Leo wants a website for his bakery. He has ten minutes before the next batch comes out of the oven.
The studio looks good. The work feels right. He clicks “Tell us about your project”.
Name. Email. Fine. Phone number. Maybe. Budget. He was hoping they could help with that. Launch date. Does “before Christmas, ideally” count? Three reference websites. Now he has to go shopping for websites before he can ask someone to make one.
He puts his phone down. There is dough to check.
The studio will never see an error message. Its form worked perfectly.
Leo is an imagined customer. The studies and documented retail examples below are real.
This is how a form can lose someone without being technically broken. It asks for more certainty, preparation or trust than the person has at that moment. To cut it well, you need to understand what each question is asking them to spend.
A small box can contain a large task
“Email address” and “Describe your growth strategy” can occupy the same amount of screen space. One asks for something you know. The other can open a meeting in your head.
Even a familiar question changes with context. “Your budget?” might mean checking a spreadsheet, getting a partner's approval, deciding what the work ought to cost, or worrying that the answer will become the price.
Those are different problems. A smaller input box solves none of them.
I know this. I can type it.
Is that in a spreadsheet? Who has it?
I haven't decided what this is worth yet.
Will answering mean a sales call?
There is research behind taking that last reaction seriously. In their 2007 review of sensitive survey questions, Roger Tourangeau and Ting Yan distinguish intrusiveness, concerns about disclosure and socially undesirable answers. Respondents may withhold or edit information because of what answering could expose.
The review concerns surveys, often on much more sensitive subjects than a project enquiry. It does not establish a conversion effect for your phone field. It does give us a better question than “How many seconds does this take to type?”: what does the person believe could happen after they answer?
In Baymard's checkout testing, some participants gave false phone numbers or abandoned when a required number came without an explanation. The lesson is practical: a completed field can still be bad data.
If a courier genuinely needs the number, explain that beside the field. If the person has requested a callback, the purpose is already clearer. If nobody needs to call them, let the number go.
Count the work, then count the boxes
The title says “in half” because a severe first edit is useful. Fifty per cent is not a research-backed optimum. Your form might need two questions, four, or considerably more.
Baymard's 2024 checkout benchmark counted an average of 11.3 fields and said most checkouts could work with eight. Its broader finding: the fields people must consider matter more to checkout usability than the number of steps alone.
So putting twelve questions on four screens does not remove eight questions. It may make a complex task easier to follow, but the work still needs a reason to exist.
There is a useful twist in a 2014 controlled study by Mirjam Seckler and colleagues. Sixty-five participants used original or improved versions of three newspaper registration forms. The improved versions were faster to complete and required fewer submission attempts.
But the researchers explicitly did not apply the guideline about removing unnecessary input. They lacked the business knowledge to decide what those companies needed.
Clear format instructions, distinguishable required fields and useful error messages helped. The study tested a package of improvements in a lab; it did not isolate a conversion lift from shortening a form.
That matters when you edit yours. Keep a useful question and make it answerable. Delete a pointless question and remove its burden entirely. Both can help; they solve different failures.
People also judge a task before starting. Galesic and Bosnjak's 2009 web-survey experiment varied the stated survey length between 10, 20 and 30 minutes. Longer stated lengths attracted fewer starts and completions. This concerns survey invitations, not a rule for four-field signup screens, but it helps explain why an intimidating promise of work can lose someone before the first answer.
Make every question earn its place
Imagine inviting everyone who requested a field into the same room. Sales wants a number. Marketing wants a source. Finance wants the company address. Design wants reference links.
Each request sounds reasonable alone. Together, they become Leo's afternoon.
GOV.UK calls its discipline a question protocol: establish why information is needed and what you will do with it before designing how to ask. The particularly useful part is forcing an answer beyond “someone might want this”.
For every field, write down the decision its answer enables. Then add the moment that decision must happen.
