Laws of Survey Design

"Respondents tend to agree with whatever you put in front of them."

When uncertain or disengaged, respondents default to agreement regardless of content. Rooted in satisficing — conserving cognitive energy by selecting the first socially acceptable answer. Research suggests 10–20% of respondents exhibit this systematically.

"People answer as the person they want to be, not the person they are."

Respondents give answers they believe will be viewed favourably by others — or by the imagined researcher. Particularly affects sensitive topics: usage frequency, effort invested, attitudes on contested subjects, and any self-evaluative question.

"When cognitive cost rises, response quality falls."

Respondents select the first 'good enough' answer rather than their most accurate one when fatigued or overloaded. The Tourangeau, Rips & Rasinski cognitive model identifies satisficing as a core failure mode in survey response. It worsens over survey length, and with matrix grid questions in particular.

"Some always go to the edges. Others never leave the middle."

Cultural and individual differences cause systematic avoidance of either extreme or middle scale options — independent of true attitude. This makes cross-segment and cross-cultural Likert comparisons unreliable without calibration.

"The container shapes what you think is in it."

Logically equivalent questions produce different responses when framed positively versus negatively. "90% uptime" and "10% downtime" trigger different evaluations of identical facts. Question stems carry implicit evaluative weight that respondents absorb before they've formed a view.

"The question already contains the answer you'll get."

Wording that implies a correct or expected answer primes respondents toward it before they've formed their own view. "How helpful was our support team?" assumes helpfulness. "Did you find the feature easy to use?" assumes engagement. The framing is the bias.

"What people say they would do and what they do are different data sets."

Responses to counterfactual or future-oriented questions ("Would you use this?", "Would you pay for that?") systematically overstate positive intent. Stated preference is not revealed preference. The gap is consistent and well-documented across product research, economics, and environmental valuation.

"One question. One thing. Always."

Questions that ask about two things simultaneously — "How easy and enjoyable was the experience?" — produce uninterpretable answers. A respondent who found it easy but tedious has no valid option. The data looks clean but is structurally ambiguous.

"Earlier questions rewrite the cognitive context for later ones."

Asking about overall satisfaction before specific features contaminates both readings — they are no longer independent measures. Prior questions prime cognitive frameworks that shape how subsequent questions are interpreted, not just answered.

"The first number seen becomes the floor or ceiling for every number after."

Numeric estimates are disproportionately influenced by the first value encountered. Scale endpoints define the range of what feels 'reasonable'. A scale that runs 0–10 elicits different distributions than one running 1–7 for the same underlying construct — both in absolute values and in discriminative power.

"First and last options are not equal to the ones in between."

In visual surveys, the first option benefits from primacy bias. In long or verbal lists, the last benefits from recency. Position systematically influences selection independent of meaning — a well-placed mediocre option will outperform a poorly-placed strong one.

"Every question borrows from a finite attention budget."

Response quality degrades over survey length. Questions later in a survey receive less thoughtful answers even when respondents are still completing it — this isn't dropout, it's silent quality decay. Complex matrix grids are disproportionately expensive and accelerate fatigue.

"Memory compresses time. Recent events feel older; vivid events feel recent."

Respondents systematically misplace events in time. Recent mundane events get pushed further back (forward telescoping); salient or emotional events get pulled closer (backward telescoping). Any recall-based question — frequency, last occurrence, duration — is affected.

"Respondents are detectives trying to figure out what you want."

When participants can infer the study's hypothesis or the researcher's desired outcome, they adjust their answers accordingly — either to please the researcher or to subvert them. Particularly problematic when the product under evaluation is obvious, and the survey context signals its origin.

"People remember the worst moment, the best moment, and the last one."

Evaluation of an experience is dominated by its emotional peak and its final moment — not a faithful average. Post-experience surveys therefore capture the memory of an experience, not the experience itself. The gap matters most for journeys with high emotional variance.