Marketing research process: Six stages

Page | 1

Business Outlook April 2015 (Volume 13, Issue 4)

Marketing Research Process: Six Stages1

Dr. Michael R. Hyman, NMSU

Dr. Jeremy J. Sierra, Texas State University

Correctly conducted marketing research requires careful attention

to intricacies. Think of marketing research as a chain only as strong

as its weakest link, where those links are stages in a process.

A professional baseball player can slump for several weeks during

the season yet win the batting title if he hits .500 for a month.

Because their regular season lasts six months, players can

overcome a slow start and ‘play to the back of their baseball card’

by season’s end. Similarly, students who perform poorly on a first

exam still can earn a good grade by performing exceptionally on

subsequent exams. Baseball and school accomplishments are

compensatory; it is possible to recover from mistakes. Not so with

marketing research.

Once you have failed to identify the correct marketing research

problem, your subsequent research efforts are wasted. If you do not

know what you need to know, then you cannot uncover the answer

to your research problem. An improperly designed and fielded study

cannot provide trustworthy data for subsequent analyses. Faulty

data analysis is meaningless at best. An inability to communicate

study results clearly to decision makers—for example, possible

investors in your new business venture—greatly increases the

likelihood of a poor decision.

Research is a multi-stage and often somewhat iterative process—

conclusions from one stage can create new ideas for other stages

in the process, and the linkages are both forwards and backwards.

Also, stages can occur concurrently.

1 Note: John Wiley & Sons, Inc. published Mike and Jeremy’s Marketing

Research Kit for Dummies. It is available in paperback [ISBN: 978-0-470-

52068-0] and Kindle [ASIM: B003CNQ4LG] versions. The following text is

based on Chapter 2 of that book.

Page | 2

Stage 1: Identifying the Problem

If you cannot define a problem properly, it is impossible to find an

appropriate solution. Unfortunately, a problem may not be obvious,

often because its cause is obscured. Hence, the ‘Iceberg Principle’

comes into play: The dangerous parts of many marketing problems

may be obscured because they are below the surface. Your job

(and that of a marketing researcher, if involved) is to identify the

appropriate problem despite its partial or total submersion.

Defining problems is a six-step process:

1. Ascertain your objectives. You and your business assoc-

iates—the loan officer at your friendly neighborhood bank,

for example—may have different yet equally reasonable

objectives. For example, you may be more interested in

growing your business, whereas your associates may want

immediate increases in sales or profits. Clearly, the goals

and types of marketing research projects may vary based

on whether a short-term or a long-term increase is sought.

Whatever the objectives, you must express them in

measureable terms. Otherwise, you cannot assess whether

the results are favorable. For example, the objective

‘improve customer attitudes’ is noble but fuzzy; in contrast,

the objective ‘increase profits by 20% in the next 24 months’

is definitive and easily monitored.

2. Understand the problem background. To avoid—or at least

minimize—that ‘iceberg’ problem we noted, you must step

back and gain perspective. An informal gathering of

background information about the environment in which

your business operates should help in that regard.

3. Isolate/identify the problem, not the symptoms. Symptoms

can be confusing; you may be so caught up in them that you

do not recognize the disease! Good marketing research can

help you structure and understand the true problem.

Consider the following example. A new cell phone with basic

computing and Internet capabilities is selling poorly.

Distributors claim competitors' lower prices for similar

products are causing poor sales. Based on the distributors'

beliefs, the company conducts a detailed analysis of

competitors' products, attending especially to pricing. In

fact, the analysis reveals the true problem is the distributors'

lack of product knowledge and concomitant inability to

Page | 3

explain the product's value to potential customers.

4. Determine the unit of analysis. Depending on the research

problem, the appropriate unit of analysis could be persons,

households, spouses, or organizations. Without identifying

the appropriate unit of analysis, you cannot draw a suitable

sample or perform suitable data analyses. In consumption

studies, for example, households rather than persons are

the appropriate unit of analysis. To understand major

purchases—an automobile or home—an examination of

spouses’ decision-making processes is critical. Marketers

who do not understand those processes are flying blind in

their efforts to provide the best possible product.

5. Determine relevant things to ask about. Although you may

want to learn about non-quantifiable matters, relevant

issues typically are quantifiable. In essence, this step of the

process entails determining what to measure and how to

measure it.

Your dependent and independent variables determine the

focus of your study, especially in forecasting contexts. As

the words denote, the dependent variable depends on one

or more independent variables. If you want to forecast next

month’s sales (the dependent variable), then you want to

identify factors that predict those sales accurately (the

independent variables). For example, realtors can predict

home-buying behavior during the next quarter or the next

year by looking at factors relating to future home-buying

behavior, such as growth in disposable income, growth in

investment income, and consumer sentiment.

