Survey Speeders: How to Detect Them Without Losing Real Respondents
Speeding is the easiest bad-data signal to measure and the easiest to get wrong. How to set a threshold that catches rushers, why the timer has to live on the server, and what to do with the people you catch.
Why speed is the first thing to check
Every other quality signal takes some judgment. Is this open-text answer filler, or just short? Is this straight-line honest neutrality? Speed is different. If a 25-question survey took 50 seconds, the person did not read 25 questions. Arithmetic settles it.
It also tends to come first in a chain. People who speed through a survey are the same people who straight-line grids and type "good" in every text box. Catch the speed and you usually catch the rest of the response with it.
Three ways to set the threshold
A fixed floor per question
Multiply the number of questions by a minimum time each. Around 1.5 seconds per question is fast enough that almost no honest person drops below it, which makes it safe but lenient. It catches the worst rushers and scripts, and misses people who skim at 4 seconds a question.
A fraction of the median
Compare each response to the median completion time for the same survey. Under a third of the median is a common cut-off. This adapts to your survey's length and difficulty automatically, but it needs enough responses for the median to mean something. Thirty is a sensible minimum before you trust it. Qualtrics takes a similar statistical approach in its response quality feature, flagging durations far from the median.
Expected reading time
Count the words on each page, divide by a reading speed of about 200 words per minute, add a little time per answer, and compare. This is the most precise and the most work. It is worth it for long surveys with very uneven pages, where one page of instructions can distort a whole-survey median.
The timer has to live on the server
Most form tools time a response in the browser and send the duration along with the answers. That works for honest respondents. It does nothing against a script, which can post straight to the submit endpoint and claim it took 9 minutes.
The fix is to have the server note when the form was loaded, usually by issuing a signed token on page load, and measure the time itself when the response arrives. A response that arrives with no valid token did not come through the form at all, which is its own signal.
Who gets caught by mistake
Speed checks have predictable false positives. Plan for them instead of discovering them in your results.
- Returning respondents. Someone who started yesterday, closed the tab, and finished today looks slow. Someone whose draft was saved and resumed can look fast if your tool only times the last session.
- Experts on their own topic. A nurse answering a clinical survey reads the questions faster than a general panel would. Relative thresholds handle this better than fixed ones.
- Short forms. On a 4-question form, normal variation is a large share of total time. Relative thresholds get noisy here.
This is why speed should lower a quality score rather than delete a response. A fast response with clean, thoughtful open-text answers is probably a fast reader. A fast response that also straight-lined a grid is almost certainly a rusher.
Under-18 respondents
Timing a respondent is a form of behavioural analysis. Several laws, including India's DPDP Act and COPPA in the US for under-13s, put limits on monitoring children. If your survey might reach minors, check whether you can use timing checks at all.
How SiliForm handles speeders
SiliForm gives every public form a signed session when it loads and measures completion time on the server. On the Pro plan it applies both thresholds: a fixed floor of about 1.5 seconds per question, and once the form has 30 or more timed responses, a check against a third of the median of the latest 500. Saved drafts keep their original start time.
A speeder is not removed. The response gets a lower quality score and a reason ("Speeder" or "Fast vs median"), and lands in Review on its own or Flagged when other signals agree. Forms marked "may include respondents under 18" skip both timing checks entirely.
Speed is one of the signals behind SiliForm for researchers. Drop-off analytics shows the other side: the questions where people quit instead of rushing.
Frequently asked questions
What is a speeder in a survey?
A speeder is a respondent who finishes a survey much faster than it is possible to read and answer the questions. Speeding is the most common sign of low-effort responding, because a person who is not reading cannot give a considered answer.
What is a good speeder threshold?
There is no universal number. Common approaches are a fixed floor per question (around 1.5 seconds), a fraction of the median completion time (a third is a common cut-off), or an expected time based on word count at about 200 words per minute. Using two of them together is more reliable than any one.
Should I remove speeders before or after data collection ends?
Relative thresholds depend on the median, which moves as responses arrive, so final removal is safest after collection ends. You can still flag responses as they come in and review them at the end.
Does SiliForm flag speeders?
Yes, on the Pro plan. It measures completion time on the server from a signed session token, flags anything faster than about 1.5 seconds per question, and once a form has 30 or more timed responses also flags anything under a third of the median. On forms marked as possibly including under-18s, both timing checks are turned off.