Short answer: Ask for the one detail that lets you continue the conversation somewhere else, plus at most two that change what you would say when you get there. For most businesses that is a contact method, what they need, and roughly when. Everything else can wait until you are actually talking.
A visitor on your site is a stranger holding a door half open. You get one question through the gap, maybe three if you are charming. The instinct is to grab everything while they are there. The instinct is wrong, and it is the reason so many businesses have a contact form nobody fills in.
Start with the only question that matters
If you could ask exactly one thing, it would be: how do I reach you?
Everything else is optional because everything else can be discovered later. A name and an email address turn an anonymous session into a person you can follow up with. Without that, the most detailed enquiry in the world evaporates when the tab closes.
So build outward from the contact detail. Add a second question only if you can name the decision it changes. Add a third only if the second one earned its place.
The three that usually earn their place
What are you trying to get done?
Open, not multiple choice. People describe their own problem in language you can learn from, and that language is worth more than a dropdown. A visitor who writes "our old system keeps dropping calls at the weekend" has just told you the service, the pain, and the urgency in one sentence.
When do you need it?
Timing sorts your callback list better than almost anything else. It also lets you set expectations honestly. Someone planning for autumn does not need a call in ten minutes, and telling them so builds more trust than pretending everything is urgent.
Anything we should know before we call?
A catch-all that costs nothing and occasionally saves an entire cycle of back-and-forth. This is where you find out about the access issue, the fixed budget, or the fact that they have already spoken to your competitor.
Questions that seem useful and are not
Company size for a business that serves anyone. If the answer never changes what you do, it is data collection as ritual.
How did you hear about us? Useful, occasionally, but visitors guess. Track attribution properly instead, which we cover in how to track where your leads come from.
Phone and email and postcode and preferred contact time. Every additional required field is another reason to close the tab. If you need the postcode to quote, ask for it. If you need it because your CRM has a column, do not.
Budget, asked bluntly, on first contact. Most people genuinely do not know yet, and being asked makes them feel judged. If price bracket matters, offer ranges rather than a blank box, and phrase it as helping you point them at the right option.
Phrasing changes everything
Compare these two versions of the same question.
"Please provide details regarding the nature of your enquiry."
"What are you after? Rough sketch is fine."
Same information. Completely different response rate, because the second one signals that a half-answer is acceptable. Most people abandon forms not because they object to the question but because they are not sure they can answer it well enough. Give them permission to be vague and they will tell you more, not less.
Similarly, avoid stacking. A chat that asks one thing, waits, then asks the next thing feels like a conversation. A chat that fires four questions in a single message feels like a form wearing a costume.
When to switch to a form
Conversation is better for first contact, but there is a point where structure wins. If the visitor has clearly decided and just wants to hand over their details, making them type it out line by line is friction, not warmth.
This is why Clerkzo supports both. Leads can be captured conversationally, question by question, or presented as an in-chat form with your own custom fields once intent is obvious. The visitor who wants to talk gets to talk. The visitor who wants to fill in a box gets a box. Either way it lands in the same leads inbox with the same pipeline behind it.
Ask early enough to matter
Timing beats wording. There is no clever phrasing that rescues a question asked after someone has already reached for the back button. The advantage of an always-present chat widget is that the door is open through the whole visit rather than only at the end, which is the crux of the exit-intent versus always-on chat argument.
The other half of the timing question is what happens after. Capturing a contact detail at 11pm only helps if somebody acts on it before the visitor has booked with someone else. Notifications should reach a human, not a dashboard, which is what after-hours answering is built around.
A worked example
A domestic electrician, three questions, in this order:
- "What is going on? Even a rough description helps."
- "What is your postcode? Just checking we cover you."
- "Best number to reach you on?"
Notice the order. The problem first, because that is what the visitor came to talk about. The filter second, framed as a service to them rather than a screening test. The contact detail last, at the point where handing it over clearly leads somewhere.
Reverse that order and the same three questions perform noticeably worse, because you have asked for something before offering anything.
Frequently asked questions
Should I make any field required?
The contact method, yes. Everything else optional. A partial lead with an email address beats a perfect lead you never received.
What if visitors do not answer the qualifying questions?
Take the contact detail and stop. You can ask the rest on the call. Half a lead is still a lead, which is the argument in how to capture more leads from your website.
Is it rude to ask for a phone number?
Not if you say what you will use it for. "Best number for a quick call about this?" reads very differently from an unexplained required field.
How many questions before people drop off?
There is no universal number, but the pattern is consistent: drop-off climbs with each question that does not obviously benefit the visitor. Audit your questions by that test rather than by count.