Methods Menu

Sort your questions by the kind of answer they need, pick methods from a short menu that fit those questions and your constraints, and plan how the methods add up to one answer.

Mixed evidence across the menu: controlled experiments are the strongest route to causal claims (Fisher, 1935; Kohavi et al., 2020), and sample size logic is settled statistics (Cohen, 1992). How many interviews are enough is a mixed literature (Guest et al., 2006; Hennink & Kaiser, 2022). The mixed methods designs are a methods typology, not a tested intervention (Johnson & Onwuegbuzie, 2004). Sample sizes and timings marked as rules of thumb are just that.

Use it when:

• You have framed questions (see 2.1 Problem & Question Framing) and are deciding how to answer them.
• Someone has already proposed "a survey" or "a few interviews" and you want to check the choice fits the question.
• You have two methods in mind and need to decide how they work together.

Skip it when:

• The answer is already sitting in existing records: start with document analysis and 2.8 First-Pass Analysis.
• The method is chosen and you need to plan the fieldwork: go to 2.5 Simple Study Design Sheet.

There is no best method, only a fit between a question and a way of answering it. The menu follows the mixed methods tradition, which holds that qualitative and quantitative approaches answer different kinds of questions and that combining them is a design in its own right (Johnson and Onwuegbuzie, 2004). Some entries stand on firm ground. Controlled experiments remain the most reliable way to support a causal claim, from the original logic of designed experiments (Fisher, 1935) to modern online tests (Kohavi et al., 2020). Sample size statistics are settled: a study too small to detect the effect it looks for has decided its result in advance (Cohen, 1992). Other entries are contested. How many interviews are enough is still debated, with saturation sometimes reached early in similar groups (Guest et al., 2006), and being observed changes what people do, because they respond to what they think the researcher wants (Orne, 1962).

So every method on the menu shows its weakness next to its strength, and Step 3 makes you write why each choice fits. The first box in Step 5, methods match the questions, is the whole page in one line.

Take your focused questions from 2.1. For each, decide whether it asks why or how (explore), how much or how many (measure), or whether one thing causes another (test). Then write what evidence would convince the person who decides.

Strong answer: "Why do small cafes stop ordering? Explore. The owner would be convinced by cafes describing, in their own words, what changed before they stopped." Not: "Customer research."

Illustrative example: the business, the roles and all numbers are made up.

The Gothenburg bakery from 2.1 has two focused questions about small cafes that stop ordering.

Step 1: "Why do small cafes stop ordering?" is an explore question; the owner wants to hear it from the cafes. "How many cafes stop soon after an order at the minimum?" is a measure question; the owner wants a count from the order system.

Step 2: Ticked: interviews, record analysis. A survey was considered and dropped: with about forty cafes in total, a survey adds little that interviews and records cannot give.

Step 3: Primary: sixteen interviews, eight with former and eight with current cafe customers, weeks 1 to 3. Secondary: two years of order history for every cafe, week 4.

Step 4: Sequential. Interviews find the reasons; the order data checks how common the pattern is.

Step 5: All boxes ticked except access. Weakness: former customers may not answer. Plan: ask through the drivers, who know them, and report how many said no.