The magazine of the Melbourne PC User Group
JDJ Software WebSurveyor 1.2.86
Ash Nallawalla
ash@melbpc.org.au

WebSurveyor is an easy-to-use survey package from JDJ Software that can be used in conjunction with a Web site or by e-mail. It does not require you to be a statistical expert, nor does it take days to analyse results. In fact, you can get results within seconds, which will appeal to business people who want to get answers to their questions without a major budget. You can download the software from the company's Web site and build a 20-question survey without paying for it. If you wish to use it for building larger surveys, you should pay for it.

Features

The program's features are:

  • Number of questions per survey limited by disk space and memory
  • Large, extensible library of questions for different types of surveys
  • Export responses to dBASE data files
  • Import questions directly from a text file. Ideal for creating large surveys quickly
  • Generate a complete set of Web pages containing analysis in a few mouse clicks
  • E-mail notifications to large groups
  • Randomly sample E-mail list.
Building the survey

WebSurveyor consists of two primary components: the WebSurveyor application, which helps you generate your survey and process the results, and the WebSurveyor Server extension, which processes incoming surveys and passes them back to the WebSurveyor application


Figure 1  Building the survey

You will probably want to use the supplied Survey Builder Wizard for your first attempts. It prompts you to make certain choices that will define the look-and-feel of your questionnaire. For example, you can populate the survey with some sample questions and you can define its appearance. Once you understand how it works, you can have a crack at customising your survey. As the survey file is merely an HTML document in the shape of a Web page, certain types of complex branching are not possible if the questions are scattered all over the survey. For proximate questions, it is possible to create a Skip Group of questions that are skipped depending on the choice of answer to a previous question. The software package and Web site contain a lot of good advice on building surveys, which will be appreciated by novice pollsters.

Publishing the survey

When you "publish" the survey, you generate the HTML file that you will later upload to the Web site. You also confirm that the server component (a Perl script running under Linux or Solaris in my trials but you might prefer a version that runs under Windows NT) has been configured correctly.

You may publicise the Web address of the survey to your potential respondents. Alternatively, you might have a list of e-mail addresses of the potential respondents and choose to e-mail to a random sample or to all of them. If you choose the latter option, you can add a unique identifier for each e-mail so that a respondent cannot submit twice or rely on their e-mail address or their Internet Protocol (IP) address to detect duplicate submissions. For the PC Update survey, which is anonymous, I did not use the e-mail option as I don't know our readers' e-mail addresses and I could not use the IP address because many of the respondents use a proxy server, which would present the same address for many hundreds of respondents.

Respondents click and type their way through the survey and when they press the Submit button, their data is appended to a text file that resides on the server. You can choose to send the respondent to a special page or simply display some closing comments.

 


Figure 2  A pie chart format used in analysis

Analysis

When we last analysed a Melb PC survey on our own, we had a small team of volunteers keying in results for days, employing complex procedures to avoid errors. With WebSurveyor, you click a mouse key while you are online, download the data, and within seconds, the graphs, charts and totals are all ready!

You can export the analysis to a Microsoft PowerPoint or Word document, making it easy to share the findings with others. You can also publish the results to a Web site (take care not to use the same address as the survey, unless you want to overwrite the questionnaire). When a viewer clicks on a question, the results for that question are displayed.

Sometimes a pie chart is not the best display format because long labels tend to squash the shape of the pie. If the displayed pie chart shows overlapping labels, you can rotate the pie until you find a suitable angle. Sometimes a bar graph or another chart shows the results in a clearer manner.

You can export the responses to a Microsoft FoxPro or Comma Separated Values (CSV) file and thence take it into Microsoft Access or a specialist statistical package.

Problem areas

Reported problems so far reflect the user's PC configuration rather than the survey software. In a long survey, the more check boxes there are, a Windows PC with low memory that is also running 16-bit applications will have difficulty in completing the survey. I had only three complaints to date, which were from Netscape Navigator users. It is a resource-allocation issue caused by the Windows GDI. Mac users seem to have a similar problem.

The solution is to have shorter surveys (which respondents prefer) and to use drop-down list boxes rather than radio buttons, check boxes, text boxes and so on. Microsoft Internet Explorer is a little more resilient in this regard.

 


Figure 3  Survey properties tab

Wish list

In multiple-choice questions, I would like to see an option to display each item as a percentage of the total responses. As an example, in the question about peripherals owned, 164 out of 174 respondents owned CD-ROMs, yet the displayed percentage (12.13) was meaningless as it was measured against 1352 total peripherals.

For numeric questions, I'd like to see the average computed to two decimal places. I would like to see the Median as an alternative to the Mean, as it is more appropriate for certain types of responses where a pure average produces a misleading answer. That option should be released very soon, perhaps by the time this article appears in print.

Conclusion

WebSurveyor is extremely easy to use if you have some experience in configuring a Web server and have administrative privileges there, although JDJ Software intends to offer the option to host your surveys. That will remove the need for special access to your Web site machine and you will only need to run the application on your own PC.

I was very impressed with the speed and quality of support by JDJ Software staff. For example, the first version of the Unix server Perl script did not support file locking. I have been burnt by other scripts that didn't have file locking and had ended up with a zero-byte file, thereby losing my data. JDJ had such a version for their own use but at my urging, they released that version shortly thereafter.

I am very heartened at the 563 responses to date in the two weeks that the PC Update survey has been placed online and the ease of the entire process. That inspired me to seek the Committee's approval to survey the entire readership with WebSurveyor so that all of you can have your say. I have no hesitation in recommending this program.

Reprinted from the June 1999 issue of PC Update, the magazine of Melbourne PC User Group, Australia