What is convenience sampling?

Where the bias comes from

Statistics Canada describes convenience or haphazard sampling as selecting units in an arbitrary manner with little or no planning, such as a vox pop where the interviewer stops anyone who happens to walk by. The trouble is that who happens to be reachable is rarely random. A survey on the street at 11 on a weekday misses most people at work; a poll on a brand's social page reaches its fans; a link shared by email reaches people like the sender. The Australian Bureau of Statistics adds that phone in polls and magazine questionnaires are subject to biased samples because only people who feel strongly about the topic respond.

Why online samples are often convenience samples

An open survey link that anyone can forward is a convenience sample, even with thousands of answers, because respondents choose themselves. Commercial opt in panels improve on this with quotas and weighting, but remain non probability samples. Pew Research Center gave an identical 56 item questionnaire to nine online non probability samples from eight vendors and compared them with government benchmarks; accuracy varied widely by vendor, and errors were largest for some demographic subgroups. Volume does not cure the problem: a larger convenience sample is a more precise estimate of a biased quantity.

When it is good enough

Convenience samples suit work where the goal is to discover rather than to estimate. Testing whether questions are understood, collecting ideas for a new menu, finding the main complaints in a service, or checking feedback after an event among the people who attended all work without a random sample. The ABS notes that non probability samples can be useful when pilot testing surveys. The safe habit is to describe the sample honestly, report counts rather than population percentages, and avoid phrases such as most Australians or customers prefer unless the design supports them.

Common mistakes

Reporting a margin of error for a convenience sample, which has no basis. Comparing two convenience samples collected in different ways and treating the gap as a real change. Over weighting a handful of people from an under represented group to make totals match the population, which magnifies their quirks. Ignoring who did not answer. And presenting results from an open social media poll as research.

Reaching the right people with a form

The form builder has no sampling or quota features: it collects whoever opens the link. What you can control is who can open it. Of the four access types, public lets anyone answer, secret key requires a key you share, login required limits the form to a list of allowed addresses or whole email domains, and domain restricted limits where it can be embedded. A login required form sent to a defined list is still only as representative as that list and its response rate. Ask the background questions you will compare against known figures as single choice questions, because each becomes its own column in the CSV export.

Questions people ask