What is a partial response in a survey?

Partial, break off and complete

Survey researchers draw a three way line. The AAPOR Standard Definitions, now in their tenth edition, describe completes and partials as a matter of the proportion of questions answered, with a break off falling below the partial threshold and treated as a kind of refusal. The document is firm that the thresholds must be defined before data collection, and it offers examples rather than a single rule: more than 80 percent of questions answered as a complete, and 50 to 80 percent as a partial, or an equivalent based on the questions that are essential to the study. It also notes that break offs are identified by the point where the survey was abandoned, while partials are identified by how much was answered.

Whether to analyse them

Keeping partials adds answers, but not evenly. The questions near the start of a survey pick up extra responses from people who later left, while the questions near the end come only from those who finished, so a table mixing the two compares slightly different groups. People who abandon also tend to differ from people who finish, often in time, interest or reading ease, which is why dropping them silently can tilt results too. The AAPOR guidance allows response rates to be reported with and without partials in the numerator. Doing both, and saying which figure a headline uses, is the honest route when the difference is large. Where partials are kept, flag them in their own column so every table can be run both ways, and never fill in answers for the questions a person did not reach without saying exactly how it was done.

Why people stop part way

Abandonment clusters at predictable points. A question that feels intrusive, such as income or health, placed before trust has been built. A required question somebody cannot answer truthfully, with no way to say not applicable. A sudden request for a document or upload they do not have to hand. A page that turns out much longer than the previous ones. And technical failures at the end, where a form rejects the submission for a reason the person cannot see. Reading the questions around each drop off point, when a tool records them, usually shows which of these is responsible within minutes.

Designing so fewer answers are lost

The best partial is one that never happens. State the real length before the first question, and keep it true. Move sensitive and effortful questions later, once people have invested, but keep anything that decides eligibility early so nobody fills in half a form they cannot use. Make questions optional unless the answer is genuinely needed, and add a not applicable choice where an item will not fit everyone. Test the form on a phone with a slow connection, since that is where unexpected errors appear. For long applications, a checklist of what to have ready does more than any save feature. Save features carry their own costs: a resume link sent by email can be forwarded, a saved draft holds personal information nobody chose to submit, and both need a rule for when the unfinished copy is deleted. Where a tool saves partials automatically, tell people on the first screen, since many assume nothing is kept until they press the final button.

Partial responses and this form builder

This form builder does not store partial responses. A response is created only when the form is submitted, so somebody who answers four screens of an eight screen conversational form and closes the tab leaves no record. The responses table and the CSV export therefore hold finished submissions only. If you need to know where people give up, watch a handful of people fill the form in. If the unanswered end of a long form is the problem, shorten it, or split it into a short first form that is worth submitting on its own and a second form for the detail.

Questions people ask