Survey completion rate
Reported as a health metric, and actually the single best predictor of whether your headline finding will survive contact with reality.
Also called Completion rate, Finish rate, Response completeness.
What it is
Completion rate is the share of people who started your survey and reached the end. It is treated as a measure of survey quality, and it is more usefully read as a measure of how much of your sample is systematically absent from every question after the drop-off point.
responses reaching the final question ÷ responses starting the survey
- starting
- Answered at least one substantive question, not merely loaded the page
- final question
- The last question on that respondent's path, since branching means paths differ
The part the formula leaves out
Size does not fix a systematic absence. A survey with two thousand responses and 40% completion among unhappy respondents is less trustworthy than one with four hundred responses and even completion, because whatever caused people to leave is frequently the thing being measured.
That makes completion a diagnostic rather than a score. The check is to compare who finished against who started, split by the variable your headline finding is about. If dropout correlates with the answer, the headline comparison is measuring dropout — and in the worked survey example, the tenure gap the survey was run to test disappears entirely once partial responses are included.
Which then forces a decision about partials, and the decision changes the result. Discarding them is the default in most survey tools and it silently removes the people most likely to be unhappy. Keeping them at the last answered question is the safer default, and either way the choice belongs in the methodology rather than in a setting nobody looked at.
How it is usually computed wrongly
01
Discarding partial responses by default
The people who abandon are not a random sample of the people who started, and dropping them biases every downstream comparison in the same direction.
Keep partials at the last answered question, report both versions of the headline, and state which you are quoting.
02
Reporting completion without reporting where people left
One rate for the whole survey hides the specific question that caused the drop, which is the actionable part.
Report drop-off per question. A cliff at one item is a broken question, and a gradual slope is a survey that is too long.
03
Counting page loads as starts
It inflates the denominator with people who never engaged and makes a fine survey look poor, which then triggers changes to a survey that was working.
Count a start as at least one substantive answer, and state the rule.
04
Comparing completion across surveys with different branching
Respondents on different paths see different numbers of questions, so the same person could be complete in one design and partial in another.
Define complete relative to each respondent's own path, and compare within a design rather than across designs.
What your file needs
- One row per response, including partial ones
- Question-level answers, so drop-off can be located
- A start timestamp, or a response order, if pace matters
Anything missing is reported as unavailable rather than substituted with something weaker computed on worse evidence.
Compute it on your own file
No account needed to start. You only pay when you like what you see.