A survey shows that new customers rate setup poorly, while interviews reveal that some are waiting for a colleague to approve access. The findings become more useful when the team connects them: who encounters the approval delay, how does it shape the experience, and what evidence would distinguish that issue from difficulty using the interface?
Mixed-methods research intentionally integrates qualitative and quantitative approaches within a study or connected programme. The integration may happen in recruitment, study design, analysis or interpretation. Running two methods is not sufficient if their findings never meaningfully inform one another.
Decide how the evidence will connect
When the team does not yet know which experiences to measure, qualitative work can come first. Interviews might identify distinct setup barriers and the language customers use to describe them, followed by a survey designed to examine those barriers in an appropriate sample. The survey does not automatically establish population prevalence; its recruitment and measurement still need to support that inference.
When a numerical pattern already needs explanation, the order can reverse. A team might select interview participants from groups with different completion histories and investigate the circumstances behind those differences. Alternatively, both strands can run together and be compared during analysis.
The NIH Office of Behavioral and Social Sciences Research’s mixed-methods guidance sets out principles for purposeful integration. Its methodological lesson transfers to product work: specify what combining the evidence will let you understand that either strand alone would leave unresolved.
Build the connection into the study
In the fictional setup investigation, plan whether survey respondents can be invited to an interview, what permission that requires and how their answers will be linked. If the second study uses a different audience, document the difference so that the combined account does not imply a connection at the individual level that was never established.
Analyse each strand using an appropriate method. A handful of favourable quotations should not stand in for qualitative analysis, and an exploratory interview sample should not be treated as a representative survey because the report includes percentages.
An evidence matrix can help the team compare claims, supporting observations and disagreements across methods. Its value is in the reasoning it makes visible, not in placing a chart and a quotation beside one another.
Preserve disagreement when it matters
Customers may give high overall ratings while describing a difficult setup process. That need not mean one source is wrong: they may value the eventual outcome enough to tolerate the initial work. Check the question, timing and audience before deciding what the apparent tension means.
Triangulation can help organise that comparison, but agreement does not prove a causal explanation, and disagreement is not necessarily a methodological failure. Both can refine the understanding of the experience.
Scope the programme to the decision and available expertise. A smaller study with a clear integration plan can be more informative than several loosely connected activities whose results arrive too late to influence one another. The final account should explain how combining the evidence changed what the team learned.