Conversational form vs standard form

The format argument usually gets decided by whoever saw a nice form most recently. That is a shame, because the two formats are not competing versions of the same thing. They make opposite trades, and one of them quietly cannot collect a signature.

What each format changes for the respondent

A classic form is a page. Everything is visible, so somebody can read the whole ask before answering any of it, skip to the hard question, scroll back to check what they wrote, and see how much is left without being told.

A conversational form is a sequence. One question fills the screen, the next arrives when the current one is answered, and nothing else competes for attention. Each individual answer is cheaper, because there is only one thing to look at, and the total number of interactions is higher, because every answer costs a tap to advance.

That is the whole trade. Attention per question against control over the whole. Nielsen Norman Group describes the general version of this as progressive disclosure, and their caution is the relevant one here: staging works when the split matches what people actually need, and starts to hurt as soon as somebody has to move between levels to find something.

It has also been measured. A 2006 experiment published in Public Opinion Quarterly fielded the same instrument, up to 268 possible questions, to a sample of over 21,000 undergraduates, with about 10,000 responding and roughly a tenth randomly assigned a scrolling version rather than a paged one. The researchers compared break-offs, item missing data and the time spent in each section. It is worth knowing the question has been studied properly at that scale, and worth being sceptical of any vendor who answers it in one sentence.

Decision rule: length, field kinds, review need

Three questions, in this order.

How long is it? Under about ten questions, conversational reads well and feels lighter. Past twenty, it becomes a corridor, and people want the map that a classic page gives them. The 55 finished forms published on this site fall out along exactly that line: the conversational ones have a median of 9.5 fields, the classic ones a median of 18.

What kinds of field does it need? This is the one that ends the argument early. If the form needs a signature, a photo, or any of the twelve design blocks, it is classic. Those questions do not degrade gracefully in a conversational form, they render as an ordinary text box.

Does anyone need to review it before submitting? A job application, a grant submission or an expense claim gets read back before the button is pressed. A classic page supports that. A sequence does not, without walking backwards through every screen.

Field support table

The table at the foot of this post lists what each format actually supports, and it is worth reading the second column carefully, because the builder is more permissive than the published form.

The conversational toolkit offers 25 components and the classic one 36, but four of the conversational entries do not have a conversational renderer. Photo, inline signature, e-signature and branding all fall through to a generic text input, so the respondent gets a box reading “Type your answer here” where the signature pad should be. You can build that form. Nobody can fill it in.

The reverse gap is smaller but real. Recording, audio or video, up to 120 seconds by default, exists only in the conversational toolkit, as does the profile block that puts name, email, phone and address on one screen. Neither is offered for a classic form.

Two smaller differences catch people out. On a classic form, a field’s description is only rendered for file upload and photo; every other field drops it, so explanatory text has to be a paragraph block above the field. On a conversational form, a number field’s minimum and maximum are not passed to the input, so the range is only checked on submit, after every question has been answered.

Cases for conversational: surveys, RSVPs, lead capture

Conversational is the right call when the form is short, every question stands alone, and the respondent has no reason to review.

A satisfaction survey qualifies. So does an RSVP, where the questions are are you coming, for how many, and what will you eat. So does most lead capture, where the point is to make the first answer feel small. The shortest form in the example set here, a six-field newsletter signup, is conversational for exactly that reason, and only one of its six questions is required.

Ratings are the strongest case. A rating question answered against a page of twenty other ratings gets anchored to its neighbours. Given its own screen, it gets considered on its own terms.

Cases for classic: applications with uploads, anything printed, anything signed

Classic is the right call when any of these are true.

Something has to be signed, inline or by a countersigner. Something has to be uploaded as a photo through the photo component. The form will be printed, exported to a fillable PDF, or read back as a document, because the design blocks that make a form look like a document only exist in the classic toolkit. Or the form is simply long, and its sections are separate jobs, in which case the paginated classic layout gives you steps without giving up the page.

Applications sit squarely here. So does anything with a consent paragraph the respondent is expected to read before ticking, because a consent clause on its own screen with a single tick underneath is easier to click past than one sitting above the signature it authorises.

Themes and welcome screens

Both formats support a welcome screen and a thank-you screen, and both can carry an estimated time. The theme sets are separate: 12 classic themes and 25 conversational ones, and they do not overlap, because a theme for a full-screen single question is doing a different job to a theme for a page of fields.

The practical consequence is that switching a form’s format changes how it looks as well as how it behaves. Budget for a look at the result rather than assuming the design carries over.

The pair of examples

Two forms on this site ask about the same subject in opposite formats.

The 360 feedback form is conversational and ten questions long. Three of those are rating grids, and giving each grid its own screen is the entire argument for the format: a rater considers one dimension at a time instead of pattern-matching down a page.

The performance review form is classic and nineteen questions long. It uses the same rating matrix component, but it also carries two table inputs, a signature at the end, and a structure that a manager and an employee read together in a meeting. It could not be conversational, because the signature would not render and because the whole document has to be visible at once.

If the underlying problem is that the form is too long rather than the wrong shape, start with the field count data across all 55 examples instead. And if you want the plain definition of the term rather than the decision, that lives on the conversational form question page.