Mixed-methods research: when to combine qual and quant

Mixed-methods research is the planned mix of qualitative and quantitative methods under one research question. Qualitative work shows why someone gets stuck. Quantitative work shows how often that happens. Together they answer what neither answers alone. NN/g describes three designs: quant then qual, qual then quant, and both in parallel. Let the question pick the design, not the urge to “do a bit of both”.

What counts as mixed-methods research

The Nielsen Norman Group article on mixed-methods research defines the approach as the intentional join of qualitative and quantitative work in the same project, under the same question. Dropping a survey next to five interviews isn’t enough. Both arms need to be planned together and read against each other.

Quantitative methods show patterns and trends at scale. Qualitative methods show motivation, frustration, and mental models. Without the first, you don’t know the size of the problem. Without the second, you don’t know what to change.

The article separates planned mixed methods from improvising halfway through. Adding a second arm later can still help. The focus here is on designs where both arms start together, aimed at the same goal.

Why one method leaves a gap

The article that compares quantitative and qualitative testing keeps the split clear. Qualitative work answers “why” and points to the interface problem. Quantitative work answers “how many” and “how much”, with success rate, time, and satisfaction.

Quantitative work alone says that 40% finish the task and doesn’t say what to do next. Qualitative work alone finds the defect and doesn’t prove it hits the whole base. That is why NN/g treats them as complementary in the iterative cycle: qualitative work guides the redesign, and quantitative work measures the gain (NN/g).

Cost changes too. A qualitative study often runs with few participants. A quantitative study needs a larger sample, often above 30, so the number can survive noise. So quantitative work shows up less often, and almost always on a product that is already live.

The three designs

The mixed-methods article lays out three common UX designs.

Explanatory sequential: quantitative, then qualitative

You measure first and only then dig into the why. Use this when the number will raise a question the number can’t answer. An A/B test shows which version converts more. It doesn’t show why the other lost. The plan already includes qualitative testing on the tasks where the quantitative arm found friction.

Exploratory sequential: qualitative, then quantitative

You explore first and only then measure. Use this when the topic is still fuzzy. Contextual inquiry with a small group yields hypotheses. Later a survey or a tree test measures what generalizes. The qualitative arm writes the instrument for the quantitative arm.

Convergent parallel: both at the same time

Both arms run together and analysis crosses the findings. Use this when the deadline can’t wait for one arm to finish, and when neither arm depends on the other’s result. A MaxDiff survey, where people pick the most and least important item in each set, ranks features while interviews ask for the reason behind the choice. At the end, you compare where the two agree and where they diverge.

How it worked on BTG+

On the BTG+ Business pages, I used both sides of the same question: where the conversion flow lost people, and why. Funnel analysis in Google Analytics showed the step with the steepest drop. Hotjar recordings and heatmaps showed what people did at that step.

I read the two arms together. The number showed where. The recording showed why. The redesign simplified the conversion flow on the business pages. In the three months after launch, credit-card adoption rose 22%, measured by the client’s data team.

This isn’t about “using two tools”. It is one question, answered by scale and by behavior. Tool-by-tool funnel reading is in the post on funnel analysis with GA4 and heatmaps.

What to decide before you start

Three warnings from the mixed-methods article work as a checklist.

Plan the time. Two protocols, two recruitments, and two data types add up. A quantitative survey may need about 40 responses. A qualitative round often needs about 10, until saturation.
Align method to the question. Use each arm for what it does best. Don’t run both just to look complete.
Integrate, don’t add. Ask how the results will connect. One method explains the other, or both triangulate the same finding.

When qualitative work enters analysis, NN/g still asks for care with the weight of each observation. The four-step analysis guide separates collecting data, assessing accuracy, explaining, and checking whether the explanation fits the set. A positive comment after a leading question doesn’t weigh the same as spontaneous behavior.

Where the sources diverge

When to improvise. The mixed-methods article allows adding an arm mid-project, but treats the planned design as the central case. The qual-versus-quant text describes an iterative cycle with distinct phases, and barely names mixed designs.
What the number proves. The qual-versus-quant text treats quantitative work as the instrument that lets you talk about ROI and measured gain. The mixed-methods article treats quantitative work as one arm, and the value shows up in the join with qualitative work.
Sample size. The qual-versus-quant text talks about few participants in qualitative work and often more than 30 in quantitative work. The article on five participants stresses that the five-user rule applies only to qualitative work. The mixed-methods article cites about 40 for a survey and about 10 for interviews as a cost order of magnitude, not as a closed statistical rule.

The common mistake

The mistake is mixing methods without mixing the question. Running interviews one week and a survey the next, each with its own goal, isn’t mixed-methods research. It is two studies side by side. Before you recruit, write the single question, the chosen design, and what each arm will answer. If one arm doesn’t change how you read the other, you can drop it.