Two Questions, One Answer
You ask: "How did you find the checkout flow, and did the shipping options make sense to you?" The participant nods, thinks for a second, and says: "Yeah, checkout was pretty smooth." You write it down and move on. What you did not notice is that half your question just vanished. The shipping options -- the thing you actually needed to learn about -- were never addressed, and because you got a fluent, on-topic answer to the first half, nothing flagged the gap.
The compound question collapse is the systematic loss that happens when you bundle two or more asks into a single question. Participants do not answer compound questions. They answer the part that is easiest to reach -- usually the last clause, the more concrete one, or the one with the more obvious response -- and let the rest quietly fall away. The transcript looks complete because you got words back. But the coverage is a fraction of what you think it is.
Why the Easier Half Always Wins
Working memory drops the tail. By the time you finish a two-part question, the participant is already composing a response to whichever part landed last or loudest. The first clause has to survive several seconds of your continued talking before they can act on it, and often it simply does not. This is a close cousin of the question length bias, where longer questions get shorter, emptier answers -- compound questions are long questions with the added penalty of competing targets.
Participants optimize for a valid-sounding reply. A user's implicit goal in an interview is to produce an answer that sounds responsive, not to maximize your data coverage. Answering one half fluently accomplishes that goal. The social contract is satisfied, the silence is filled, and the unanswered half creates no visible friction for either party.
The concrete clause beats the abstract one. If you ask "What did you think of the design, and how does it fit into your workflow?", the design question is easy to answer with a quick reaction while the workflow question requires real reconstruction of behavior. Participants reliably take the low-effort path -- the same dynamic behind the articulation gap, where people default to what is easy to say over what is true about their behavior.
You mistake fluency for completeness. A confident answer to half the question feels like a complete answer to the whole thing, because your brain registers "they responded" and releases the tension of the ask. The missing half does not announce itself. It just never comes up again.
The Hidden Cost
The damage is not that you get bad answers -- it is that you get half-answers you record as whole ones. Your coverage matrix says the shipping-options question was asked in twelve of twelve sessions. In reality it was answered in three. The other nine got absorbed into a checkout comment. When you synthesize, you treat the topic as thoroughly explored and conclude there were no problems, when the truth is you never actually asked in a way that got answered.
This produces false saturation. You believe a theme is settled because the question appears in every guide and every transcript, when the signal is far thinner than the paper trail suggests. Silence on a topic reads as absence of a problem, when it is really absence of a question that survived. The gap between what you think you measured and what you actually measured is exactly the kind of untraceable failure that mature engineering teams design against when they enforce structured, schema-constrained outputs so every required field must be explicitly filled rather than silently skipped. A well-formed answer to the wrong subset is still a hole.
Breaking the Collapse
Ask one thing. Then stop. The single most effective fix is mechanical: one question, then silence. Let the participant fully answer before you introduce the next target. This also opens room for the productive silence that pulls out the second, more honest layer of a response.
Catch yourself mid-question. Compound questions often form live, when you tack a second clause onto the first because the silence felt too long. Notice the word "and" arriving in your own question and cut it. Park the second half; ask it as its own question afterward.
Track answered, not asked. In your notes, mark whether each question was actually answered, not just whether you said it out loud. When you review, the unanswered halves become visible instead of hiding inside adjacent responses.
Return to the dropped half explicitly. If a participant answers only one part, come back: "That's helpful on checkout -- and specifically on the shipping options, walk me through what you expected to see." Re-asking the abandoned clause is not redundant. It is the only way it gets answered. Pushing past the first easy reply is the same discipline behind effective follow-up probing, up to the point where returns diminish.
The Discipline of One
The compound question feels efficient. You cover more ground per breath, the conversation flows, the participant seems engaged. But interview efficiency is measured in answers received, not questions emitted, and every bundled ask trades a real answer for the illusion of one. The researchers who get the deepest data are almost boring in their restraint: they ask one small thing, they wait, and they only move on once the answer is actually in the room.
Qualz.ai flags compound questions in your discussion guides before you run the study and surfaces which questions actually got answered versus merely asked across your sessions -- so half-answers stop masquerading as full coverage. Book a demo to see how your current guide holds up.
