How do you determine when data saturation has been reached in qualitative research?

There is considerable variation over what constitutes data saturation in a research study. To be rigorous, it's always helpful if we can justify or explain why we feel we have reached saturation or not, and how much information is expected to be produced the more data we collect in a study. Empircal studies suggests the most novel information in a qualitative dataset follows an asymptotic curve and there is a relatively sharp decline in new information after just a handful of interviews (see Guest, Namey, and Chen 2020). I would argue that one factor to consider in applied research is risk mitigation, and whether your organization can tolerate risk of the minimum sample size or if it requires going above and beyond to meet saturation.

Determining when data saturation has been reached in qualitative research often hinges on recognizing patterns. As you collect and analyze data, reaching a point where no new information or themes are observed in the data signals saturation. This means the data collection can be deemed sufficient and complete, as further data would only reinforce existing findings without contributing new insights. Recognizing these patterns is crucial as it ensures that the research is comprehensive and the conclusions drawn are robust and reflective of the dataset.

Beyond this, one effective method that can be used is the triangulation of data sources. This means gathering data from different types of sources or using various methods, such as combining interviews with focus groups or observations. By corroborating information across multiple sources, you can verify that the emerging themes are robust and consistent, reinforcing the indication that no new information is being uncovered and that saturation has likely been reached.

Data scientists are the first to cite “garbage in, garbage out” when it comes to statistical modeling or predictive analytics. The same principle holds water in qualitative research. If you are not owning the requirements for identifying target users you run the risk of not talking to the right people or finding useful data that can inform your analysis. This dilutes the quality of your research and wastes precious time which is almost always in short supply. It’s best to spend time up front really defining the requirements and identifying target characteristics that matter for the goals of your study. Sometimes in qualitative research finding the right participant can unlock a wealth of insights so it’s critical to prioritize quality.

Once you establish data saturation, don't forget to periodically check for data drift, which can mark the introduction of new information and patterns, leading to desaturation of previously saturated data.