"What's a good conversion rate?" The question comes up in almost every conversation about a small business website, and the answer you find online fits on one line: 2 to 3%. The number is easy to remember and close to useless for a service business. It comes mostly from e-commerce, and when you trace it back, nobody is at the other end.
The real problem sits elsewhere. A conversion rate is a fraction, and most small business sites get both halves wrong: they count actions that aren't conversions, and divide by visits that will never buy anything. The benchmark matters far less than the way you count.
The short answer, for the busy
There is no reliable benchmark conversion rate for a Quebec service business website. The "2 to 3%" that circulates comes mostly from e-commerce, and its source dissolves: Canada's BDC cites 2.5 to 3% from Shopify, whose French-language guide credits the range to unnamed "industry leaders." The figures published with a sample disagree because they count different things: a 6.6% median across 41,000 landing pages according to Unbounce (late 2024), and 6.1% qualified leads or sales for professional services among Ruler Analytics customers (2026), more than half of them arriving by phone. The useful number is your own, measured honestly. On top, real requests: forms the server accepted, calls, bookings, with spam and job applications removed. Underneath, prospect visits, with your own team excluded and blog readers set apart. Then aim to beat your own average from the last three months.
Where "2 to 3%" comes from
In Canada, one of the first pages Google shows for the French query "taux de conversion site web" is BDC's glossary entry. It gives a 2.5 to 3% range for e-commerce, "according to e-commerce platform Shopify." Shopify's French guide attributes that same range to "industry leaders," with no study and no date, then cites a measurement a paragraph later from Littledata, an analytics app for Shopify stores, that found a 1.4% average. Its English guide has moved on to other sources: Statista, reporting 1.4% of global e-commerce visits converting in Q1 2026, and Dynamic Yield at 2.66%. The Crown corporation points to a vendor, and the vendor's numbers change from one page to the next.
The other figure you run into, 2.35%, gets presented by a well-ranked agency article as "the average website conversion rate." It traces back to a 2014 WordStream post about landing pages in Google Ads campaigns (thousands of accounts, $3 billion in annual spend): paid traffic, sent to pages built for a single action.
Serious measurements exist, and they land elsewhere. Unbounce analyzed 41,000 landing pages in Q4 2024, with 464 million visitors and 57 million conversions: a 6.6% median. Ruler Analytics, an attribution software vendor, publishes its customers' rates every year. The May 2026 edition covers more than 110 million sessions and 5 million conversions across 13 industries, for an average of 5.13% and 6.1% in professional services. Ruler counts a conversion as a qualified lead or a sale, by form or by phone.
Each study measures a different fraction. A landing page receives visitors sent on purpose for one action; a whole website also gets the curious, existing customers and job seekers. And both solid studies come from vendors whose customers, by definition, measure carefully. For the website of a Quebec service business, we found no published study with its method and sample.
Define the conversion for a service business
An online store has an obvious conversion: the order. A service business closes its sales somewhere else, on the phone or in a meeting. Its website doesn't sell; it produces requests.
So the primary conversion is a request someone on your team can act on: a contact or quote form, a call, a booked appointment. Everything else gets counted separately: newsletter sign-ups, guide downloads, clicks on "Contact us." BDC recommends calculating a rate per action. Its own example adds 30 purchases, 10 catalogue downloads and 5 forms to show a 30% overall rate, a figure that, in the words of the adviser it quotes, "doesn't tell the whole story."
A simple test: if the action doesn't lead to a conversation with a potential customer, it isn't your primary conversion.
Count what the site actually produces
The top of the fraction inflates easily, and almost always by accident.
First trap: counting the intent instead of the result. Fathom's documentation shows how to fire an event at the moment a form is submitted. That's simple, and it also counts submissions the server then rejects, filtered spam or errors. On our own site, the event only fires after the server responds successfully, which means after the spam check.
Second trap: spam that gets through. A poorly protected form produces conversions that are bots, and if you count submissions, you count them too. The invisible defenses described in our article on web forms and spam clean up the number and the inbox together.
Third trap: the phone. A click on a phone number only measures intent. And that's often where the sale starts: in Ruler's data, calls make up 52.6% of professional services conversions. A site that counts only its forms can miss half of what it produces. The simplest fix is a phone number reserved for the website, which counts calls without knowing anything about visitors. Same logic for bookings: count confirmed appointments in the calendar tool, not views of the booking page.
Last trap: testing. Fathom warns that events are also recorded on local and staging environments. Block them, or every test by your team becomes a conversion.
All of this can be measured without cookies, so without a banner and without lost data, as we explained in our article on cookieless analytics. But the honest numerator gets checked by hand. Once a month, reread the requests you received and sort them: prospect, existing customer, job application, vendor pitch, spam. Only the first column is your conversion.
Clean up the denominator
The bottom of the fraction raises a question few people settle: a conversion out of what? The tools don't agree. By default, Fathom divides events by pageviews; a site setting lets you divide by visitors instead. Plausible divides unique conversions by unique visitors. The same site with the same requests shows different rates depending on the tool and the setting. Before comparing your number to anything, check the formula.
Fathom and Plausible already drop known bots and data center traffic at ingestion, but Plausible admits no filter is perfect: be wary of a sudden spike where nearly every visit leaves immediately. Your own team, meanwhile, isn't filtered by anyone. Exclude your office IP address, which Fathom lets you do in the site's firewall settings.
What remains is human traffic that will never buy: job seekers reading your careers page, existing customers looking up your phone number, readers who arrived through a blog post. Jakob Nielsen of the Nielsen Norman Group describes a humorous article, published April 1, 2013, that quadrupled his site's traffic that day without changing the number of sales. The day's conversion rate fell to almost a quarter of its usual level. The site was as good as the day before; only the denominator had changed.
The fix is two moves. Calculate the rate on the pages that sell (home, services, contact) by filtering your dashboard on those pages, rather than on the whole site. And track the absolute number of requests next to the rate, as Nielsen advises: if traffic jumps and requests stay flat, the new visitors aren't your customers, and that's fine.
Small numbers, false trends
A small business site produces few requests, and small numbers lie. Take a made-up example: 1,500 prospect visits a month and 12 requests, or 0.8%. The next month, 15 requests: 1%. Nobody touched the site, and the rate just rose by a quarter. Three extra requests can be pure chance.
Nielsen recommends a month as the measurement period, and warns that on a low-volume site, a handful of conversions per period makes the rate bounce for no real reason. Look at three-month averages, and only judge a change once enough requests have come in. For Nielsen, a good conversion rate is simply one that's higher than what you had before.
What actually moves the number
Once the measurement is clean, few changes matter.
The homepage first: one primary action, visible on the first screen. We laid out that structure in the homepage that converts, and the classic leaks (slow pages, a missing call to action) in five mistakes that hurt your conversions.
The form next. According to the Nielsen Norman Group, removing a question from a form is a very safe bet to raise its completion rate. Every "interesting to know" field has a cost, and a visible anti-bot puzzle adds another.
Traffic last. The conversion rate sets the real cost of a request: at a given cost per click, a site that converts twice as well pays half as much for each request. That's the arithmetic to run before choosing between SEO and Google Ads.
So what number should you aim for?
Not another industry's. Measure three months with a strict definition and a cleaned-up denominator: that number is your starting point, and the goal is to beat it. Track the number of qualified requests next to it. A rate that rises while requests fall is not a win.
That's the measurement we run on our own site and propose when we build or rebuild yours: events that only count accepted requests, and a first reading of the number together.
→ Let's talk about your website and what it actually produces
Written byXavier Peich