|
Here is a brief guide to the options you have for sourcing respondents and planning your sample.
SAMPLE SIZE AND CONSIDERATIONS
At a basic level, the larger your sample size the more reliable the results. Therefore, the usual determination of sample size is your budget. If you can afford a large sample, go for it!
However, to make the most of a limited budget, the starting point is to precisely define the 'universe' you are trying to sample. The more tightly you define the universe, the smaller the sample you can use and still get reliable results.
As a rule of thumb if you want to compare the results of two groups from your research, you'd need to ensure there were at least 100 people in each of the two comparison groups for your analysis to have validity.
SAMPLES REPRESENTATIVE OF THE GENERAL POPULATION
Many Research Connections clients start by requesting a sample 'representative of the general population across age, gender, and region'. A minimum sample for this type of study is often 500 respondents or 'Completes', and even then it is not possible to extensively segment the results. If you don't need to segment the final results and do need a general-population-representative study, we'd recommend a sample size of 500 as a minimum.
If you wish to do some segmentation, or is a small statistical margin of error is essential, you may need to consider a minimum sample of 1,000, and possibly more.
If you can't afford a large sample, give thought to more tightly defining the population whose behaviours and views you wish to understand. For example, for research about a supermarket brand, a sample of 300 may be adequate if all these people are mothers of young children, main and joint household shoppers, and living in main urban areas.
'Hard' and 'Soft' Quotas
When dealing with different panel providers you may (or may not) hear the terms 'Hard' and 'Soft' quotas. This is an important distinction and impacts on the cost and quality of your sample.
The best way to demonstrate the difference between the two is by a simple example.
Just say you requested a sample that was proportionally the same as the general population for gender, age group, and region, and that the population was structured like this:
Gender |
Region |
Age |
Female 50% |
Northern 50% |
Under 30 years 30% |
Male 50% |
Central 40% |
30 to under 60 years 40% |
|
Southern 10% |
60 and over 30% |
With a 'Soft' quota, your final sample will reflect this structure, however conceivably (although reputable panel companies would rarely do this) all the respondents in the Northern region could be females, and all the respondents in the Central and Southern regions could be males. All the younger people could be males, and all the older people could be females. Despite these obvious imbalances the Soft quotas set for the study would still be met at the macro level.
With a 'Hard' quota, the sample is structured right down to the lowest level so your requirements could actually be specified as follows:
female, under 30, northern |
7.5% |
female, under 30, central |
6.0% |
female, under 30, southern |
1.5% |
female, 30 to under 60, northern |
10.0% |
female, 30 to under 60, central |
8.0% |
female, 30 to under 60, southern |
2.0% |
female, 60 +, northern |
7.5% |
female, 60 +, central |
6.0% |
female, 60 +, southern |
1.5% |
male, under 30, northern |
7.5% |
male, under 30, central |
6.0% |
male, under 30, southern |
1.5% |
male, 30 to under 60, northern |
10.0% |
male, 30 to under 60, central |
8.0% |
male, 30 to under 60, southern |
2.0% |
male, 60 +, northern |
7.5% |
male, 60 +, central |
6.0% |
male, 60 +, southern |
1.5% |
Obviously it is much easier and cheaper for a panel company to meet 'Soft' quota targets as many fewer people will be filtered out of the survey. It is very important to ask a panel provider how they work out their quotas so that you know what you will be delivered and what the estimated prices cover.
When Research Connections is sourcing panel respondents for you either from our own panels or third party providers we will discuss with you the options for quotas.
QUOTA OR SEGMENT SAMPLING
A common methodology for online sampling, especially using online research panels, is Quota Sampling (not to be confused with 'setting quotas' to achieve a representative sample). A less confusing term for this could be 'Segment Sampling'. This can be a really effective way to get the maximum value from a smaller budget and smaller sample.
With Quota or Segment sampling you define the precise number of each type of person you want to talk to. Examples of studies we have done like this are:
- 50 each of males aged 15 to 24, 25 to 34, 35 to 44;
- 100 respondents from each of 5 client branch regions;
- 150 people who have tried a product in the past year, and 50 people who have not tried it.
MIXED SOURCE RESPONDENT SAMPLING USING EMAIL LISTS AND ONLINE PANELS
Another option is to use two sources of respondents. If you have a lower incidence of people in the general population who fit your sampling criteria (e.g. used your product in the past 3 months, aware of your service) it can be very expensive to source these people from an online research panel.
On a number of occasions we have used a client database to reach 'users' and our online research panel to reach 'non users'.
The different groups come into the same survey but we automatically vary the survey branding depending on where they are coming from, and final results can be split easily into the two groups.
Using you're an end-client email lists is invariably the cheapest way to access their own customers.
PRE-POLLING
Where you don't have your own list to use, another option to cost effectively reach a 'difficult-to-reach' group is pre-polling called a Buzz Snap Poll.
With this approach, we send a very simple short survey to a wide group of people to see if they meet the survey criteria.
Only people that qualify with this first survey are then invited into the main survey. |
