Short Surveys and On-Site Questions That People Answer

An on-site survey earns answers when it is short, timed to a natural moment, and specific enough that replying is easy. One question is the default; three is the ceiling. Ask about the visit the person is actually having: what stopped you completing this today, what were you hoping to find here, what would you need to know before booking? Offer a few written options drawn from real support language, plus an open box, because options make answering fast and the open box catches what you failed to imagine. Treat the results as leads to investigate, ranked by frequency, not as votes that decide. People are generous with a well-aimed question and ruthless with a questionnaire that interrupts their errand.

The same question converts or annoys depending on when it appears. On arrival it is an interruption before any value has been given, and it will be dismissed by exactly the engaged visitors you want to hear from. Natural moments are behavioural: after a search returns nothing, after a visitor has read a guide and is leaving, on an exit pattern from a product or pricing page, or after a defined depth of engagement on the page with the puzzle. Set the trigger from the behaviour your recordings show, covered in recordings and heatmaps, and cap frequency hard: a visitor who declined gets peace for a long while. Timing is most of survey craft; wording is the rest.

Write questions in the visitor's vocabulary, singular and concrete. What stopped you, not how would you rate your experience. Offer options you have genuinely heard from customers and support, in their phrasing, ordered without steering, and always include a way to say none of these. Avoid leading options that smuggle your hoped-for answer into the list, and avoid asking people to predict their own future behaviour, which they do charmingly and inaccurately. For the open box, a light prompt helps: the delivery detail, the missing size, the unanswered question, in a few words. Pilot the survey on staff from another context, or a friendly customer, and fix any question they have to read twice.

Survey replies over-represent the motivated: the annoyed, the delighted, and those with time. Read them as a ranking of reasons among respondents, not as shares of all visitors, and never convert them into percentages of your traffic in a report. Their strength is vocabulary and discovery: they name hesitations you had not listed and phrase them as visitors phrase them, which is gold for page copy and for the observation clauses of hypotheses. Cross-check the top reasons against recordings and step rates before spending on them. A reason that appears in replies, in recordings and at the funnel step it describes is as close to certain as small-business evidence gets.

On-site questions hear from everyone, including the leaving, which is their unique value: the buyer survey can never interview the person who did not buy. Their weakness is shallowness and interruption risk. The post-purchase survey hears only from buyers, at a calm moment, and can ask about the whole decision. Use on-site questions to explore a specific page puzzle and to catch non-buyers' reasons, and the post-purchase question for the standing ranking of near-misses across the business. Running both permanently is rarely needed; run the on-site question in campaigns of a few weeks aimed at one puzzle, and keep the post-purchase question as the always-on instrument.

Every survey creates an obligation. Collate replies on a rhythm, publish internally the top reasons with counts, attach each to an action, a fix, a test or an explicit decision to accept it, and record that routing in the same log as your changes. If the same reason tops the list for three cycles with no action, either act or retire the question, because asking without acting trains the business to treat visitor words as decoration. Where a fix follows from replies, say so in the follow-up work and watch the relevant step and the next survey batch together. The comparison in choosing behaviour tools places surveys among the other instruments; their distinctive output is language, so use it where language is written: pages, forms and offers.

Start with one page, one question, one trigger. Choose the page the funnel indicts and the moment recordings suggest. Write options from the last twenty support conversations about that page. Run for a few weeks or a set number of replies, whichever your traffic reaches first, with frequency capping on. Read the results in one sitting, rank, route, and file. That single cycle teaches the mechanics, the response quality your audience gives, and the reading discipline, at almost no cost. Only then consider a second question elsewhere. Survey programmes fail by sprawl, ten half-read questions across the site; they succeed by sequence, one answered question at a time feeding the weekly routine in from data to decision.