Conversion rate optimisation: five checks before you buy traffic

Conversion rate optimisation, CRO, is the work of getting more enquiries from the visitors you already have, before you pay for more. For a small business site it comes down to five checks: is the phone answered, is the offer on the first screen of a phone, is there one action per page, is the proof next to the ask, and is the form short. Each takes ten minutes to test. Do them before you buy a single click, because traffic sent to a page that leaks is money sent to a page that leaks.

What a leaking site costs before you buy traffic

A conversion rate is a plain fraction: visitors who enquire, divided by visitors. If 1,000 people visit and 10 enquire, the rate is 1%. The number matters because every pound you spend on traffic, ads, SEO, a leaflet with the web address on it, is multiplied by it.

Here is a sum with real numbers in it. When Bad Boy Labs built the site for Walk On The Wild Side, a pet shop in Tunbridge Wells, month one brought 1,500 visits and 39 enquiries, with 75% of those visits from Google. That is 39 divided by 1,500, or 2.6 enquiries per 100 visits. Treat it as one verified reference point, not a benchmark every site should hit.

Now the example. A trade site gets 1,000 visits a month. Put your own figure in from Google Analytics.

At 1 in 100 enquiring:    1,000 x 1%   = 10 enquiries
At 2.6 in 100 enquiring:  1,000 x 2.6% = 26 enquiries (the pet shop's month-one rate)
Same visitors. 16 more enquiries. Nothing spent on traffic.

What does that mean? The difference between a page that leaks and a page that works is more enquiries than most ad budgets would buy. How do you know? Google Analytics shows visits; your inbox and call log show enquiries; divide one by the other and you have your rate today. Why does it matter? Because if you buy traffic first, you pay to send more people to the leak. If you fix the leak first, every visitor you later pay for is worth more. On 13 October 2026 the advertised cost per click for "plumber kingston" was £15.12 (Google Ads UK, 2026 run-rate). At 1 in 100, that is £1,512 per enquiry. At 2.6 in 100, £581. Same click price. The page decided the rest.

What the search data says about CRO in 2026

The term is searched a great deal, mostly by marketers, and almost never locally. On 13 October 2026 the UK run-rates (Google Ads UK, Jan to Aug 2026) were:

Eleven thousand people a month search the idea. Eleven search it with "surrey" attached. "Website redesign" at 425 is the telling one: owners who suspect the site is not working reach for a redesign, which is the expensive version of the five checks below, done without first finding out which of the five is broken. An honest note: no Surrey town shows meaningful searches for CRO or landing page design. The checks are the same in Woking as in Leeds, and you do not need a local specialist to run them.

What are the five checks before you buy traffic?

Check 1: Is the phone answered? (10 minutes)

What it means: ring the number on your site, from a mobile, at 12.45pm, at 5.30pm and on Saturday morning. Count how many rings, and whether a person or a machine answers. How you know: this is the only conversion step that happens off the website, and it is the one most sites fail. A page can be perfect and the enquiry still dies on a ringing phone. Why it matters: a missed call is a visitor who did everything you asked and got nothing. The money is in what an AI website conversion rate actually depends on, and the phone is most of it. Pass: a person or a system that takes the details answers within five rings at all three times. Fail: voicemail at any of them.

Check 2: Is the offer on the first screen of a phone? (10 minutes)

What it means: open your homepage and your main service page on your own phone, and do not scroll. What is on the screen? How you know: if the first screen is a logo, a photo and "Welcome to", the visitor has to work to find out what you do and whether you do it where they are. Why it matters: most local visits are on phones, and the back button is one thumb away. Pass: without scrolling, the screen says what you do, where, and shows a phone button or a form. Fail: you have to scroll to find any of the three.

Check 3: Is there one action per page? (10 minutes)

What it means: on each main page, count the things you are asking the visitor to do. Call, email, fill a form, book online, WhatsApp, download a brochure, follow on Instagram, read the blog. How you know: list them. If the count is above two, the page is asking for nothing in particular. Why it matters: a visitor given eight choices usually takes the ninth, which is leaving. Pass: one clear action, repeated at the top and the bottom, with a second quieter route at most. Fail: three or more different asks competing on one screen.

Check 4: Is the proof next to the ask? (10 minutes)

What it means: find your best evidence, the Google rating, the review count, a named customer's words, an accreditation, and check where it sits relative to the phone button and the form. How you know: if the proof is on a separate reviews page or at the bottom of the homepage, it is not where the decision is made. Why it matters: a person decides at the button. The review that would have persuaded them is one tap away, and they do not tap. Pass: rating, review count and one real review visible within a thumb's scroll of the call button. Fail: proof on its own page.

Check 5: Is the form short? (10 minutes)

What it means: count the fields on your enquiry form, and time yourself filling it in on a phone. How you know: a form that asks for name, email, phone, address, postcode, job type, budget, preferred date and "how did you hear about us" is a job application. Why it matters: every field is a reason to stop, and on a phone each one costs more. The long version of this failure is in the five-minute booking form that loses customers. Pass: three fields, name, mobile, what is the problem, and under 30 seconds on a phone. Fail: more than five fields, or any field the visitor has to think about.