“We need their email to send the reply” has a clear action and timing. “We need their team size for our CRM” describes a storage destination. Keep asking what the information changes for the person or the service.
What becomes impossible if we ask this after the next useful step?
If you cannot name anything, the question has a strong case for moving. If nobody will act on the answer at all, delete it. “Later” should not become a warehouse for questions nobody wanted to defend.
An eight-field enquiry, cut to four
Let's give Leo a different entrance. For this fictional studio, the first job is to understand the request well enough to send a useful reply. It can discuss scheduling and scope afterwards.
Tell us everything.
- Full name
- Phone number
- What do you sell?
- What would you like help with?
- Project budget
- Preferred launch date
- Three reference websites
Start a conversation.
- Full name
- What do you sell?
- What would you like help with?
A few sentences is enough. We'll discuss scope and timing in our reply.
Phone: if you choose a call.
Budget: when discussing scope.
Date: when checking availability.
References: when shaping the direction.
The shorter version still asks for thinking. “What would you like help with?” can be intimidating if someone assumes a perfect brief is required. “A few sentences is enough” gives them permission to be unfinished. A studio should use that line only if it can actually work from a short description.
Now give the removed questions an appointment in the process.
| Question | Ask when… | It now helps to… |
|---|---|---|
| Phone numberOptional contact route | The person chooses a call. | Arrange the conversation they requested. |
| Budget rangeIf scope can be discussed first | You have outlined possible approaches. | Choose a scope they can afford. |
| Launch dateIf it does not determine initial eligibility | You check project availability. | Agree a realistic schedule. |
| Reference sitesOptional inspiration | You start discussing visual direction. | Talk about specific choices together. |
There are exceptions inside this example. If the studio has a firm minimum price, show it before the enquiry. If budget determines whether any help is possible, a simple range may belong upfront. If availability is the offer, the date is part of the first useful answer.
The test is whether moving the question helps the real conversation. Do not make someone explain their whole project before revealing that you cannot take it.
Real checkouts show two kinds of “later”
Sometimes a question belongs after a completed task. Sometimes it belongs behind a small, clearly labelled choice within the task.
John Lewis: buy first, consider an account afterwards
In Baymard's published example of John Lewis, the guest-checkout option tells shoppers they can create an account later. After checkout, they can create one or add the order to an existing account.
The timing preserves the purchase as the main task. An account can then offer a way to keep the details the shopper has already entered, rather than another decision before placing the order.
- 01 · During checkoutContinue as guest.
Know that an account can be created later.
- 02 · Purchase completeSee the confirmation.
The order stands on its own.
- 03 · Optional next stepSave the details?
Consider an account with a clear benefit.
Baymard reports usability observations, without a conversion-lift estimate. Finish the promised task before inviting someone into an additional relationship.
Best Buy: a field appears when the shopper needs it
Baymard also documents a Best Buy participant finding an option to add an apartment, suite or floor, then entering her unit number. The extra address field was available without making every shopper interpret it.
That is a different kind of timing. The address detail still belongs before delivery. It appears when relevant, rather than being postponed until after the order.
These examples come from the researchers' published observations. They illustrate two mechanisms: delay an optional task, or reveal a conditional question. Neither mechanism excuses hiding information someone needs to make a purchase.
“Later” needs an event, an owner and a reason
Imagine removing five signup fields, celebrating the cleaner screen, and immediately opening a modal asking for all five.
The entrance got shorter. The first session did not.
Give each deferred question a trigger that makes sense in the person's task. In a collaboration product, “Who should join?” becomes useful when they choose to share something. A delivery instruction becomes relevant when arranging delivery. A profile photo can wait until someone wants to be recognizable to other people.
These are design proposals, not measured product results. Their strength is that you can explain why the answer matters here.
Write the follow-up into the actual workflow. Who asks? At which event? What happens if the person skips it? If the answer is essential, how will you recover it before promising something you cannot deliver?
