The arithmetic, and why it beats buying traffic
Conversion rate is conversions divided by visitors. A site with 1,000 visitors a month and 20 enquiries converts at 2%.
The reason this is the first place to look is that the two levers are not equally priced. To double enquiries by buying traffic, you buy another 1,000 visitors, every month, for ever. To double them by conversion, you do a piece of work once and it keeps paying. The first is a subscription; the second is an asset.
| Route | What changes | Ongoing cost |
|---|---|---|
| Buy traffic | 1,000 → 2,000 visitors at 2% | Every month, for ever, rising with competition |
| Improve conversion | 1,000 visitors at 2% → 4% | Paid once. Compounds with every future visit. |
This is also why conversion work should come before an ad budget rather than after it. Sending paid traffic to a page that converts at 1% is buying an expensive demonstration of a problem you could have fixed first.
Getting the number right (this is most of the job)
Almost every business I start with is making decisions on a conversion rate that is wrong, usually in both directions at once. Four reasons, all of them common.
- The denominator is wrong. Consent banners mean analytics only counts visitors who accepted cookies. Depending on your audience, that can be well under half of them — so the true rate is lower than the dashboard says, sometimes by a lot.
- The numerator is incomplete. Form submissions get counted. Phone taps, WhatsApp clicks, live chats and people who read the page and rang the number from memory do not. For trade and service businesses this is frequently the majority of enquiries.
- Nothing is deduplicated. A returning visitor who enquires once but visited four times distorts a session-based rate.
- Bot and internal traffic is counted. Your own team, your developer, and a great deal of automated traffic all look like visitors who did not convert.
A practical test you can run today: add up last month's enquiries from every route — inbox, phone log, chat, forms — and divide by your analytics visitor count. If that number is materially different from the conversion rate in your dashboard, your dashboard is not measuring your business.
What is a good conversion rate?
I am going to answer this in a way that is less satisfying and more useful than a table of industry averages, because published CRO benchmarks are close to worthless: they aggregate businesses with different traffic sources, different definitions of a conversion, and different tracking accuracy, and they are quoted with a confidence nobody has earned.
Rough shape, from sites I run: a service business with a form as the main enquiry route typically sits somewhere between 1% and 5%. E-commerce is usually lower and varies enormously by traffic source, because a visitor from a branded search and a visitor from a cold social ad are not the same person. High-intent traffic converts several times better than discovery traffic, which is why a single blended figure hides more than it shows.
The benchmark that matters is your own site last month, measured the same way. That comparison is valid. A comparison against an industry average from an unnamed sample is not, and making decisions on it is how businesses end up rebuilding pages that were working.
What actually moves the number
In the order I usually find them, which is roughly the order of impact for small and medium UK businesses.
- Making the next action obvious on every screen. Not one call to action at the bottom of a long page — the enquiry route visible wherever the visitor happens to make up their mind.
- Removing fields. Every field on a form is a chance to leave. Ask for what you need to reply, not what you would like to know. Name, contact, and one sentence about the job beats an eleven-field qualification questionnaire every time.
- Speed and stability. A page that takes four seconds on a phone loses people before the offer is read, and a layout that jumps as it loads loses the tap. This is why the case studies on this site quote cumulative layout shift: zero layout shift means the button is where the thumb expected it. Pro Pave loads completely in 0.68 seconds with zero shift.
- Answering the objection on the page. Price, timescale, area covered, whether you do this kind of job. Most enquiry forms are competing with a visitor's unanswered question, and the question wins.
- Giving people the route they prefer. Some will never fill in a form and will always ring. Put the number where a thumb can reach it and count the taps.
- Legibility of the offer. Vehicles 2 Lease sells in a market where every competitor shows a wall of vehicles and prices. Restructuring the listings so the monthly figure and the terms behind it are readable at a glance was a conversion change as much as a design one.
What is usually not the answer: button colours, and a testing programme a small site does not have the traffic to run. A site with 1,000 visitors a month cannot reach statistical significance on a 15% improvement in any sensible timeframe. On that traffic you make changes based on evidence and judgement, and you measure the direction over months. Pretending otherwise sells a lot of software.
A worked example: Vehicles 2 Lease
Vehicles 2 Lease is an FCA-authorised vehicle leasing broker in Bury, Greater Manchester. Van leasing is a price-comparison market: every competitor shows the same vehicles at nearly the same prices, and nothing on any of the sites tells you why to pick one broker.
The work was not a redesign for its own sake. The listings were restructured around a clear pricing hierarchy so the monthly figure and the terms behind it are legible at a glance. The enquiry journey was rebuilt with fewer fields and an obvious next step, and the phone number was kept in reach throughout for the people who would always rather ring. The build was kept fast: 0.19 seconds to first byte, 0.89 seconds to first contentful paint, and zero layout shift, all measured on the live site on 4 September 2026.
The client reports form conversion moving from roughly 1% to roughly 4%.
One more piece of honesty about that engagement: Cybrial rebuilt the site — the previous one was dated and partly broken — and works on it monthly. No search performance figure for them is published anywhere on this site, because there is not yet a window long enough to report one honestly.
How to start, if you are doing it yourself
- Fix the tracking first. Every enquiry route as an event, internal traffic excluded. Do not skip this to get to the interesting part.
- Watch ten session recordings. Not a hundred — ten is enough to find the thing everyone is doing. It is almost always something structural and slightly embarrassing.
- Ask five recent customers what nearly stopped them. The answers are more useful than any heatmap and they cost you five phone calls.
- Fix the biggest drop-off, once, and leave it alone for a month.
- Compare against your own previous month, measured the same way. Not against an industry benchmark.
- Repeat. This is a habit, not a project, which is exactly why it fits a monthly block of hours rather than a one-off engagement.
