The average website conversion rate: channel benchmarks in 2026, and how to build your own

1.7% of sessions converted for new visitors, 2.9% for returning visitors (Contentsquare, 2026): the average website conversion rate is read by channel and by type of visitor, never as a single figure. This page is for the business owner or marketing director who wants to place their own rate.

It gives the benchmarks by channel with their denominator, explains why two published averages contradict each other without either being wrong, then how to build a benchmark of your own. What makes a rate comparable, and what makes it good, is the subject of the good conversion rate.

Benchmarks by channel and by type of visitor

The benchmarks below relate conversions to sessions, the same denominator as the GA4 “session conversion rate”. They are broken down by acquisition channel and by type of visitor, not by sector:

These lines are not compared with one another: a site fed by unpaid social and a site fed by paid search have no reason to show the same overall rate. Three reading rules follow:

• Read channel by channel: each line places an order of magnitude for one channel, which is worth more than a single all-channel average.
• Check the denominator: this reading counts sessions, as yours does; a reading built on unique visitors or on landing pages does not sit on the same line.
• Do not turn it into a target: a panel remains a panel, and its window carries its own seasonality; it tells you whether you are in the right order of magnitude, not what a channel has to reach.

What a conversion rate average measures

A conversion is the action tied to revenue, an order or a qualified contact request, and the rate relates it to the sessions of the period. Micro-conversions are tracked separately and do not enter the rate. If the formula escapes you, it is set out in the definition of a conversion rate. Here we start from the result: yours on one side, a published average on the other.

That average is not a property of the web. It is the result of a calculation made by someone, on something, and four decisions determine it entirely, not one of which is yours:

• The calculation base: sessions or users, the same performance produces two different numbers.
• The scope of channels: all the traffic, or only part of it.
• The panel of sites: those the publisher of the study observes, its customers or its users, never a sample of the web.
• The period: a quarter, a year, a high season or a low season.

Until these four elements are known, an average compares with nothing, and they are rarely published alongside it.

The scope of channels is the one that misleads most. The overall rate includes organic search as well as paid advertising: a distinction therefore has to be drawn between the SEO conversion rate and the rate for paid traffic, which obey neither the same volume nor the same intent on arrival.

Hence the principle that governs what follows: an average announced as “all sites” that measures nothing but paid traffic is not an all-channel average. The study did measure what it measured, but its label is wrong. Nothing in the figure flags it, and an accurate but badly named figure is no better than a wrong one: it has you decide on a comparison that does not exist.

Why two published averages do not say the same thing

Two serious studies, published in the same year, can announce averages twice as far apart as each other without either of them being mistaken. The disagreement is not about the measurement, but about what was measured. Going back through the four decisions one by one is almost always enough to explain it:

The calculation base deserves a clarification: a visitor who came back three times before buying counts three times in the denominator of a per-session rate, and only once in a per-user rate. They are not two versions of the same indicator, and no calculation lets you move from one to the other.

The practical consequence comes down to one question, to be asked before any comparison: what is this an average of? If the answer is not given, the comparison is not made. That refusal is the only way of not committing a budget on a gap that exists only in a calculation method.

Simple average, median, overall rate: three numbers for the same data

A misunderstanding hides under the word “average” itself: depending on how a set of rates is summarised, you get different numbers from the same data. Take five sites. Four show 1% conversion and receive 1,000 sessions each; the fifth shows 5% and receives 96,000 sessions. Three calculations circulate under the same name:

Three numbers, no calculation error, and a spread of one to nearly five on the same data. None is systematically preferable: the median resists extreme values but deliberately ignores volume. What matters is knowing which of the three you are reading: a study that does not write it down compares with nothing, because the gap between these calculations is often wider than the one you set out to comment on.

The sector is not enough to explain the gap

The first reflex, when a general average looks too far from your own, is to go and find the one for your sector. The reflex is a good one: the conversion rate varies widely by business, and if you sell cars you are better off referring to the rate for your sector than to an all-markets value.

But the sector explains only part of the gap. Within one sector, several factors make the result vary as much as sector membership does:

• The business model: direct sale, quote request or subscription.
• The maturity of the site: age, brand awareness, content history.
• The share of brand traffic: a visitor who searches for you by name converts more.
• The price level: the higher it is, the longer the decision takes.

A sector benchmark places you, it does not judge you. It too measures one precise quantity, which has to be known before anything is drawn from it: a lead, a sale and a converted session are not counted in the same way. The values by sector, their source and what they count are brought together in the conversion rate by business sector.

