Statistics are inferences drawn from a sample population

Useful statistics are valid inferences drawn from a representative sample population

Be careful of drawing too many inferences from a small sample.  Any results that are very unexpected should be carefully reviewed and perhaps even conducted again using another surveying instrument – e.g. interviews.

Recommended sample sizes (from ARL LibQUAL+™ Policies and Procedures Manual)

Institutions with less than 10,000 students may wish to survey their whole population instead of sampling.

The minimum recommended sampling figures for large academic library systems (>10,000 students) are as follows:

900 Undergraduate students
600 Graduate students
600 Faculty
600 Non-library University Staff (optional)
All Library Staff (optional)

Representative Random Sample

Try to ensure that your sample is representative of the target groups, e.g. If social science faculty represent 30% of the faculty, your sample should include 30% social sciences faculty. 

No matter how representative your original sample, the representativeness of your survey population will ultimately depend on the dempographics of your repondent population.


Email surveying tends to elicit lower response rates that more direct survey methods (e.g. telephone or person to person interviews) and is dependent on the reliability of the email addresses.  So, over-sampling is encouraged.  Also bear in mind that response rates for undergraduates tend to be lower than other academic library users.


A reliable set of results depends on (a) representative sample (b) size of result set. Your results may provide enough data to infer how your overall population, e.g. faculty, perceive a service.  However, if the respondents' demographics are not representative of your population, you may not know with certainty whether a particular group of faculty have not skewed the results.  A sufficiently large response from a target population can yield reliable results with a low margin of error.  A sample of 1,100 gives you a low margin of error of 3%. At that sample size, the margin of error will not change whether the total population is 250,000 or 250 million.   So, increasing the number of target respondents beyond 1,100-1,300 is overkill and will not increase the reliability of the results.

Formula for margin of error = 1 ÷ square rootno. of people in the sample

A major benefit of the LibQUAL™ survey is the opportunity to correlate your results with ARL's aggregate scores for similar types of libraries. 

Source of Sample Data

The library patron file can be a handy, reliable source for extracting email addresses provided that it is updated regularly.

Importance of Free-text Comments

While statistics provide indicators of patterns and trends, respondent comments can help direct you to possible solutions and actions. LibQUAL+™'s free-text comments offer valuable information about the reasons for respondents ratings and suggestions for improving services - even in categories where the number of respondents may too small to draw reliable inferences from the statistics by themselves.