6. Translate the marketing problem into researchable

objectives. Because researchers must create researchable

objectives concordant with their problem definition, often

they want to express those objectives in the most rigorous

terms—something called a hypothesis. A hypothesis is a

formal, testable statement refutable by empirical data.

Whereas you may have a hunch about your customers, a

hypothesis about your customers is a formal statement of

that hunch testable by marketing research. To generate one

or more hypotheses for formal testing, start with a purpose,

which helps generate research questions answerable by

exploratory research, your experience, and basic marketing

theory.

Exploratory research often is a necessary prelude to developing

hypotheses. Perhaps you do not understand the underlying

Page | 4

process sufficiently to develop a formalized, testable statement,

in which case exploratory research is a preliminary step. Types

of exploratory research include reviewing secondary data,

conducting pilot studies, doing in-depth interviews with people

who have requisite experience, and implementing case studies.

Turning Problems into Objectives: Examples

Marketing problem #1: Should a brick-and-mortar women's

clothing retailer create an Internet shopping site?

Possible research question: Are current and potential new

customers comfortable shopping online for women's

apparel?

Research objective: Assess current and potential new

customers' online shopping attitudes and behaviors for

women's apparel.

Marketing problem #2: Which group of potential customers

should this clothing retailer target?

Possible research question: Which group of potential

customers spends the most on apparel each year?

Research objective: Assess previous purchase behavior and

purchase intentions of different groups of potential customers.

Stage 2: Designing the Study

The research design is the master plan for the research that follows.

This stage specifies the methods you or the researcher will use to

conduct the study.

Here are several basic questions about research design:

What types of questions need answering? You must decide

the questions needing answers and whether those answers

can be provided by some combination of surveys, experi-

ments, or analyses of secondary data. If you are uncertain

about those questions, preliminary exploratory research

may be necessary.

What is the data source? If you conduct a survey, then your

initial design issues relate to your questionnaire and data

collection method, which are intertwined. For example,

complex questions and questionnaire structures are ill-

Page | 5

advised for surveys administered via telephone.

If secondary data are needed—for example, to conduct a

site location analysis—how timely and compatible are

existing sources? In essence, are the available data a

square peg you are trying to stuff into the round hole of your

research needs?

If you conduct an experiment, then you must ensure a

proper design with tight controls; otherwise, you cannot

know if your results are bogus or reflect the conditions to

which you exposed participants.

Can you get objective answers by asking people? Often,

people are unaware of their reasons for doing things or are

incapable of responding meaningfully to questions about

their attitudes and behaviors. When these issues arise,

asking people directly will not work. Alternatively, you may

use observation to answer your research questions

indirectly.

How quickly is information needed? You must decide how

quickly your research study must be completed. Marketing

research can be relatively accurate, relatively fast, and

relatively inexpensive, but it can only be two of those three

simultaneously. If you needed to know yesterday, then the

expense for a study of sufficient quality increases markedly.

How should survey questions be worded? Wording survey

questions so answers accurately reflect people’s attitudes

and behaviors is both an art and a science.

How many questions can you expect to ask respondents?

Respondents’ patience is finite, especially when you phone

them at home or intercept them at the mall. Thus, the survey

data collection method you choose depends on the number

of questions you need to ask.

Are descriptive findings sufficient, or will an experiment be

necessary? Surveys are helpful for assessing people’s

attitudes and preferences for current products, and some-

what useful for self-reports about previous consumption, but

not especially good for predicting people’s reactions to new

products. For example, a survey about features for ebook

readers administered to people who have never used such

devices is unlikely to produce accurate forecasts of future

reader purchases. An experiment, in which different people

Page | 6

use different reader devices with different features, might

provide far more predictive data.

If an experiment, what will be tested? If an experiment is

needed, what treatment or condition will the researcher

test? In what circumstance will you place one group of

people and how will you compare their responses to the

responses of a different group of people placed in a different

circumstance?

For example, if you want to identify the most effective among

several print ads you might run in a local newspaper, how

will you expose people to those ads? You want people to

respond naturally to these ads, yet to show them only the

ads and then ask them what they think is an artificial task

likely to produce untrustworthy results.

Stage 3: Selecting a Sample

If you have only 25 customers to whom you might offer a new

service, you can afford to survey all of them. However, if you have

100,000 customers, surveying all of them is neither cost effective

nor necessary. Instead, you can select a representative sample to

ask about this possible new service.

Here are several basic questions about sample selection:

Is a sample necessary? If the population is small and

reasonably accessible, then you can query every person in

the population, in which case you are taking a census rather

than drawing a sample.

