Surveys can be used to infer some attribute or property about a population which have been derived from a sample taken from that population. Surveys can be carried out by mail, phone, face-to-face and/or across the internet. The results can then be collated and broken down (both graphically and through statistical analysis) to provide the relevant information which can be acted upon.
BDRI Ltd conducted an online survey for one of the Faculty libraries at a Midlands University (confidentiality agreements prevent the disclosure of which one) and this provides a case study of how a survey can be deployed and the results generated for a reasonably large number of potential respondents.
By way of background to this survey, the University has two sets of students within this particular Faculty: the traditional undergraduate course which is lecture based and a graduate entry course which uses a problem based learning approach. The library staff wanted to know if there was a difference in the way the two sets of students (ie their customers) used the library and valued their services. This questionnaire was developed to obtain this information, mostly through quantitative, tick box answers, although there were also a couple of open questions in which the students could explain their reasoning.
Emails were sent to all the students (about 350 in total) with a link back to the questionnaire. This was hosted on a page on the BDRI website, so as to ensure independence from the University. Students clicked on the link and filled in the questions online. In this case, a census was possible – ie all members of the population were surveyed due to the web based method chosen – but the same principle would apply for a sample of a larger population.
A copy of the online survey can be viewed by clicking on the link below:
Library Use Survey
The survey results can then be “picked up” by the survey software which is used. In this example, the results could then be exported as a .csv file directly into Microsoft Excel, and all the graphical and statistical analysis could be done in Excel. The survey software also provided tracking to which students filled in which answers. There was also the option for students to fill in their email addresses to allow further contact for a focus group.
The same type of questions (eg tick box and/or open questions) could be used for any survey depending on what information was required. The pages can also be made to look a bit more visually interesting, for example, including company logo and/or brand colours. In this example, no branding was required. Statistical analysis can include mean, median, range, standard deviation, hypothesis testing, t-tests, chi-square tests, statistical significance etc.
A selection of the types of results and analysis which were conducted for this survey are given in the Excel spreadsheet which can be viewed by clicking on the link below:
Library Use Example Results
This spreadsheet (Library Use Example Results.xls) has five worksheets which give the results from some of the individual questions. The first sheet (FC_Data1) is the base information which was collected by the software from the survey responses of, in this case, the graduate entry students. The data is used in the subsequent analysis and in generating the latter graphs. The responses to the main open question are also included in this sheet and the text can be analysed to identify common words, phrases or themes.
An example of the results and the basic bar graph which is produced from one of the questions is given in the next sheet (Q9. Skills Train’g Satisfaction). This bar graph shows the absolute numbers of respondents who clicked on a particular answer.
The next two sheets (Q3. Frequency – Course Books and Q4.Importance – Course Books) take the analysis further by showing the percentage of respondents rather than just the absolute numbers. This allows comparison between data sets. This is a simple segmentation exercise and in this case, the two data sets are for the two different types of students. One is coloured green and the other is coloured orange to aid comparison.
The graphs look similar as the same number of respondents happened to came from each group. However, if one set of respondents was much larger in number than the other, a direct comparison of absolute numbers would not be meaningful and hence the preference for percentages. The results can also be shown as a pie chart instead of a bar chart. All these graphs are easily copied into any report or Powerpoint presentation.
Finally, the last sheet (Add’l Comparison Analyses) shows a comparison of average levels of, in this case, frequency of use and importance across several individual areas. This provides an overall picture whilst the previous sheets provide more of the detail as necessary. The Y-axis error bars on these graphs show the sampling error at a 95% confidence interval. This means that we can be 95% sure that if the sample was repeated, the results would still be between the upper and lower limits shown ie that the true result lies between these limits as well. This is important as it means that if the error bars for a particular attribute eg frequency of use of course books do not cross over each other, we can say that there is a statistically significant difference between the two results. This means that there is very likely a real difference rather than one due to sampling error. The reasons for this could now be investigated and resolved.
