"I don't Know How to Create a Customer Survey."
Ideas to Consider:
look at: Using Focus Groups
an Opinion Research Survey to Increase Sales
By Gary Witt, Ph.D.
In another article we looked at qualitative research and focus
groups. Now we'll turn to quantitative research and surveys, the
most popular method of quantitative consumer research.
We covered the importance of having a clear, specific purpose for
your research, and how to build a list of top questions to ask. A
survey allows you to ask a large number of consumers these questions,
and turn their answers into numbers which can be manipulated and graphed.
There are four major ways to contact those being surveyed: telephone,
personal (like door-to-door or mall intercept), mailed, and Internet.
Telephone, mail and e-mail give you the most control over the
characteristics of those who are surveyed. The Internet (survey on
your Web site) and e-mail are the cheapest to do, but those who
answer may not be representative of your target population.
Selecting an unbiased sample is critical if you want to
extrapolate the results to all your consumers. For example, if
you wanted to know if buyers of your cookies would pay another
quarter for them, you wouldn't just send your researchers to malls in
affluent areas. The people surveyed would likely have a much higher
income than your overall buyers. So when the survey results showed
people would gladly pay an extra quarter, that may not be true of
your overall buyers, just the most affluent of them. A pricing
decision made from this biased sample could be disastrous.
The two keys to a good sample are to (1) accurately define the
population of customers you are interested in, and (2) select enough
random members of that population. If you wanted to find out your
buyers' attitudes about the features of a new type of expensive
laptop before planning its marketing campaign, your target population
would likely be computer users who are affluent. You might find that
population by buying a mailing list from a computer magazine,
specifying just those zip codes near your store which have the
highest household incomes.
If the list is large, you will only need to sample part of it.
Statistical analysis formulas show that even for a population of
10,000,000 people, you only need to accurately survey 384 people to
get results that have a 95% chance of being accurate for your entire
population. If your population is 10,000, you'll need to survey 378
people. If its 5,000, survey 357.
How do you pick those to survey? The most accurate way is to assign
each person on your list a number, then use a Random Number Table (at
your library), a "fishbowl" drawing, or any other method in
which (a) every person has an equal chance of being picked, and (b)
the selection of any one person does not affect the chance for anyone
else to be picked.
Now you've got your sample of about 360 people. (By the way, if you
think most of them are very similar, statistical laws say you only
need to survey 240 people out of 10,000.) What's the best method to
find out what you want to know, to find information you can really use?
Your decision about the method of sampling will probably be
influenced by your budget and your timetable. Mail is low cost, but
slow, and often doesn't have a good response rate without incentives.
Telephone surveys have a moderate cost, and they are fast. Personal
(mall) interviews are expensive and relatively slow (it takes a while
to stop 360 qualified shoppers), but the quality of answers may be
superior. Of course, there are other ways to elicit quantitative
information besides surveys, such as experiments and observation, but
we'll stick with this most popular approach.
Your survey questions can be "open-ended" (like fill in the
blank, or "tell me what you think"), or "closed-ended"
(like a multiple choice test.) The open-ended questions will give
you richer data, but they'll be hard to turn into numbers and are
more time-consuming to score. That's why most surveys either provide
the answers or furnish the scale on which to rate specific responses.
They're easy to administer, quick to score, and easy to understand.
But they do reduce the richness and variability of the responses.
Some surveys use a combination of both types of questions.
Good questions are difficult to write, but seem easy. Why?
Because most people will create questions that have a hidden bias
which can skew the results. For example, "Do you prefer the old
or new package for Hamm's Hotdogs?" will skew answers toward the
latter answer, since many people are conditioned to like
"new" over "old" for nearly every product. For
detailed advice on how to do write good questions, consult a book or
articles on consumer research. One good book is How to Conduct
Your Own Survey by Priscilla Salant.
The major problems with questions center around unintended bias and
clarity. For example, a question which presented more positive than
negative choices would show an unintended bias. A question which used
unfamiliar abbreviations or vague words may be unclear. Finally, be
careful about creating "double-barreled" questions, ones
that are really two questions linked at the hip, like "Do you
think green is an attractive color for Spring fashions, but not for Summer?"
Here are the three most common types of survey questions and their uses:
(1) Likert Scale questions: These questions give the consumer a five
or seven step set of choices, like "Do you Strongly Agree,
Agree, Neither Agree or disagree, Disagree or Strongly Disagree with
the following statements? (a) I like to shop for clothes. . ."
Use the Likert Scale if you want to find out how much consumers agree
or disagree with various statements about your product, service,
competition, price, or anything else. The answers help reveal the
judgments consumers make. The next scale helps reveal the emotions
which influence those decisions.
(2) Semantic Differential Scale questions: These questions ask the
consumer to evaluate a product on a series of scales ("Please
check the blank closest to your feeling or opinion about Starbuck's
Coffee on the following scales.") Each scale is comprised of a
set of two adjectives with opposite meanings (hot/cold, like/dislike,
expensive/inexpensive), separated by five or seven blanks, like this:
HOT ___ ___ ___ ___ ___ COLD. Each blank is given a numeric value in
scoring, so the overall average result may be plotted exactly on the scale.
