SURVEYS
Primary Reference Sources
Kirakowski, J and M Corbett (1990) Effective Methodology for the study
of HCI. North Holland Elsevier.
Summary description
A survey involves administering standard questionnaires to a large sample
population. Surveys can help determine customer preferences, work practice
and attitudes. There are two types: Closed, where the respondent is asked
to select from available responses and Open, where the respondent is free
to answer as they wish.
Typical Application Areas
Open surveys are applicable in the same kinds of situations as semi-structured
interviews, closed surveys as structured interviews. Interviews are preferable
in situations where there is a lot of person-power available for requirements
elicitation and the user group is easily accessible. Surveys should be
used, by preference, in situations where the user group is spread out geographically
or where person-effort is small. Unless the user group are especially motivated,
response rates of 20% or less are common with questionnaires.
Benefits
Quick and relatively inexpensive to administer. Results can be subjected
to statistical analysis
Limitations
Questionnaire design is not straightforward. It may be hard to follow up
on interesting comments as it is often not desirable or possible to keep
records of who wrote what comment.
Cost of use
Depends very much on the complexity of the survey and the number of respondents
needed. Mailing costs are likely to be high, as are telephone costs if
a follow-up is to be used.
Costs of Acquisition
Survey design requires some expertise and can be costly if inexperienced
staff attempt to use it for the first time.
Suitability for requirements engineering in Telematics:
Many RESPECT partners
have expertise in the area. Widely used in industry.
How to get it
Surveys have to be tailored to individual requirements. Techniques are
widely documented in the literature.
Detailed description of method
Initial steps are the same as for interview design, keeping in mind that
semi-structured interviews are similar to open-ended surveys (ie, the issues
are known, but the range of user responses to them is not); and structured
interviews are similar to closed-ended surveys (ie, the ranger of user
responses is pretty well understood, but the strength of each response
category needs to be determined).
Questions should be posed in as factual way as possible. Evaluative
questions about feelings and interpretations lead to attitude questionnaires
and opinion surveys, which are all notoriously difficult to develop by
the researcher with little knowledge of psychometrics.
User sampling should be used, and if done properly, surveys should employ
a rigorous statistical sampling method to ensure that results are not biased.
However, this recommendation is rarely if ever observed in industry. It
is sometimes done to offer respondents a little gift in exchange for a
returned survey: if chosen appropriately, this can raise response rates
to 80% and above. A low response rate may be followed up with either a
re-posting or better still a telephonic contact. However, these methods
require that users be identified by name to the researcher at least: some
surveys may require total anonymity. It is usual to include a short covering
letter requesting the respondent to reply and a stamped addressed envelope
if possible to make the return as easy for the respondent as can be.
If user information is being kept on computer (as is almost inevitable
these days) care should be taken to ensure that the data privacy legislation
in your country is not breached, and respondents should be assured of this
in the covering letter.