Guide | Good practice for sampling

A guide on targets and tactics for making the most out of your sampling practice...

If you can’t collect data from every audience member you need to take a sample.

A sample is a representative portion of a population that you are interested in – and in the case of arts organisations the population you are most interested in is your audience. Unlike a census, which aims to include everyone in your target population, a sample gives an indication of the attributes of the wider population by looking at those of a smaller portion.

Choosing an appropriate sample size

Generally the larger the sample size the more accurate the data is, therefore the more accurate and robust your conclusions are about the whole audience. A sample size should be large enough that the results from the sample are applicable to the wider audience, to an appropriate degree of accuracy. The following table shows appropriate target sample sizes for different audience numbers, assuming you want to describe the entire audience at an event or organisation.

Total number of visitors (per year/event)

Sample required for

5% margin of error

Sample required for

8% margin of error

Sample required for

10% margin of error

20,000 or more

377

149

96

10,000

370

148

95

5,000

357

146

94

1,000

278

131

88

500

218

116

81


The ‘margin of error ’ relates to the extent to which you can generalise findings about your sample to the wider audience. Although it is important for the robustness of your data to have a good sample size this is very dependent on the resources at your disposal. However it is recommended that a 10% margin of error is this upper limit.

To calculate your sample size more accurately, there are many online sample size calculators.

Setting targets

Setting hourly and weekly targets helps to ensure the sample size targets are met.

The hourly targets for face-to-face interviews depends on the length of the interview process and the flow of your visitors. When setting targets for collecting email addresses it is also important to take in to account that not all visitors who volunteered their details will end up completing the survey.

The table below gives estimates of reasonable targets per hour based on a continuous flow of visitors. The fourth column accounts for the fact that many people who give an email address will not go on to complete the survey.

Methodology Typical length of interview Max number of responses per hour Max completed surveys per hour
Face to face interview

5 mins

5

5

Collecting e-mail addresses for e-survey

1 min

20

5

A weekly target will depend on the length of your data collection period. The shorter the period, the higher the number of responses that will need to be collected per week.

The table below outlines estimated weekly target based on a four week months and an overall total target of 380 surveys.

Methodology 10 month data collection period 5 month data collection period 3 month data collection period
Face to face interview

10 per week

20 per week

30 per week

Collecting e-mail addresses for e-survey

Collect 40 email addresses for 10 completes

Collect 80 email addresses for 20 completes

Collect 120 email addresses for 30 completes

Sample frame

A ‘sample frame’ should sketch a rough data collection schedule aiming to reduce ‘day of week’ and ‘time of day’ bias and as such should proportionally reflects an organisation’s whole offer and flow of visitors . It should also highlight optimum periods for fieldworker data collection.

In practice, if you expect twice as many visitors on Saturday as you do on Friday, you should aim to conduct twice as many interviews on Saturday than Friday. Similarly, if you expect Saturday afternoon to attract more people than Saturday morning, more interviews should be conducted in the afternoon than in the morning.

Random sampling

Random sampling is based on the premise that each person attending an event in theory has the same chance of being included in the survey. A truly random sample will be as representative of an audience as possible, and not favour certain types of visitors over others.

There are always certain elements that affect a random sample, which can be hard to overcome e.g. language barriers, physical or practical considerations. But there are some elements that can be minimised such as interviewer bias, time of day or location of interviews.

1) Minimising interviewer bias

Interviewers tend to be more likely to approach audience members who look ‘friendly’. To ensure that the sample is as random as possible:

  • Ask fieldworkers to approach every second person or group passing by.
  • If approaching a group, interview the group member whose birthday is soonest rather than a group member who ‘volunteers’.

2) Sampling in a crowded environment

Interviewing member of the public can seem daunting at first, however there are some ways of making it easier to single respondents out from the crowd:

  • Position the interviewer near an entry/exit where there is a steady flow of people passing, in one direction.
  • Try to be methodical – choose a small, central area and approach anyone entering that space.
  • Walk to and fro in a straight line and ask, for instance, every third person that comes within one metre.


3) How to maximise response rates

If you are collecting information face-to-face ensure that:

  • The interviewer is confident to approach and engage with people.
  • The respondent is clearly informed why it is important for them to take part in the research, how long it will take and what is in it for them.
  • The respondent is assured of confidentiality.
  • The interviewer is properly briefed about the purpose and scope of the research.
  • Questions are only asked if useful and relevant – keep it short!