A website is not converting when the visits it gets don't turn into contacts, quote requests or orders at the rate you expect. That sentence contains an assumption almost nobody checks before spending money: that there are enough visits to be able to say so.
The practical difference is this. With 150 sessions in a month and zero enquiries, your site's real conversion rate could still be 2%, in other words in line with your sector. With 1,000 sessions and zero enquiries, it can't. In the first case rebuilding the site is a gamble, in the second it is a decision.
How many visits before "not converting" becomes a fact
There is a statistical rule that settles the question in one calculation, and it comes from clinical research. It is called the rule of three: if an event has never happened in n observations, the upper limit of the 95% confidence interval for its probability is 3/n. James Hanley and Abby Lippman-Hand formalised it in JAMA in 1983, to answer a question of the same shape: if nothing bad happened during a study, can you say the drug is safe?
Applied to your website: divide 3 by the number of sessions and you get the highest conversion rate still compatible with the zero you observed. The rule holds for samples above 30 observations, which on the web is always true.
| Sessions in a month, zero contacts | Real conversion rate still compatible | What you can conclude |
|---|---|---|
| 100 | up to 3.0% | Nothing: 3% is above the average for several sectors |
| 200 | up to 1.5% | Nothing |
| 300 | up to 1.0% | The site converts below 1%, if it converts at all |
| 600 | up to 0.5% | Below the average of every sector surveyed |
| 1,000 | up to 0.3% | Here the problem is real, and measured |
The consequence is awkward for anyone selling redesigns: most small business websites don't get enough traffic to know whether they convert. When that is the case, the work to be done isn't on the page, it is on the amount of demand you capture, and the two jobs have different costs and timescales. If you don't know how many sessions really arrive, the first step is to check your website traffic before looking at anything else.
How well a normal website converts, and who measured what
The benchmarks that circulate are quoted as if they measured the same thing. They don't, and the detail matters more than it seems, because what changes is not only the market but above all the object being measured.
| Survey | What it measures | Sample and market | Result |
|---|---|---|---|
| Ruler Analytics, 2026 | Conversions on company websites, 13 sectors | Over 110 million sessions and 5 million conversions, UK market | Average 5.13%. Travel 1.9%, retail and ecommerce 2.4%, professional services 6.1%, software 7.6%, legal 7.9% |
| Unbounce, Q4 2024 data | Dedicated landing pages, mostly with paid traffic | 41,000 landing pages, 464 million visitors, 57 million conversions, international sample | Median 6.6% |
A whole company website is not a landing page. It mixes job seekers, suppliers, existing customers looking for a phone number and the curious who arrived from a blog post. Comparing your site's 1.2% with Unbounce's 6.6% measures two different things. The useful comparison is with your sector's row in the Ruler survey, which for a UK business is also the right market.
One figure turns up again and again in articles on this topic, usually without a source: "53% leave if a site takes more than 3 seconds to load". It comes from a DoubleClick by Google study, cited by Google in its own documentation, and it concerns mobile visits abandoned, not contacts generated.
The conversions you see in GA4 are fewer than the real ones
Before blaming the page, it pays to know how blind the tool you are judging it with is. People who reject cookies on the banner aren't tracked, and Google only models their behaviour statistically above two specific thresholds: at least 1,000 events a day with analytics_storage set to denied for at least 7 days, and at least 1,000 daily users with consent granted on 7 of the previous 28 days (Google Analytics documentation). The vast majority of small business websites sit below those numbers, and below them modelling doesn't kick in: conversions from people who rejected cookies don't appear anywhere.
Then there is the part no tool sees: the phone call made after reading the pricing page, the WhatsApp message, the email typed by hand after copying the address from the footer. Search Console counts clicks from search and doesn't depend on cookie consent, so it remains the most honest yardstick for how many people really arrived. What Analytics shows and what it doesn't, and the difference between the two tools, is explained in the Google Analytics guide; which numbers are worth watching every month is covered in SEO KPIs.
A page problem or a demand problem?
The two diagnoses can be separated with an hour's work in Search Console. Export the queries for the last three months and split them into three groups: branded (they contain the company name), informational (how to, what is, difference between, guide), and commercial (price, quote, supplier, best, near me, a competitor's name).
If sessions come almost entirely from the informational group, a low conversion rate is the expected result, not a defect: someone searching "what is an LED wall" isn't buying today. If, on the other hand, you have sessions on commercial queries and no contacts, the problem is on the page, and that is where to work. Splitting by search intent is the step that makes everything else readable, and the commercial queries you don't cover yet can be found with keyword research on the data you already have.
An example from our own projects. For Macropix, a Milan-based LED screen manufacturer, the query that brings the most mature enquiries in Italy is "ledwall prezzo" (LED wall price), 720 searches a month: whoever types it is preparing a budget. The same company went from position 88 to position 2 for "monitor pubblicitario" (advertising display) and now holds a 25.55% share of voice on the LED wall cluster, ahead of Amazon's 18.08% (Semrush, July 2026). The full figures are in the case study on SEO for manufacturers. The point for you: the page that converts isn't the prettiest one, it is the one built on a query where the decision is already under way.
