Adaptive surveys

What adapts#

• The next question — written from the previous answer, not selected from a preset branch
• The depth — a theme that keeps producing signal gets followed further
• The target — a question goes to the respondents whose answer would resolve the open hypothesis

There is no question tree to maintain, which is also why a new theme can be explored the day it appears.

Surveys are a last resort, not a default#

Poth collects only what the existing evidence cannot supply. A question is generated when two hypotheses survive and no data on hand separates them. That keeps the volume low, which is what keeps response rates alive.

Question quality#

Each generated question tests one specific explanation, which is what makes the answers usable:

• "When you say dirty, what specifically did you notice?" separates a vague impression from a concrete failure.
• "What time did you visit on Thursday?" tests a timing hypothesis without leading the respondent.
• "Have you noticed this issue before, or was this one-time?" separates a systemic pattern from an incident.

None of these are satisfaction ratings. They exist to move a hypothesis from open to settled.

Next#

• Hypothesis engine explains where each question comes from.
• Ask Poth a question covers the other direction, when you are the one asking.