Conversion rate optimisation: visitors but no enquiries

Traffic without enquiries is a demoralising problem, because every number on the dashboard goes the right way while nothing happens in the inbox. Somebody is arriving. Nobody is speaking.

The temptation is to answer it with a redesign, which usually relocates the fault rather than removing it. Treat this as a sequence instead: six checks, in order, each eliminating something. There is no point rewriting your homepage for buyers if two thirds of your visitors were never buyers.

Check one: is the traffic real, and is it the right traffic?

A visitor count is not a count of potential customers. Filter out your own address first, and your developer's: at a few hundred monthly visitors, your own team is not a rounding error.

One caution before you trust these numbers. A consent banner means your analytics is undercounting by an amount you cannot see. The EU rule is Article 5(3) of the ePrivacy Directive (2002/58/EC, as amended by Directive 2009/136/EC): reading or storing anything on a visitor's device needs consent unless it is strictly necessary for the service they asked for. What counts as an exception is where it stops being settled. France's regulator exempts tightly scoped audience measurement on conditions. The UK exempts more: since 5 February 2026, Schedule A1 of the Privacy and Electronic Communications (EC Directive) Regulations 2003, inserted by the Data (Use and Access) Act 2025, allows storage solely for statistics about how a service is used with a view to improving it, given clear information and a simple free way to object. That exception is narrower than it sounds. It covers only analytics whose data goes to nobody else, except to help you improve the site, and in their ordinary settings the common third party tools do not meet that condition. Other regulators still expect consent, and the ePrivacy Regulation meant to harmonise this was withdrawn by the European Commission in 2025, so get your consent setup right rather than guessing.

Search Console gives a second opinion on volume: it counts clicks on Google's side, before your page loads, where no banner can suppress them. The gap against your analytics is your blind spot.

Check two: does the page work on the device it arrives on?

Owners are the worst placed people to run this check. You open the site on a desktop, on a fast connection, in a browser that cached everything. Your visitor opens it on a four year old phone, on mobile data, cold.

Across 1,089 Dutch business websites and small online stores I ran the same technical check over, 77 (6.8%) did not load at all, and of the 1,033 that responded, 53% were slow to answer. A slow first response happens before a single pixel is drawn, so nothing you do to the page hides it.

Run your homepage and main service page through Google PageSpeed Insights and read the mobile numbers. Google's good thresholds for its three Core Web Vitals are Largest Contentful Paint within 2.5 seconds, Interaction to Next Paint of 200 milliseconds or less, and Cumulative Layout Shift of 0.1 or less. Interaction to Next Paint replaced First Input Delay in 2024, so ignore older advice about that one. Time to First Byte is not one of the three: it is what my audit measured.

Then do what no tool can. Take your phone off the office wifi and complete an enquiry with your thumbs, watching for the layout jumping as you read, a banner over the button, a field that zooms in and never back out. Any one ends the visit and none appear in a score. Where the server is at fault the answer is a hosting decision rather than a project, and technical work is not a rebuild.

Check three: can a stranger tell what you do, and who for?

Show your homepage to somebody who does not know your business, for five seconds, take it away, and ask: what does this company do, who is it for, and what would I do next if I wanted it. If they cannot answer all three, you have found your problem in under a minute. Use three or four people, and not the kind ones.

The most common failure is not ugliness, it is abstraction. The heading says something about partnership, or solutions, or being driven by results, the words a business reaches for once it has stopped thinking about the reader. The heading that works says the plain thing: what you do, for whom, where. It reads as unambitious to the owner and as a relief to everyone else.

Second most common is a site written for people who already know you. If every page assumes the reader has met you and knows your sector's vocabulary, it serves referrals well and strangers not at all. Where traffic lands on one page for one service, a page built around that one thing beats a homepage serving six audiences.

Check four: is there anything to respond to, and is responding cheap?

Count the fields on your form, then delete everything you do not need in order to reply. Company size, budget range, how they heard about you, job title: all askable once a conversation exists. Four will usually do. My own contact page asks five and I have argued with myself about the fifth.

Look next at what you demand. A required phone number turns a low commitment action into a high one, because the visitor is no longer sending a message, they are agreeing to be rung. Plenty of people who would happily type will not, and they do not tell you why. Making that field optional costs nothing.

Check whether an email address appears anywhere as readable text. A form is not a substitute: some people will not use forms at all, having sent things into one before and heard nothing back. There is a legal floor too. Article 5 of the e-Commerce Directive (2000/31/EC) requires member states to make anyone running a commercial online service keep their name, their address of establishment and contact details including an electronic mail address easily, directly and permanently accessible. It survives the Digital Services Act, Regulation (EU) 2022/2065, which deleted Articles 12 to 15 of that directive and left the information duties intact. The national versions are the ones that bind you: Article 3:15d of the Dutch Civil Code, in Germany section 5 of the Digitale-Dienste-Gesetz, which replaced section 5 of the Telemediengesetz on 14 May 2024, and in the UK regulation 6 of the Electronic Commerce (EC Directive) Regulations 2002. That is what an imprint page is for.

