Editor’s note: Beth Clarkson is a Wichita, Kan.-based independent consultant.

As a researcher, you know how time-consuming and expensive it can be to conduct surveys. The Internet is changing that fact by providing a fast and inexpensive way to contact people with your questions and get their responses. People answer by computer, so their responses can be automatically added to your database...no data entry hassles and no waiting. Finally, the Internet allows you to build an unprecedented amount of flexibility into your surveys. You can collect data from a broad variety of people, yet still get the kind of detailed information usually available only through extensive pre-screening of respondents.

Of course, there are some difficulties. Aren’t there always? Internet respondents may and probably do form a biased group. Therefore you can’t automatically extrapolate the results to a more general population. Anonymity is harder to guarantee and a lot of people won’t believe you even if you do promise it. And if you haven’t worked with the Web before, it can be difficult to get started.

Luckily, there are ways to deal with all of these problems.

Who’s doing surveys and why

There are a variety of reasons to put a survey on the Internet. A commercial site may want to establish the demographics of those who visit in order to help sell advertising on their site. Some Web surveys are aimed at researching the potential market for a new product or service, while others are establishing whether or not current customers are satisfied.

Dr. William Hammers of the Clearwater, Kan.-based Advanced General Aviation Transport Experiment (AGATE) consortium says the organization decided on an Internet survey to conduct market research after determining that the profile of people who buy and fly airplanes and the profile of people who are on the Web are a very close match.

Drugstore.com, a Bellevue, Wash.-based Internet retailer, is using periodic surveys to stay in touch with its customers. Such surveys, says Stephen Jensen, Drugstore.com’s market research manager "are a natural extension of how we would communicate with our customers. It’s not as intrusive as a phone survey, it allows customers to complete the survey on their own and when it’s convenient for them. It’s given us some great feedback on our customers." Jensen is very pleased with the response rates of 30 to 40 percent.

Internet surveys are becoming common as more and more companies are providing computers, e-mail, and Internet access to their employees as a standard part of doing business. Many organizations are finding that such surveys are a fast and inexpensive way to reach their employees and solicit their opinions. The University of Wisconsin - Madison used eListen software from Tustin, Calif.-based Scantron to set up a survey on the workplace for its information and technology group. The Internet was considered a natural choice because everyone already had access to the Internet, so it was easy and convenient to send the invitations by e-mail.

Tara Wolfe, president of International Survey Systems (ISS), Bellingham, Wash., has a client using Internet surveys to assess the return on investment for training and evaluate the effectiveness of different courses. Because the client’s employees are used to working on their computer desktops, an electronic survey fit nicely into their flow of work.

Hosting the survey

There are a number of problems that can come up with hosting the survey, including making sure that your survey will work with a wide variety of browsers, some of which may be quite old. One way to avoid these problems is to have a vendor host the survey. Drugstore.com chose to have Scantron host its surveys. "At some point we will move to hosting it on our site, but right now we are focused on our Web store," Jensen says.

An additional consideration for intracompany surveys is confidentiality. "People still have the same old confidentially issues," says ISS’s Wolfe. "They are probably a little more nervous about electronic surveys." Given the current publicity about privacy on the Internet, particularly when using your company computer and account, this is a realistic concern. One of the ways Wolfe decided to help her client with that issue was to host the surveys through an outside company, which helps people to feel more confident that their comments will remain confidential.

Internet survey design

The usual steps in doing a survey are slightly different for Internet surveys. The first phase - developing the survey itself - is still there, but surveys designed for the Internet are usually a little different from those developed for other media such as pencil-and-paper. Sure, you can take a paper-and-pencil survey and load it onto the Web, but you’re not taking advantage of the Internet’s unique capabilities.

The administrative step is still there too, but with the right software, it’s easier. Wolfe found this to be one of the key advantages of using software to develop Internet surveys. "Once the set-up and prework is done, the administration is really fast and easy compared with traditional paper and pencil. Input and analysis of results and even the distribution is so much easier."

If you want to get a good response to your survey, it needs to load as quickly as possible. People won’t wait; they’ll move on to another Web site. Graphics and other fancy stuff increase the time it takes the survey to load. Although it’s tempting to make your backgrounds beautiful, all those graphics can be a problem. What will the background and font look like on different browsers and at different font sizes? Best to choose something that will look good and load quickly no matter what kind of computer set-up the respondent has.

"Our own reaction at the beginning was to design a fairly fancy questionnaire," says Bob Tortora, chief methodologist at The Gallup Organization. "You’re much better off using a very plain questionnaire - something that looks like a mail questionnaire. When you start getting fancy, you start slowing down transmission times."

Make it easy to respond as well. It’s a lot easier to click on a response than it is to type one in. Match the entry mechanism to the question. Use radio buttons or drop-down menus for a single response, check boxes for multiple answers. It’s useful to include an option for "other" responses, but few people will use it, so it needs to be supplemental, not part of your core data.

Radio buttons are best with five or fewer answers to choose from. More than that, and a drop-down menu is easier to use. For really lengthy lists with short identifiers, such as the 50 states, it’s wise to allow respondents the option of typing in the two-letter abbreviation rather than forcing them to scroll through the list all the way down to Wyoming.