Who or what is the data source? Is/are the group(s) of

interest—your sampling unit—individual consumers, house-

holds, or organizations? A scientific sample—one you can

comfortably generalize to the group(s) you want to query—

requires drawing respondents from a representative list (or

sample frame). Such lists are available from commercial

suppliers, but you will need to identify the supplier and the

characteristics of your respondent pool. Regardless, the

next step is to identify the unit of analysis and a sample

whose constituents are consistent with that unit of analysis.

Can the target population be identified? Typically, there is

no single correct population to sample; sampling from any

of several alternative populations is acceptable. Suppose

you want to explore consumer preferences for a reformu-

Page | 7

lated soft drink. You could sample from any population of

heavy soft drink consumers, such as high school students,

college students, or young professionals. In this case,

convenience and cost should dictate your population choice.

Perhaps your target population is ill-defined; for example,

the population of potential new customers for your store (if

you are a retailer). Even if you assume these customers are

similar to your ‘current customers’, it is silly to ask ‘current

customers’ what would cause them to switch to your store.

Instead, you could pay a commercial supplier for a list of

people with demographics similar to your ‘current

customers’ and then disqualify ‘current customers’ through

a filter question on your questionnaire. Alternatively, if you

also wish to survey ‘current customers’, then you would use

those filter question data to sort respondents into receivers

of your ‘current customer’ questionnaire and receivers of

your ‘potential new customer’ questionnaire.

How accurate a sample is needed? Many questions about

sample size relate to accuracy and the way in which the

sample is drawn from a larger population. For many com-

mercial studies, researchers use commercially available

lists, yet those lists may be deeply flawed. For example, one

established commercial list of people who had recently

moved included people who had not moved in 57 years!

Is a probability sample necessary? Researchers may need

to assess whether a probability/scientific sample is required.

For some research purposes, convenience sampling—a

type of non-probability sampling—is much less costly and

may be appropriate.

Is a local sample sufficient? The need for only a local or

regional sample rather than a national sample may affect

the methodology you choose. For example, if your research

requires a national or international sample and you are

concerned about cost, you probably will opt to collect

respondent data via some social media community. For a

local sample, the phone or other data collection technology

may suffice.

How large a sample is necessary? Knowing the scope of the

sample is useful, if only to keep data collection costs within

budget. Because the cost of data collection is a large share

of total study cost, staying within budget becomes

impossible once data collection costs soar.

Page | 8

Stage 4: Gathering the Data

When conducting research, arguably the most important stage is

the data collection stage. Research questions cannot be answered,

consumer needs cannot be met, and your business cannot benefit

from pertinent findings without data.

Here are several basic questions about data gathering:

Who will collect the data? If it is an independent field

service, then you likely prefer some control over the way

interviewers query respondents, whether by phone or in-

person.

How long will data collection take? You must decide on a

time horizon for completing your study, as that horizon may

dictate many aspects of your study, such as the data

collection method, sample size, extensiveness of pre-

liminary qualitative and secondary research, and so forth.

How much supervision is needed? Such supervision

depends on the data collection method. For example, phone

surveys often are fielded by data collection services with

supervisors who monitor—some might say eavesdrop on—

many calls placed from extensive phone banks. Supervising

such fieldwork is straightforward. For personal interviews,

immediate oversight is impossible, so the type of super-

vision differs markedly. Supervising personal interviews

typically entails verifying at least 10-15% of interviews were

conducted as indicated by field service workers.

Stage 5: Analyzing the Results

Data without analysis is rubbish. Although universities offer

statistics courses where hand tabulation is required, the

available software trumps the use of such archaic computation

methods. Because good business decisions depend on

trustworthy empirical analyses, you should learn to use the

software needed to perform such analyses. Although advanced

statistical analysis software—such as Statistical Package for

the Social Sciences (SPSS) and Statistical Analysis Systems

(SAS)—is available, you can run many analyses on Excel.

Here are basic questions about data processing and analysis:

Will standardized editing and coding procedures be used?

How will data be edited and coded? (For survey research,

Page | 9

editing means clearing the data of impossible and incon-

sistent responses, and coding means creating a database in

which numbers rather than words represent respondents’

answers.) Expertise is required to edit and code open-ended

questions. Alternatively, if you are creating a database of

responses to close-ended questions—the type of questions

scaled 1 to 6 or 1 to 7—then minimal expertise is required.

How will the data be categorized? The ability to analyze data

depends on the way they are grouped. Depending on the

type of data, the way they are categorized enables certain

types of statistical analyses. For example, nominal data like

‘sex’ or ‘ethnicity’ enables descriptive statistics like total

number and percentage in each group. Interval data—for

example, attitudes measured on a 1-to-7 scale—can be

grouped by their relationship to a calculated mean score.