The email address on that list is settled; what the courts narrowed is the rest. In Case C-298/07 the Court of Justice held that a provider must offer, alongside the email address, a second route allowing rapid, direct and effective communication, though not necessarily a telephone number. An enquiry form counts, as long as somebody who has already written in and cannot use email is given a non-electronic route when they ask for one. Publishing no telephone number for calls is therefore defensible. My own second route is WhatsApp beside the email address, not a phone line. Publishing no way of reaching a human is not.

Last, and the one owners resist hardest: is there a price signal? Not a full list if your work genuinely varies, but something: a starting figure, a typical range, the smallest sensible project. A visitor with no idea whether you are a small fixed price job or a six figure programme has to risk looking foolish to find out, and many find it easier to ask somewhere that told them. Almost everything I charge is on a public page, with one exception the page names, which is not generosity but the fastest way to find out whether there is a fit.

Check five: do your website enquiries actually reach you?

This is the one I would put money on. It is the least glamorous cause and the most under-diagnosed, because it leaves no trace. The form submits. The page says thank you. Nothing arrives.

The reasons are mundane. A plugin update changed the address notifications go to. A mailbox was closed when somebody left. Or the form sends its notification pretending to be the visitor's own address, so the receiving provider sees a message from a domain that never authorised your server. If that is how yours is set up, change it to send from your own domain with the visitor as the reply address: providers weigh a domain's published authentication records, SPF, DKIM and DMARC, when deciding where a message goes.

I have some standing to nag here. The form on this site both stores every submission and sends it on, so a delivery failure cannot lose an enquiry. That is the minimum, not cleverness, and it is startling how rarely it is done.

How to test it properly

Filling in the form at your own desk, on your office network, only proves it works for you. What follows is about twenty minutes of your own time, spread over a day or two.

Submit the way a stranger would

Phone, on mobile data, not the office wifi. A personal address you genuinely check. Write a plausible message rather than the word “test”, because filters treat one word messages differently. Note the time, then watch the screen: a silent reload that just clears the form is the classic signature of a submission that went nowhere.

Check every folder, not just the inbox

Spam, junk, promotions, any quarantine your provider runs at server level, and any filing rule you set up years ago and forgot. Check again next morning: some failures are delayed, and you will otherwise call it a pass too early.

Reply to the notification

Hit reply and see where it goes. If the reply address is wrong, or a no-reply mailbox, every enquiry you answer goes into a void, which from the visitor's side is indistinguishable from being ignored.

Repeat with the banner declined and a blocker on

Reject the cookie banner, then submit. Turn on an ad blocker, then submit. Plenty of forms depend on a script that one or the other stops loading, so the button does nothing for a slice of your visitors while working perfectly for you.

Have somebody outside the business do it

Ask a friend at another company to send a real enquiry and tell you the timestamp. Their network, their device, their mail provider. This catches the failures specific to how the world outside reaches you, and it is worth starting early, since it waits on their day.

Then put it in the calendar: once a quarter, and whenever somebody updates the site, changes host, switches mail provider or touches the form. The gap between breaking and noticing is measured in months of silence.

Check six: are you asking for a commitment they are not ready to make?

If the previous five come back clean, this is where the problem lives. The site works, the visitor understands it, the route functions. They are not ready to do what you asked.

Most sites offer one action, and it is the biggest available: book a call, request a quote, get a proposal. For a visitor in the last week of a decision that is perfect. For one in the first week it is like being asked to marry on a first meeting, and they leave, because the only door was too heavy.

The fix is a smaller door beside the big one. Something concrete a stranger can accept without exposing themselves: a written assessment, a checklist, an answer to one question. It has to be useful whether or not they buy, and it must not be a sales call wearing a hat, because people can tell instantly.

What to do once you have the answer

Fix things in the order you found them, one at a time. Change six things in a week and you will never learn which mattered. Give each a fair run, too: on a site with modest traffic a fortnight of data tells you almost nothing.

Notice how much of this costs nothing. Filtering your own visits, reading your query report, trimming form fields, making a phone number optional, publishing an email address and a starting price, testing your own form: all free, all doable this week, and between them they cover most of the cases I see. Only two items here are real projects: better hosting, if the server is genuinely at fault, and a rebuild, if the page truly cannot say what you do. Both waste money when they are not the answer, which is the whole argument for diagnosing first.

Common questions