Conducting and Analyzing a Customer Survey
Surveys can be used to infer some attribute or property about a population which have been derived from a sample taken from that population. Surveys can be carried out by mail, phone, face-to-face and/or across the internet. The results can then be collated and broken down (both graphically and through statistical analysis) to provide the relevant information which can be acted upon.
BDRI Ltd conducted an online survey for one of the Faculty libraries at a Midlands University (confidentiality agreements prevent the disclosure of which one) and this provides a case study of how a survey can be deployed and the results generated for a reasonably large number of potential respondents.
By way of background to this survey, the University has two sets of students within this particular Faculty: the traditional undergraduate course which is lecture based and a graduate entry course which uses a problem based learning approach. The library staff wanted to know if there was a difference in the way the two sets of students (ie their customers) used the library and valued their services. This questionnaire was developed to obtain this information, mostly through quantitative, tick box answers, although there were also a couple of open questions in which the students could explain their reasoning.
Emails were sent to all the students (about 350 in total) with a link back to the questionnaire. This was hosted on a page on the BDRI website, so as to ensure independence from the University. Students clicked on the link and filled in the questions online. In this case, a census was possible – ie all members of the population were surveyed due to the web based method chosen – but the same principle would apply for a sample of a larger population.
A copy of the online survey can be viewed by clicking on the link below:
Library Use Survey
The survey results can then be “picked up” by the survey software which is used. In this example, the results could then be exported as a .csv file directly into Microsoft Excel, and all the graphical and statistical analysis could be done in Excel. The survey software also provided tracking to which students filled in which answers. There was also the option for students to fill in their email addresses to allow further contact for a focus group.
The same type of questions (eg tick box and/or open questions) could be used for any survey depending on what information was required. The pages can also be made to look a bit more visually interesting, for example, including company logo and/or brand colours. In this example, no branding was required. Statistical analysis can include mean, median, range, standard deviation, hypothesis testing, t-tests, chi-square tests, statistical significance etc.
A selection of the types of results and analysis which were conducted for this survey are given in the Excel spreadsheet which can be viewed by clicking on the link below:
Library Use Example Results
This spreadsheet (Library Use Example Results.xls) has five worksheets which give the results from some of the individual questions. The first sheet (FC_Data1) is the base information which was collected by the software from the survey responses of, in this case, the graduate entry students. The data is used in the subsequent analysis and in generating the latter graphs. The responses to the main open question are also included in this sheet and the text can be analysed to identify common words, phrases or themes.
An example of the results and the basic bar graph which is produced from one of the questions is given in the next sheet (Q9. Skills Train’g Satisfaction). This bar graph shows the absolute numbers of respondents who clicked on a particular answer.
The next two sheets (Q3. Frequency – Course Books and Q4.Importance – Course Books) take the analysis further by showing the percentage of respondents rather than just the absolute numbers. This allows comparison between data sets. This is a simple segmentation exercise and in this case, the two data sets are for the two different types of students. One is coloured green and the other is coloured orange to aid comparison.
The graphs look similar as the same number of respondents happened to came from each group. However, if one set of respondents was much larger in number than the other, a direct comparison of absolute numbers would not be meaningful and hence the preference for percentages. The results can also be shown as a pie chart instead of a bar chart. All these graphs are easily copied into any report or Powerpoint presentation.
Finally, the last sheet (Add’l Comparison Analyses) shows a comparison of average levels of, in this case, frequency of use and importance across several individual areas. This provides an overall picture whilst the previous sheets provide more of the detail as necessary. The Y-axis error bars on these graphs show the sampling error at a 95% confidence interval. This means that we can be 95% sure that if the sample was repeated, the results would still be between the upper and lower limits shown ie that the true result lies between these limits as well. This is important as it means that if the error bars for a particular attribute eg frequency of use of course books do not cross over each other, we can say that there is a statistically significant difference between the two results. This means that there is very likely a real difference rather than one due to sampling error. The reasons for this could now be investigated and resolved.