Where contacts get lost, in order of how much it costs to fix
1. The offer and who it is for. The line at the top of the page should say what you do, for whom, and what happens if someone gets in touch. "Innovative solutions for your business" doesn't say it. "We design and build custom LED screens for events and retail, quote within 48 hours" does. It costs an afternoon of rewriting and usually moves more than any technical fix.
2. Form friction. Every extra field is a request made before you have given anything. In B2B the opposite of the usual advice applies: a few generic fields produce leads the salesperson has to call back to understand what they are about, while the right fields (quantity, size, timing, location) produce fewer contacts that are already qualified. The choice depends on which of the two problems you have, and on a B2B website it is almost always the second.
3. Proof. Named case studies, numbers, photos of installations, references that can be checked. It is the part companies have in-house and don't publish, out of habit or confidentiality. One project page is worth more than ten adjectives: that is why we keep our projects public with the results in them.
4. Speed. It comes last because it costs more and returns less than the first three, but the data exists and it is solid. In the Milliseconds Make Millions study, carried out by Deloitte and 55 for Google on 37 European and American brands and more than 30 million sessions collected in late 2019, a 0.1-second improvement brought +8.4% conversions in retail and +21.6% more users reaching the form on lead generation sites. Before doing anything, though, you need to know which speed metrics Google actually measures, because half the work sold as optimisation doesn't touch any of them.
If traffic lands on a page built to convert rather than on a whole website, the order is the same but the margin is higher: it is worth reading how to structure a B2B landing page and how to balance conversion with landing page rankings.
The two moves that look obvious and are nearly always premature
The first is rebuilding the website. A rebuild changes dozens of variables at once, so the following month you won't know which one worked, and if traffic was low you will be left with the same doubt as before plus an invoice.
The second is A/B testing. It needs traffic that almost no company website has. With a starting rate of 2%, recognising an increase to 3% with reasonable confidence takes about 3,800 sessions per variant, 7,600 in total: the figure comes from the standard formula for comparing two proportions, at 5% significance and 80% power, and anyone can redo it. With 500 sessions a month that test lasts more than a year. Until you reach that threshold, you work on the things you can see with the naked eye, not on button colour variants.
Thirty days, in order
- Count the sessions for the last 90 days and apply the rule of three. If the number isn't enough, stop here: your problem is demand.
- Check that conversions are tracked and that the cookie banner isn't hiding half the data.
- Export the queries from Search Console and split them into branded, informational and commercial.
- Take the three pages that get the most commercial sessions and rewrite the first line of each.
- Add to the form the two fields your salespeople ask for on the first call anyway.
- Set a number to check at 90 days, not 30: SEO and content rewrites have known response times, and they aren't weeks. How to write that number so that it holds up is covered in the method for SEO goals.
If after the first step the answer is "there isn't enough traffic", the game shifts to how much demand you can capture: it is the work we do in lead generation projects, and for a small organisation it is best tackled one channel at a time, as in the approach described for small business marketing.
Frequently asked questions
It depends on which rate you want to rule out. With the rule of three, if not a single contact arrives in a month, the real rate could still be up to 3/n: with 100 sessions a 3% rate remains possible, with 300 sessions 1%, with 1,000 sessions 0.3%. Below 300 sessions a month, saying the site doesn't convert is an impression, not a measurement.
There is no number that suits everyone. The most relevant survey for the UK is Ruler Analytics, over 110 million sessions and 5 million conversions: an average of 5.13%, with 1.9% in travel, 2.4% in retail and ecommerce, 6.1% in professional services and 7.9% in legal. The most useful comparison is still with your own site three months earlier.
With the queries. Export the last three months of searches from Search Console and separate the informational ones from the commercial ones. If traffic comes almost entirely from informational searches, a low rate is the expected result and the work should go into pages that answer searches with price, quote or supplier in them. If you have commercial traffic and zero contacts, the problem is on the page: the first line, the offer, the form.
Rebuilding the site changes many variables at once and makes it impossible to tell what worked. If traffic is below a few hundred sessions a month, a rebuild won't even produce the data needed to judge it. Start with the two or three pages that get the most commercial traffic, rewrite them and measure at 90 days.
Yes, for two reasons. People who reject cookies aren't tracked, and Google Analytics only models their behaviour above high thresholds: at least 1,000 events a day with consent denied for 7 days and at least 1,000 daily users with consent granted on 7 of the previous 28 days. Below those thresholds it models nothing. On top of that, phone calls, messages and direct emails don't go through any form and don't appear in any report.
There is one number almost no company collects that is worth more than half the reports: the answer to "how did you find us?", asked of everyone who calls or writes in and noted next to the quote. In the projects where we have introduced it, the same thing comes up regularly: a share of enquiries the owner put down to word of mouth actually came from a page on the site read two weeks earlier, and didn't appear in any dashboard because the person then picked up the phone. Adding that line to your management system takes half a day, and in several cases it has shifted the decision from "let's rebuild the site" to "let's write three more pages like that one".
