What is an Intercept Survey?

An intercept survey stops a person in the course of an activity and asks them a few questions about it, on the spot. The method's origins are physical (mall intercepts, exit interviews at stores and events, street surveys), and its digital form is the website or app intercept: a brief survey triggered by arrival, behavior, time on page, or exit intent, presented within the experience. It overlaps heavily with the in-product survey, and the distinction in practice is scope: intercepts are typically used on websites and public-facing experiences with anonymous visitors, where the researcher knows nothing about who arrived, while in-product surveys usually target known users inside a product. What both share is timing: the answer comes from someone in the middle of the thing the question is about, which is the format's entire value and its entire limitation.

The Two Questions It Answers Best

Intent: what did you come here to do? The single most valuable question in web research, because analytics shows what visitors did and cannot show what they wanted. A one-question intercept on arrival ("What brings you here today?") with an open box, coded into categories, reveals the mismatch between the audience's goals and the page's design, and the visitors the site was never built for.

Outcome: did you get it, and if not, why? On exit, or after a task: "Did you find what you were looking for?" followed by "What stopped you?" for the nos. This turns an exit rate into a list of reasons, and it is the cheapest diagnostic a site owns.

Beyond those two, intercepts serve for quick satisfaction reads (CSAT), for profiling who visits (a demographic or role question), and for recruiting: an intercept that ends with "would you be willing to take part in a short study?" is how many teams find real visitors for recorded sessions.

Sampling and Timing

1. Sample a fraction, at random.
Showing the intercept to a random percentage of visitors, rather than everyone, keeps the interruption rare and the sample less skewed by who tolerates pop-ups; response rates still make it a self-selected sample, and the readout should say so.

2. Trigger with care.
Intent questions on arrival but after a moment (a few seconds, or a scroll); outcome questions on exit intent or after a meaningful action; never on top of a form, a purchase, or the first paint of a page.

3. Cap and suppress.
Once per visitor per period, suppressed after any response or dismissal.

4. Keep it to two or three questions.
An intercept is a favor asked of a stranger; the open intent question and one follow-up is the usual ceiling.

5. Run long enough for the mix.
Weekday and weekend, campaign and non-campaign traffic; a two-day intercept samples two days' visitors.

Analysis

Open intent answers are coded (content analysis with a codebook, increasingly machine-assisted and human-checked) into a small set of goals, then crossed against page, source, and outcome: which intents fail most, which sources bring visitors with which goals, which pages are being asked to serve intents they don't serve. The analysis is quick, and its findings route straight to design (the page for the unserved intent), to content (the language visitors use), and to deeper research (a recorded study in which screened visitors with the failing intent attempt it on the live site, a Ballpark task on the real page, which shows what the "no" answers only describe).

The Short Version

An intercept survey asks a few questions at the moment and place of an experience, and its two great uses are learning what people came to do and whether they managed it. Sample randomly, trigger gently, cap and suppress, keep it short, run it long enough, and code the open answers against outcomes. It is self-selected and shallow and it answers the one question analytics never can, which is why every site should run one and almost none run it well.