And be honest about what is optional. GOV.UK's question-page guidance recommends explicit optional labels and allowing uncertainty when it is a valid answer. If “I'm not sure” is a perfectly workable starting point, do not force a confident fiction into a required dropdown.
You may never collect some of the deferred information. That is a real trade-off. If the service genuinely needs it, plan for the follow-up. If only a dashboard wanted it, the missing cell may be a reasonable price for helping someone get started.
Some questions should stay at the entrance
A shorter form can waste more time if it delays the question that would have stopped an unsuitable request.
For a local repair service, a postcode can establish whether you cover the address. For an appointment, the date and service type can determine whether a booking exists. For a physical order, you need enough delivery information to make the delivery promise.
Ask those questions before the person invests in an impossible outcome. State relevant prices, restrictions and commitments before submission as well. Moving surprises later is a different operation from moving questions later.
There is also no prize for compressing six questions into a giant “Anything else?” box. That pushes the organising work onto the user. A few specific, answerable prompts can be kinder than one vast blank rectangle.
Check whether the whole job got easier
When a shorter form produces more submissions, it is tempting to declare victory. Look at what happened next.
Did people receive the promised result? Could the team act on the answers? Did the removed questions return as a week of email tennis? Did more suitable customers make it through?
Watch where someone has to leave the form
Ask a few prospective users to work through the real task. Look for searching, asking colleagues, uncertainty and concern. A pause is a clue to investigate, not proof that a field should be removed.
Trace one useful outcome from start to finish
For an enquiry, follow through to a useful reply or qualified conversation. For checkout, follow through to a completed order. Check the follow-up workload and answer quality alongside the first submission.
Compare equivalent journeys
If you run a controlled test, define success and quality checks beforehand. Compare similar traffic and track the journey beyond page one. A field-count reduction does not establish its own benefit.
Leo does not need to become better at writing project briefs. He needs an opening that lets him say: “I run a bakery. People keep asking whether they can order online. Can you help?”
That is enough to begin a useful conversation. The rest of the questions can arrive when there is a conversation worth having.
What the research supports
Usability experiments, survey research and checkout observations answer different questions. They support examining effort, clarity, sensitivity and timing. They do not establish a universal ideal field count or a guaranteed conversion gain from cutting a form in half.
Research and documented examples
- Seckler, Heinz, Bargas-Avila, Opwis & Tuch (2014). Designing Usable Web Forms: Empirical Evaluation of Web Form Improvement Guidelines. CHI, 1275–1284. Full paper consulted, including the explicit exclusion of the field-removal guideline and the laboratory limitations.
- Tourangeau & Yan (2007). Sensitive Questions in Surveys. Psychological Bulletin, 133(5), 859–883. Author-provided paper. Review of sensitivity, disclosure concerns and reporting; its application to enquiry forms is editorial interpretation.
- Galesic & Bosnjak (2009). Effects of Questionnaire Length on Participation and Indicators of Response Quality in a Web Survey. Public Opinion Quarterly, 73(2), 349–360. Publisher's abstract consulted for the stated-length manipulation and participation findings.
- Scott (2024), Baymard Institute. Checkout Optimization: 5 Ways to Minimize Form Fields in Checkout. Public benchmark summary and documented Best Buy address-field observation.
- Scott (2023), Baymard Institute. Save Account Creation for the Confirmation Step. Public usability findings and John Lewis example.
- Reeves (updated 2025), Baymard Institute. Checkout UX: Best Practices. Public findings on unexplained required phone numbers and false information.
- GOV.UK Service Manual. Structuring forms. Practical guidance on a question protocol and early eligibility checks.
- GOV.UK Design System. Question pages. Practical guidance on optional labels, purpose and valid uncertainty.
Sources checked 18 October 2026. Retail examples describe Baymard's published tests, not live interfaces inspected for this article. Only publicly available research material was consulted. The bakery story and studio forms are fictional; the figures are original illustrations.