Progressive profiling

Progressive profiling answers a specific problem: every field added to a registration form costs sign-ups, and every field left out costs the product something it wanted to know. Rather than choose, the questions are spread out. Registration asks for the minimum that makes the product work, and each later visit asks for one more small thing, at a moment when the reason for asking is visible. The record fills up over weeks instead of in one sitting, and nobody was ever facing twelve empty fields.

Three things have to be true for each ask, or the pattern becomes a tax on returning. It has to be one question, not a section. It has to be skippable, and the skip has to be remembered, so the same question is not waiting again tomorrow. And it should buy something the reader can see: asking for a postcode next to the delivery estimate it improves reads as useful, while asking for it beside nothing reads as data collection, which it is. A reading of how complete the profile is helps, for the same reason steps left helps anywhere else: it turns an open-ended series of asks into a thing with an end.

The neighbouring word is first run experience, and the difference is where each one lives in time: a first run experience is bounded to the very first open and is then gone for good, while progressive profiling is the long tail of every later visit. They are often designed together, since what the first open decides not to ask is exactly what the later visits inherit.

The failure mode is easy to name and common to see. An ask that arrives on every visit, ignores the last refusal, or blocks the thing the reader came to do has stopped being progressive profiling and become nagging with a progress bar attached. The same applies at the other end: pairing this with lazy registration is the honest version, because the account is built out of things the reader actually did, but only while each question stays optional and each answer stays useful to the person giving it.