent, and we later discarded her data. However, once she was engaged in the process, her approval was tremendously helpful, and we successfully recruited her qualifying patrons to participate.
While soliciting bids for our study, we were warned that pre-recruiting tends to yield a large number of no-shows. The research services told us that an over-recruitment of as much as 40-50% may be necessary to yield the intended number of completed responses. Intercepts appear to be more successful, since a stage where attrition may occur is eliminated. If respondents are recruited by intercept, an over-recruitment of 20-25% may be sufficient to ensure a completed sample of sufficient size. We were able to generate the necessary number of completions with an over-recruitment of approximately 20%. Our high level of success was, in part, attributable to the diligence of the interviewers and to the nature of the study, which provided an inherently pleasurable sensory experience.
Further, as an enticement to full participation and completion, the study's incentives were packaged in bright, multicolored miniature shopping bags overflowing with tissue paper and ribbons, and displayed on a counter in the salon in full view of qualifying respondents. Although the women did not know the specific contents of the bags, their appearance was attractive and evoked a high level of perceived worth.
U.S. Census Bureau data show that approximately 65% of Hispanic females are married. Traditionally, the male is the head of the household and does not condone his wife's working outside the home. We found this to be true at one of our selected geographic locations, where this tradition holds strong. The female generally depends upon her husband for transportation; he accompanies her wherever she needs to go. Interestingly, we found that in these instances, when approaching a couple, it is appropriate to first ask the husband's permission for his wife to participate in the survey. Once he consents, the interviewer may speak to the wife.
We also found that the daily routines of non-working females are very different from those of working females. The majority of married Hispanic women do not work. They tend to rise and go to bed later than working women. Therefore, the most productive time to screen and recruit respondents is after 10 a.m.
Screening and interviewing
Our study required that the respondents interact with the product under normal use conditions at several timepoints during the day. Under similar circumstances, with a more general U.S. sample, we'd found that a self-administered questionnaire worked well.
When working with this specific population, we found it necessary to screen the potential respondent for literacy in either Spanish or English. We accomplished this by simply asking the potential respondent to reply to an early screening question by reading her response from a card pre-printed in the appropriate language.
We invested a greater than usual amount of time in explaining the logistics of the study to these women. The bilingual interviewers really engaged the respondents, and convinced them that we valued their personal opinions highly, while going over the questionnaire, point by point, with each of them. We found that once these women really understood what was expected of them, they became engaged in the process, and were usually committed to the study.
Study design
Central location tests are generally more successful in the achievement of completed quotas than home use tests, because one has a captive audience. In addition, interviewers are on hand if questions arise. The nature of the research, however, often dictates the appropriate technique to be used, as it did in our case.
In designing this study, we gave consideration to the fact that a number of Hispanics do not have easy access to a telephone. Alternatives to phone interviews should be explored if this subgroup is part of your target market. Hispanic women may regard several telephone callbacks as a nuisance; the calls may anger them to the point where completion of the study is jeopardized.
Our study spanned a length of time, with product evaluations throughout. Further, it was important that the evaluations be completed at specified timepoints. In the past, when we have conducted similar studies with a more general population, the respondents were asked to stay at a location where they could be reached by phone during the day. The interviewers called the respondents near their specified evaluation times, and reminded the respondents to evaluate the product and complete the appropriate questionnaire.
For this particular study, we had to explore alternatives to the "reminder" calls. We created a reminder sheet that was clear and simple. It contained the pertinent information and was handwritten in large, easy-to-read block print. We asked the respondents to use the sheets as reminders and to post them in a place where they could be easily seen during the day.
In part, central location tests work well because the questionnaires are administered by the interviewer. In the case of home use tests, phone interviews or self-administered questionnaires must suffice. In the specific case of our study, a self administered questionnaire worked well because 1) the respondent was told exactly how to fill it out; 2) the study was short; 3) the questionnaire was kept simple; 4) receipt of the incentive hinged upon the respondent returning the questionnaire in-person; and 5) the interviewer edited the questionnaire before giving the respondent the incentive, to ensure that the questionnaire was correctly and completely filled out.
