It's not possible to research each and every one of your users. Creating a strategy to find the most appropriate or effective group of users to recruit will make the most of limited time and budget.
Creating a sample is the first step for many methods of understanding users, including one-to-one interviews and focus groups.
Start by brainstorming the user attributes that you think influence behaviours in relation to your project. Then choose the most important attributes to determine a useful range of people to study. For example, if you were designing a bike for female commuters you might want to study people who commuted by different means (train, car, bus), and in geographical areas with different weather, as well as people of different size and strength.
Other common attributes to consider might be age, life stage, ethnicity and socio-economic background, as well as emotional characteristics or attitudes.
Be wary of speaking to more people than you have time to analyse. There is a trade-off between the number of people you speak to and the number of insights you will get. Often a sample of six to nine people will be enough.
Remember, your sample doesn't have to be representative. In fact, talking to non-representative or extreme users will often yield the most insight and inspiration for your project. For example, if you were designing a bike for commuters you might want to include a cycle courier in your sample.And don't make the mistake of only talking to people who use your products or services - speak to those who don't use them too. Don't only talk to people who love commuting by bike, but include those who have tried it and hated it, as well as those who love commuting by car. Always be mindful of who you haven't met.
The sample you choose will depend on what you are trying to get from your users. If you are researching to identify opportunities, a diverse sample including extreme users can often yield the best results. If you are researching to validate a resolved design (for example in a focus group ), on the other hand, then a more representative and less diverse sample may be more appropriate