The issue with creating variables about how the product is used and then trying to fit / map your research data to these variables is that you may be skewing / projecting bias as to how you think a product is being used and not letting the research data "speak for itself". As you've collected all your research data (and I'm not sure how you've conducted your research process so I'll make some assumptions), the following is the approach that I'd be taking.
This step should have already been done during your user research phase when identifying what people you wanted to survey and interview, but I'll reiterate as it helps explain the next step. It's important to identify the external and/or internal stakeholders' roles that will be using the product as it will help begin the segmentation process. For example, you wouldn't want to be comparing the research data for an internal customer support staff member against an external customer as there uses of the product will be substantially different.
Identify patterns for each role in the research data collected
The next step is to start identifying patterns for each role in the user research data collected. Though there are different techniques available for identifying patterns, I'd recommend going through the following process:
- Identify traits (such as behaviour, attitude, demographic, goals) from the user research you conducted. I'd recommend going no fewer than 7 but no more than 20. Examples of traits could include the time they use the product, how old the user is, how they feel when an error occurs with the product etc.
- Once you have identified these traits, the next step is to put the trait on a spectrum (similar to a Likert Scale). If the trait can't be put on a spectrum, an acceptable fallback is to make the trait into a multi-choice question. An example of putting a trait on a spectrum could be the time of day they use the product e.g. morning, midday, afternoon, night.
- Once these traits have been put on a spectrum / made multi-choice, the next step is to put each of the users on each trait spectrum based on the user research you gathered. Pending on how you've structured your surveys and line of questioning, this process can be done pretty quickly.
- As you are starting to put these users on the trait spectrums you've made, you'll begin to notice patterns are forming e.g. John, James and Julie both use the product at night, are around 40 years of age and like to talk to support staff when facing an issue. In this example, you would classify John, James and Julie as one persona due to the strong pattern of traits they possess (they won't always be identical and this is ok).
For further reading on creating personas, I'd recommend having a read of this Smashing Magazine article and the Fluid Project website as it really helped me to learn about the process of defining personas.
If you have any additional questions please feel free to ask.