The Speed option was added to the first question on Oct 4, 2026; earlier visitors did not see it. The wording of the questions did not change.
Research question. What do VPN.com visitors say they want from a VPN: what they would use it for, what matters most to them and what they would spend?
Sources. VPN.com's VPN finder quiz asks three questions: what the visitor would use a VPN for, what matters most and their monthly budget. A response is recorded when the first question is answered and updated with each later answer. Provider clicks come from VPN.com's affiliate click log, joined on the first-party visitor cookie that both surfaces read. A visitor's click-outs count as genuine only if that visitor left proof of having been on a vpn.com page (our page script ran, the browser named a vpn.com page as the referrer, or the visitor carried the landing-page cookie our site sets on a first pageview) and made no more than 20 click-outs in the study window. Visitors with no such proof, or with more clicks than that, are treated as automated and left out. Quiz responses have their own, simpler cleaning rules (see the table below). Test traffic is excluded. The shorter page quizzes are a second, separate source.
One visitor, one vote. The first question and drop-off use each visitor's first valid response (14,792 visitors). The second and third questions, every combination and every click-out figure use each visitor's first valid completed response (10,907 visitors). 9.3% of visitors took the quiz more than once; only the first counts.
Statistics. Shares carry 95% confidence intervals (Wilson). Lift is the share of a group choosing an answer divided by the share of all finishers choosing it. Each cell is tested against everyone else, with one false-discovery correction (Benjamini-Hochberg) across the combination tables, and each pair of questions is tested with a chi-square test and reported with Cramér's V, the effect size to read: with thousands of visitors almost any difference is statistically significant. Cells with n under 100 carry a dagger and are directional; counts under 10 are withheld. The confidence intervals describe sampling error only. They do not correct for self-selection (who chose to visit and to open the quiz) or for errors in how answers were recorded, so a narrow interval does not make a share representative of anyone beyond these visitors.
Conflict of interest. VPN.com earns an affiliate commission when visitors buy through our links, including some of the clicks this study measures. The study publishes only aggregated answers and click-out rates, and it ranks no provider. No provider paid for the study.
Ethics and consent. The quiz asks three questions and nothing else: no name, email address or free text. A response is linked to a first-party visitor cookie so that repeat takers and later provider clicks can be matched. That identifier is not published: the files contain aggregates only, with no visitor, session or response identifiers, no individual timestamps and every count under 10 withheld.
Reproducibility. The aggregates and the method are public in the dataset; the raw answers stay private to protect visitors' privacy. The analysis code that regenerates every number is not published; it is available on request to journalists and researchers at press@vpn.com.
The wording of the three questions did not change after launch; the answer options changed once, when Speed was added to the first. What changed around them: the descriptions under each option, a budget hint, translated homepages and, most recently, a homepage test. Answer labels are quoted as visitors saw them.