Score yourself out of five. Most small business sites we look at pass one or two. Nothing below needs to happen until you know your score.

What are the three ways to fix what the checks find?

Option 1: Fix the failures with whoever built your site

What it means: send the failed checks, with the pass condition for each, to your current web person. How you know it works: four of the five checks are an afternoon's edits: move the offer up, cut the asks to one, move the proof under the button, shorten the form. Check 1 is a phone decision, not a web one. Where it falls down: it fixes what you found, not what you did not. The checks are the biggest leaks, not all of them. And if the site was built as a brochure, moving things around may not be enough. Who it is right for: most small businesses, first. Do this before anything else.

Option 2: Buy a testing tool and run experiments yourself

What it means: a heatmap tool, software that shows where visitors tap and how far they scroll, and an A/B tool, software that shows two versions of a page and counts which gets more enquiries. How you know it works: it replaces opinion with measurement, and the tools are mature. Where it falls down: a test needs enough visitors to tell the difference. At 1,000 visits a month and a handful of enquiries, most tests will not reach a clear answer for months, and in the meantime someone has to build the variants and read the numbers. For a site under a few thousand visits a month, the five checks will find more than the tools will. Who it is right for: a business with several thousand visits a month, a person who will own the tool, and the five checks already passed.

Option 3: Rebuild the page as a measured conversion system

What it means: a page built from the enquiry backwards, with the offer first, one action, proof under the ask, a three-field form, the phone answered, and every call and form counted so you can see the rate move. This is what Bad Boy Labs means by a conversion-focused website. How you know it works: the pet shop figures above come from a site built this way, measured from day one. Where it falls down: it is a build, with a cost and a timeline, and it is the wrong first move for a site that fails the checks for reasons an afternoon of edits would fix. Who it is right for: a site that passes none or one of the five, or that was built as a brochure, or a business about to spend serious money on traffic and unwilling to pour it into a page it cannot measure.

Which one should you pick?

Bottom line: run the five checks today, fix the failures with your current web person this week, and only rebuild or buy testing tools if the score is one or zero, or the traffic spend you are about to make is bigger than the cost of the page it lands on.

The reason is the multiplication. Every pound of traffic is multiplied by the conversion rate. In the example above, the same 1,000 visits produced 10 or 26 enquiries depending on the page. Fixing the page first is the only move that makes every later pound worth more, and four of the five fixes cost an afternoon. The rebuild is the honest answer when the afternoon is not enough, and the testing tool is the honest answer only once there is enough traffic to test.

How to do it this week

1. Get your current rate (Monday, 15 minutes). Google Analytics: Reports, Acquisition, Traffic acquisition, set the date range to last month, read Users. Then count last month's enquiries from forms, emails and calls. Divide. Done looks like: one percentage written down.
2. Run check 1 (Monday, 10 minutes). Three calls, three times of day, from a mobile. Done looks like: pass or fail, with the number of rings.
3. Run checks 2 to 5 (Tuesday, 40 minutes). On your own phone, homepage and main service page. Done looks like: a score out of five, and a line on each fail saying what the pass condition is.
4. Fix the phone first (Tuesday). If check 1 failed, decide who answers at 12.45pm, 5.30pm and Saturday, or what system does. Done looks like: a repeat of the three calls passing.
5. Brief the edits (Wednesday). One email to your web person: "Move [offer line] and the call button into the first screen on mobile. Remove [list of extra asks]. Put the Google rating, review count and this review [paste] directly under the call button. Cut the form to name, mobile, what's the problem." Done looks like: a reply with a date.
6. Check the edits on a phone (when live). Repeat checks 2 to 5. Done looks like: a new score.
7. Measure for a month (next month). Same Analytics report, same enquiry count, same division. Done looks like: two percentages side by side, before and after.
8. Only then buy traffic (month after). Ads, SEO, whatever is next. Done looks like: a known rate the new visitors arrive into.

Mistakes we see

• Buying traffic first. The site converts at 1 in 100 and the owner doubles the ad budget. Twice the spend, same leak.
• Redesigning instead of checking. "Website redesign" gets 425 searches a month. Most of those sites needed four edits, not a new site.
• Testing with too few visitors. An A/B test on 800 visits a month will not tell you anything this year. Run the checks instead.
• Treating the phone as not the website's problem. It is the last step of every conversion. A ringing phone fails the site.
• Measuring nothing. Without the before-and-after rate, nobody can tell whether the edits worked, and the next decision is a guess too.

Do you want the five checks run for you?

If you send me your web address and last month's visit count, I will run the five checks on your site, give you the score with the pass condition for each fail, and a straight answer on whether an afternoon of edits or a rebuild is the honest fix, before you spend anything on traffic. Yes, and you will have it within two working days. No, and the checks above take under an hour on your own phone. Which is it? What a measured, conversion-focused site looks like is on the websites page.