Questionnaire construction
As with the design of any other questionnaire, the key point is to keep it simple! The length of the questionnaire itself is not the most important factor as long as, 1) the questionnaire is laid out in an easy-to-follow format; 2) the directions are clear and easy to follow; and 3) the questions are easy to understand and require little effort on the respondent's part to complete.
The language
The translation stage required several rounds of discussion and revisions before we were satisfied with the questionnaire. Even so, it proved not to be perfect.
We found that translating the questionnaire from English to Spanish was very tricky. It was imperative that the two versions be equivalent; nuances of the words used in one language had to carry over to the other language. For instance, one of the descriptive terms used in the English version of the questionnaire was "sexy," which was best translated as "sensual" in Spanish, since this has a different meaning than "sensual" in English. Despite our best efforts, the occasional gaffe slipped through. In our pilot study, we had translated eau de toilette as "aqua de tocador" for a product usage query in the screening questionnaire. The usage response to "aqua de tocador" was suspiciously low. We were chagrined to learn that "aqua detocador" literally means "water from the bowl"! This episode taught us that foreign words or phrases that are universal in all languages should not be translated. Other words, such as "hello" or "bye-bye" are universally understood by people, regardless of their ethnic background, and also may not require translation.
We would recommend that colloquial, rather than formal Spanish be used. However, it is important not to use colloquial expressions that are specific to a sub-group of the Hispanic population. For example, a phrase used by Puerto Ricans may be incomprehensible or misunderstood by a Hispanic of Mexican origin. Keep in mind that some words in the English language may not exist in Spanish, or vice versa.
The scales
Again, simplicity is the key! The Hispanic respondents with which we tested had limited education and had difficulty with the concept of gradation. For example, we found that they either liked or disliked something; it was difficult for them to quantify the degree of their liking or disliking.
We have also found this to be true in our research with female respondents in our Mexico subsidiary. Our questionnaires for this type of study, when conducted with our customary population, usually contain a seven-point hedonic scale. Early on in our work with our Mexican counterparts, upon their return of the completed questionnaires in Spanish, we found that they had collapsed our seven point scale to three points-"like," "neither like nor dislike," and "dislike." Their concern was that their testing population would not be comfortable with a seven-point scale. A flurry of telexes later, we had convinced them of the need for the expanded scale.
Whether word scales or pictorial scales are used, the interviewer must explain the use of the scale to the respondent and ensure that she understands it. As an extension of this, we noted that Hispanic respondents, as a group, tend to rate higher or use the upper end of the scale more readily. We speculate that this could be due to their desire not to offend the interviewer. It could also reflect that the product category we were testing is designed to provide a pleasurable experience; thus, we got "good," "gooder," and "goodest"! This tendency must be taken into consideration during data analysis and interpretation.
Incentives
Large cash incentives are sometimes used in an effort to bolster returns. We've found that over-compensation has its repercussions. It sets a precedent that, given the vagaries of research budget funding, may be difficult to meet the next time around.
Cash may be generally preferred. However, in this particular case, we discovered a way to get around using cash as the incentive. We gave an assortment of beauty products manufactured by a leading beauty company, and worth more than $50. This approach had several advantages. Since the products were manufactured by our company, we "bought" them at cost, which substantially lowered our out-of-pocket expenses. Further, the perceived value of the incentive was greater than the monetary equivalent. Combine that with the fact that our company was receiving free advertising and sampling of major products, and you have an incentive strategy that can't be beat. Since the respondents did not know the identity of the manufacturer until after completed questionnaires were resumed, the incentives did not present a study bias, nor did they breach the security of this research guidance study.
Summary
Our experience has led us to several conclusions regarding what it takes to conduct successful research among Hispanics.
The Hispanic population represents a multitude of segments. Each differs in its cultural and historic background, as well as in its attitudes and beliefs. It should not be regarded as homogeneous, and segmenting it should not be oversimplified.
The largest requirement for conducting research among this population will be your investment of time. Never underestimate it. It will directly affect the quality of your questionnaire and the clarity of your questions.
Be aware of cultural implications, as well as the implications of the Spanish language. Both will help you in understanding the complexity of this population. Successful execution of your study requires that you accommodate to the habits and practices of the Hispanic people.
Lastly, we highly recommend that you run a pilot study, to refine the test procedure and questionnaire, before initiating the full-scale study.