One question may require complementary choices across all three method dimensions

What it is

A product question often contains several inferences, requiring separate evidence choices across attitude–behavior, qualitative–quantitative, and natural–controlled dimensions rather than one all-purpose method. For checkout abandonment, telemetry estimates where behavior occurs naturally, interviews examine interpretation, and a controlled task isolates wording or latency. Multi-dimensional method combination is organized by subquestion and evidence role, not method count.

Why it happens

Each source observes one projection of a phenomenon. Connected projections separate prevalence, process, meaning, and the effect of changing a factor. Combination also compounds bias: unmatched samples, windows, or constructs can create artificial disagreement. Value arises when one source reduces ambiguity in another or makes an unresolved conflict explicit.

Studying it

Decompose the question into estimation, interpretation, causation, and generation, then assign required positions. An evidence-connection table records participant linkage, construct mapping, sequence, and integration rules. Sequential designs let one result guide later sampling; parallel designs prespecify comparison. Remove methods with neither an independent question nor a connection role instead of collecting them for decorative triangulation.

Where it stops holding

More methods do not automatically improve validity and can increase cost, privacy exposure, and analytic flexibility. Coverage does not require one method on every side; a narrow question may need one. Conflicts are not resolved by majority vote but by checking operationalization, sample, and setting. Method diversity cannot replace appropriate causal or sampling design.

Applying it

• Split the problem into claims and tag each with its required evidence dimensions.
• State what each method answers independently and how it changes interpretation of another.
• Align objects, time windows, and construct definitions; preserve differences that cannot be aligned.
• Audit methods that contributed no inference or integration and reduce the next evidence set.

Related

• Same group: Q1.10.1 Methods vary across three dimensions · Q1.10.2 Attitudes and behavior can diverge · Q1.10.4 A qualitative–quantitative binary hides context
• Adjacent: Q1.03 Choosing qualitative and quantitative methods · Q3.11 Telemetry and instrumentation
• Search terms: method complementarity · mixed methods integration · triangulation protocol