Make sure you’re aware of any default answers. A common mistake for inexperienced Internet survey programmers is to unintentionally have the first response in a list become the default answer. This can seriously skew the results. It isn’t a problem if you use survey design software that avoids such errors automatically, but if you’re doing the programming yourself, you’ll need to test for this potential flaw.

Advantages of Internet surveys

Internet surveys have some real advantages over other methods of collecting data, particularly with regard to interactive branching. Such branching allows you to target the questions, or even which survey the respondent receives, based on the answers they give to earlier questions.

The survey I developed for the AGATE consortium presents pilots, former pilots, and potential pilots each with a different set of questions. Such branching is transparent to the respondent. They answer a question at the beginning regarding their experience and desire to be a pilot and thereafter they only see the questions that are applicable to them.

ISS is using 60 different surveys in one current project. "Each of those surveys is customized for the course taken," Wolfe says. "It’s very easy to do a lot of customization and not as confusing to administer that volume of surveys. If a person responds with option A on particular question, the software allows you to open up a different set of questions. It’s really user-friendly for the respondents."

How easy it is to do such branching depends on the software you are using. Some programs can’t do interactive branching. You can program branching logic for a survey in Java (this is what the AGATE team did) but it may require a considerable amount of time and effort to do so.

Another technique developed for the Internet is to use random numbers to present each respondent with a particular subset of questions rather than expecting each person to answer all questions. It allows the surveyor to collect a large amount of information without burdening respondents with excessively long questionnaires.

This technique is particularly useful in determining prices. Given a selection of prices and asked "What would you pay?", people typically give answers at the low end of the price range. A yes/no question about a specific price, while more accurate, only yields information about that particular price. By using random numbers to present each survey respondent with a specific price, and asking if they would pay that much, you can build a pricing curve for the market potential of the product or service by looking at the percentage of positive responses at each amount.

The AGATE survey at http://apats.org/ uses a combination of such techniques to tailor the questions presented to each survey taker. The result is that, while the survey comprises over 100 questions, no one person sees more than a few dozen. "It’s very efficient for people to take the survey," says Hammers, "yet we gather a large amount of data."

Internet surveys virtually eliminate data entry. Responses can go directly into your database. Essentially, the people taking the survey are performing the data entry task for you, including typing in any text responses.

Because users are putting their data directly into a computer, the answers can be automatically checked for validity. You can make sure that a particular question has been answered if that data is crucial, or check to make sure that the response falls within predefined values. While you can’t screen out all errors, you can eliminate a great many, up front. If the respondent has neglected to answer a key question, the program can let them know and they can make a correction on the spot.

You can program your survey to be available in a variety of languages if you feel that the application warrants it. Some programs will facilitate this and allow you to store phrases that are used regularly in your surveys. Of course there are Internet tools which allow users to translate Web pages into their native tongue if necessary, although the automatic translators do sometimes come up with some interesting interpretations.

Explanations and graphics can be integrated into your survey without detracting from the overall flow. This can help confused respondents understand questions, terminology, and literally see what you mean. On the other hand, since those explanations are only accessed when needed, those who don’t need them are not distracted or slowed down by having to read through definitions of terms they are already familiar with.

Biased results

Biased results are the most potentially devastating problem associated with Internet surveys. They can be handled, but the expense and effort involved with the survey rises. However, in a lot of cases, it simply isn’t an issue. A survey that’s placed on the Web in order to receive feedback from Web site customers, such as the ones that Drugstore.com is running, is an example of such a situation.

In other situations, it may be acceptable if the demographics of the Web are close to the demographics of the people you’re trying to reach, such as for the AGATE survey. It was felt that the data collected would lead to good insights regarding the potential market and the skewed demographics could be corrected by using statistical techniques to weight the actual responses to bring them in line with the demographics of the population at large.

If you do need scientifically accurate results, it is possible to obtain them, but the costs go up considerably. Gallup is meticulous in its use of the Internet in order to guarantee the authenticity and scientific validity of their results. The company first qualifies all respondents with a short phone interview. After that, the qualified recruits are sent a letter with the site address and a personal identification number (PIN). The firm also sends them $5 cash for completing the survey. All of that is done to increase the response rate. Using the PIN, Gallup can determine who hasn’t completed the survey. After two weeks, they receive an e-mail reminder. If they stopped before completing it, they can log back on and finish.

An internal corporate survey may or may not be appropriate to place on the Web, depending both on the company and the employees being surveyed. Boeing Company, for example, has used the Internet for surveys designed for managers and professional workers who are provided with computers and Internet accounts by the company, but when surveying factory floor employees who may not have the access and equipment, the more traditional pencil and paper methodology prevails.

A matter of time

If you’re not using the Internet to conduct surveys yet, it’s only a matter of time. While the Internet will never entirely replace the more traditional forms of communication, for surveys or anything else, if you plan on doing market research in the future, you need to be looking into how to make use of the Internet now. Sure, problems exist, but that’s true with every methodology. The problems can be dealt with.

The combination of access, convenience, and low cost give the Internet some major advantages in collecting information. Add to that the flexibility and ease of use that can be programmed into Internet surveys and it’s a winning combination for market researchers.