Conversational forms — one question per screen, advancing as you answer — get talked about as though they are strictly better. They aren't. They are better at some things and measurably worse at others, and which you need depends almost entirely on why the person is filling in your form.
This is not a case where one side is right. It is a case where the two options fail in opposite directions, and knowing which failure you can afford is the whole decision.
What each one actually does
A conversational form hides length. You cannot see how much is left, each individual step looks trivial, and the rhythm of answer-advance-answer builds a small momentum. It also controls the order absolutely: you see question four only after answering three.
A single-page form reveals length. The whole thing is there, you can scan it before starting, you can answer out of order, you can leave a hard question and come back. It asks for a decision up front — is this worth my time — and then mostly gets out of the way.
Everything else follows from that difference.
Conversational wins when people want to be there
A quiz. A calculator. A personality test. An application for something the person actually wants. In all of these the respondent has intrinsic motivation, and hiding the length converts that motivation into completions: each answer is a small commitment that makes the next one more likely.
It also wins decisively when there is branching. A single-page form with half its questions conditionally hidden is visually unstable — things appear and disappear as you answer, the scroll position jumps, and the page feels like it is arguing with you. One question at a time makes branching invisible, which is how it should be.
And it wins on mobile more often than people expect. One question, one input, one large button is a much better phone experience than a long scroll with a keyboard covering half the screen.
Hiding the length helps when someone wants to finish. It hurts when they are deciding whether to start.
Single page wins when people are being made to be there
An expense claim. A support request. A mandatory registration. A compliance form. Here the respondent has extrinsic motivation — they are doing this because they have to — and the thing they want most is to know how long it will take.
Hiding that from them is actively hostile. A conversational form that turns out to be thirty questions long feels like a trick, and the person who discovers this at question twelve is angrier than if you had shown them thirty questions at the start.
It also wins whenever people need to see later questions to answer earlier ones, whenever they will want to gather information before starting, and whenever someone might reasonably want to fill it in out of order — which covers most forms that involve looking something up.
The things people believe that aren't reliably true
"Conversational forms convert better." Sometimes substantially, sometimes not at all, and on unmotivated audiences they are often worse. The studies people cite for this are mostly vendor case studies on marketing forms, which is exactly the case where conversational does well.
"Single-page forms are faster to fill in." Usually true for short forms and often false for long ones, because scrolling and re-finding your place has a cost that people underestimate.
"Progress bars fix the length problem." They help, and they are not the same as seeing the form. A progress bar tells you how far through you are; it does not tell you whether question nine is going to be difficult.
Just test it
This is a two-minute experiment if your builder renders both modes from the same form — no rebuilding, no second version to keep in sync, no risk of the two drifting apart. Ours does it with a radio button, and I would say that whoever built it.
Publish one mode. Wait a week, or until you have a couple of hundred starts, whichever is later. Switch. Wait the same again. Compare completion rate, and while you are there compare per-question drop-off, because the two modes often fail at different questions and that tells you something on its own.
Run it for whole weeks. Form traffic is strongly weekly — a B2B form on a Tuesday is a different audience from the same form on a Sunday — and a five-day test comparing a Monday-to-Friday against a Thursday-to-Monday will tell you about the days, not the design.
Everything above is a prior, not a finding. It is what to expect before you have data. Your audience is the finding, and getting it costs you two weeks and no engineering at all.
What to measure besides completion rate
Completion rate is the headline, but on its own it can mislead you in a specific way: a mode change can raise completions while lowering the quality of what you collect.
Watch the open-text answers. Conversational forms reliably produce longer, more considered free text — one question on a screen with nothing else competing for attention is a good place to write a paragraph. If your form's value is in one open question, that difference can matter more than a few points of completion.
Watch time-to-complete as well, but interpret it carefully. Conversational forms usually take longer in wall-clock terms and feel shorter, which is the whole trick. A longer median time with a higher completion rate is a win, not a regression.
And watch where the drop-off moves. Switching modes rarely changes the total by much without also changing which question loses people. If a wall appears in one mode and not the other, that is information about the question, not just the layout.
You can also mix them
The choice is not actually binary, and treating it as one is how people end up with a thirty-question conversational form.
A single-page form split into three steps with page breaks gets you most of the benefit of both: each step is short enough not to intimidate, the step indicator tells people how much is left, and within a step they can still scan and answer out of order.
This is usually the right answer for anything long and obligatory — applications, onboarding, intake. It is almost never the right answer for a quiz, where the momentum of one-at-a-time is the point.
The rule of thumb I would use: under six questions, single page. Six to fifteen with branching, conversational. Six to fifteen without branching, either — test it. Over fifteen, steps, whichever mode you pick.