Here's another example: "How do you rate the Pet Store on each
of these scales?
Expensive __ __ __ __ __ Inexpensive
Warm Feel __ __ __ __ __ Cold Feel
Exciting __ __ __ __ __ Not Exciting
Good Selection __ __ __ __ __ Poor Selection
Helpful Clerks __ __ __ __ __ Unhelpful Clerks," and so on.
Notice that the choices are not biased (we use "expensive/inexpensive,"
not "expensive/cheap"), since that would bias the results
(some people would hesitate to admit they shop in a "cheap"
store.). Using common opposites (good/poor) and common semantic tools
(helpful/UNhelpful, courteous/DIScourteous) are good ways to keep
The Semantic Differential Scale is useful when you want to know in
more detail how consumers see your product. Taken together, these
beliefs play an important role in the consumer's overall judgment
about the product or store, and likelihood of shopping there. Of
course, consumers may rate you high on scales which aren't important
to them ("attractive decor"), and low on ones which are
("price," "quality"), or vice versa. To
understand better how to evaluate these results, you could include
the third kind of survey question.
(3) Rank-Order Scale questions: Subjects are asked to rank items in
order of preference, enjoyment, importance, value, or some other
criterion. Each item is listed, with a blank next to it. For example,
"Rank in order of importance these four attributes of a pet store:
__ Feeling I get when I enter
__ Price level of merchandise
__ Knowledge of Clerks about the merchandise."
Rank ordering is often used to find out where your product, or
certain features of your product stand in relation to your
competition. It shows you where you are weak, and it can present some
good positioning opportunities when you find an important feature on
which your competition ranks poorly.
Finally, include some demographic questions which are important to
you, such as the person's age, sex, income level, race, Internet
usage, and so on. These questions are usually found at the end of a
survey. Don't go overboard. People are reluctant to answer many
personal questions. Respectfully ask only what you need to know about
One advantage of demographic questions is the opportunity to run
"cross-tabulations" to show relationships between answers
on some of the survey questions and certain demographic
characteristics. For example, you might find that men are most
interested in the power of your new laptop, while women focus on its
weight. "Cross-tabs" are very useful to have when you're
planning targeted advertising.
Quantitative research results are turned into numbers, most often
averages. The results of each question can also be graphed to give a
visual illustration of the results. Graphs also give you a way to
longitudinally compare previous results with current results, showing
trends. Your first survey will give you a "baseline" to
which future answers can be compared. Knowing these trends can give
you the edge you need over your competition.
Even if you are mathematically-challenged, it is important for you to
know how to determine a few types of basic results. The Average or
Mean is found by adding up all the individual scores and dividing by
the number of scores. If you have a few outlandish scores which seem
to be skewing your Mean, look at the Median. Find the Median by
putting all the scores in order from top to bottom. Count the number
of scores and divide by 2. Count down the scores to that number,
which is in the exact middle of the list. That's the Median.
Finally, you should know the Standard Deviation (SD) around the Mean.
That is the average amount of variance around the Mean. For example,
if you asked people to guess the price of your ($25/room) carpet
cleaning, and most answers ranged from $23 to $27, your deviation is
small. But if most answers stretched from $15 to $45, you would have
a larger deviation.
To find the SD, subtract each individual score from the Mean (treat
all answers as positive numbers), and add up the answers. Divide it
by the number of scores. That's your standard deviation of scores
around the Mean. It's useful to know in comparing the Means of
different questions, such as the rating of your product's value
compared to the rating of your competitor's product. A small SD shows
that most people have about the same opinion. A larger SD shows a
wider variety of opinions, and probably indicates unfamiliarity with
your brand or product class.
If you have good data, there is an enormous amount of useful
information you can get out of a survey by applying statistical
analysis. If you don't understand statistics, it's tough to do,
so call someone who does. Try a college or a small business center.
Of course, professional research firms will also be happy to help you
analyze your data. If you have the funds, you can probably learn the
most that way.
If you leave this article with nothing else, remember that consumer
research is important for businesses of all sizes, that it should
form the bedrock of your marketing strategy, and that even
inexperienced business people can do some basic, useful research
themselves. Good research will let you see your problems before your
competition does, tell you why consumers are buying (or not buying)
your product, and suggest new avenues for stronger sales growth. It's
worth a regular investment.
>> For more on doing research using Focus Groups, click HERE.
(c) Gary Witt, 1998
* Solve My
Problem * About
Us * Free Information *
Products & Services *
Newsletter * Contact
Us * Rewards *
NOTICE: All material on this
site is Copyrighted, and may not be reproduced without the written
consent of the Marketing Psychology Group, Inc., Scottsdale, Arizona.
If quoting text, please provide attribution.