Categorization in this context is more than a statistical

notion; it has practical implications. Perhaps you want to

compare current customers to non-customers. Alternatively,

you may want to compare frequent customers, infrequent

customers, and non-customers. Assessing differences

between groups of current customers may provide useful

marketing insights.

What data analysis software will be used? Commercial

packages like SPSS and SAS enable almost any type of

statistical analyses. Prices for these packages, depending

on configuration, range from $500 to $2,000. Although they

are ideal for seasoned marketing researchers, you can run

worthwhile analyses with Excel for a far lesser price. For

example, you can run standard descriptive statistics, cross-

tabulations, correlations, and difference tests in Excel.

What is the nature of the data? If the data are qualitative,

you are looking at people’s open-ended and rambling

responses to questions. If the data are quantitative, you are

looking at close-ended data, which are far easier and more

straightforward to analyze.

Stage 6: Communicating the Findings

and Their Implications

The value of a research study is only as good as the weakest link

in the process chain. Even the best conceived and conducted study

is useless if its results are not presented meaningfully.

Page | 10

Understanding your audience is a good idea if you are required to

present your study. The following are some basic questions to ask

yourself about the type of report you should create:

Who will read the report? Readership is critical because it

determines the amount of technical expertise. Marketing

jargon and statistical analyses decipherable to a venture

capitalist may be meaningless to your banker. Many

audiences prefer well-constructed graphical displays to

detailed tables and extensive exposition.

Do you want/need managerial recommendations? If specific

recommendations are required, then they should be

included and justified in your report. If you merely are

providing information others will use to draw their own

conclusions, then providing recommendations is needless.

Will presentations be required? If so, how many presenta-

tions and to whom (for example, possible lenders or

franchisees)? If presentations are required, you should

ascertain the audience for and number of those

presentations as part of the budgeting process.

What format will the written report take? The degree to

which a written report should be formal or informal may

depend on corporate culture and the need to please the

people who will read it.

The more you understand such preferences and constraints, the

more likely your report will achieve its intended goals.

Anticipating Outcomes

Creating dummy tables—blank tables to be completed once data

analysis results are available—helps guarantee the most useable

report is created. By providing dummy tables to report readers, you

give them an opportunity to provide feedback about how helpful that

set of tables, once completed, would be to their decision-making

process.

Before beginning a study, consider these checklist-type questions

to confirm the wisdom of conducting it:

How much will the study cost? You must confirm the study’s

cost because it is critical to assessing its value. If a study

costs more than the value of your reduced uncertainty about

the best course of action, then you should not conduct it.

Page | 11

Is the time frame acceptable? Studies inherently require

different times to complete, so you need to confirm the time

frame for completion is acceptable.

Is outside help needed? If outside help is needed, you need

to identify that outside help and make certain it is available.

Will the research attain your stated objectives? You must

confirm your research plan addresses your research

objectives. To do so, return to your research problem and

check your objectives are consistent with it. Ensure your

objectives, if attained, are actionable; leave research-for-

research-sake to academicians.

When should research begin? Given budgetary and other

concerns, you must confirm a starting date for the research

commensurate with decision-making deadlines.

About the Authors

Dr. Michael R. Hyman is Distinguished Achievement Professor and

Ph.D. Coordinator of Marketing at NMSU. He is Executive Editor of

NMSU Business Outlook and Marketing Ethics Section Editor for

Journal of Business Ethics. Attesting to his writing compulsion are

80 academic journal articles, 45 conference papers (10 which won

a ‘best paper’ award), four co-authored/co-edited books, 25 other

academic contributions, and 30 non-academic works. He is known

for his collection of Looney Tunes shirts, inability to chip a golf ball

correctly, encyclopedic knowledge of classic Hollywood movies,

overly neat office, and loyalty to the New York Yankees.

An avid golfer and ardent Nebraska football fan, Dr. Jeremy J.

Sierra is Associate Professor of Marketing at Texas State

University. He serves on six editorial review boards, including

Journal of Marketing Theory and Practice, Journal of Services

Marketing, and Psychology & Marketing. He has published his

scholarly work in such outlets as International Journal of Market

Research, Journal of Advertising, Journal of Business Research,

Journal of Marketing Theory and Practice, Journal of Product &

Brand Management, Journal of Services Marketing, and

Psychology & Marketing. Dr. Sierra also has published, with

Michael R. Hyman (his mentor and long-lost golfing buddy),

Marketing Research Kit for Dummies. His research interests

include advertising effects, brand tribalism, dual-process theory,

and superstitious beliefs.