What the average basket and the margin change in the reading

Some KPIs are strongly correlated with the analysis of the site conversion rate. They are the indicators tied to the commercial value of the offer:

• The average basket: at an equal rate, a higher basket produces more value per session.
• Price and margin: lowering a price or thinning an offer raises the rate by lowering what it brings in.
• Retention: a customer who comes back is not worth the same as a customer you will never see again.

Two sites with the same rate are therefore not in the same place, and an average that adds them together distinguishes neither of them. For averages, the consequence is a rule of method: a gap against a published value is not commented on without knowing what a conversion is worth on your own site. How that value is calculated, and the conditions under which a rate is good, are dealt with in the pillar on the good conversion rate, linked above.

Building your own benchmark

It is important to gather data on your objectives as early as possible. Your history, on a constant definition, will serve to assess how far your conversion rate optimisations carry, and it is the only reference whose calculation method you know end to end. Building it takes four decisions, to be taken once and written down somewhere:

• The depth of history: twelve months at least, one full seasonal cycle, read on a rolling month-by-month basis. Any shorter, and it has you comparing a back-to-school month with a summer month and calling it a trend.
• The granularity: an overall benchmark is only good for communicating. The ones that support decisions are the benchmark by channel and the benchmark by entry page, the only levels where action is possible.
• The form: the overall rate, conversions divided by sessions over the period, is the only one of the three calculations that connects to revenue. A median places the series, it does not set a threshold.
• The review frequency: a deadline decided in advance and kept, and not a review every time the figure displeases. The reading cadence belongs to KPI tracking.

Next, freeze what holds the series together over time, the point on which most dashboards fail:

• The calculation base: sessions, from the first to the last month.
• The scope of the events counted: the day a new action is tagged as a conversion, the rate rises without anything having changed for the company. What counts as a conversion is set from the definition of conversion.
• The attribution model: a change of model redistributes conversions between channels.

These three elements are noted in the report itself, in one line: what is divided by what, what counts as a conversion, which attribution window. They must not stay in the memory of whoever built it.

When do you open a new series? In three cases:

• A change of offer: what is sold, or at what price, is no longer the same.
• A change of measurement: tool, tagging or definition of a conversion.
• A lasting change in the channel mix: the make-up of the traffic no longer resembles that of the series.

In those cases, the old series is not recalculated: a new one is opened and kept beside the first, with the switchover date noted. Outside those cases the benchmark does not move, even if the year has been a bad one: a benchmark adjusted for reassurance no longer measures anything.

The channel reading given above then comes as a complement, and in that order only: your history says whether you are progressing, your segments (channel, device, page type) say where, and the published benchmarks, on an identical denominator, say whether you are in the right order of magnitude.

Tracking the rate over time

How do you find out a site's conversion rate? By configuring analytics tools such as Google Analytics correctly, it is possible to follow the conversion rate over time, and to identify which pages convert best and which perform least well.

It is this second reading that has the most value, and it is the least used: a site rate does not say what to do; a ranking of pages by their gap against their own history does. Grouping these pages into families, and the order in which to work on them, belong to the site conversion rate.

Two rules avoid false signals. Conclude nothing about a page whose conversion volume is too small for the gap to be anything but chance. And check, before blaming a page, that its traffic source has not changed: a page that loses its brand traffic to a broad query sees its rate collapse without a word on it having moved.

Traffic from generative engines distorts the average

Benchmarks are shifting with the arrival of traffic from generative engines. That traffic does not show the same rates as classic organic traffic, and that alone makes an average published last year hard to compare with an average published this year:

• Benchmarks in motion: AI-referred traffic has its own line in the channel reading, with its own conversion level. A site that does not measure it separately sees its average move with the weight of this channel, and may credit its classic SEO with a result that does not belong to it.
• A variable effect by sector: informational queries (health, finance, tech) are covered more by AI Overviews, while local transactional queries are still little affected.
• A new ROI calculation: including AI traffic gives a fairer view of content ROI, an article ranking well in SEO and cited by AI systems generating value on two channels.

An average that aggregates a fast-growing channel whose conversion differs from the others becomes illegible within a few quarters. Separating this segment makes it possible to know whether your benchmark is rising because your pages convince better, or because the composition of your traffic has changed.

Continue reading

Depending on what you want to decide next:

• You sell to businesses: the all-sites average does not concern you, and your cycle pushes the conversion several weeks past the click. The B2B benchmarks by sector are in the average B2B conversion rate.
• You only have an overall rate: your average aggregates channels, to be broken down by channel, entry page, device and campaign before comparing. The breakdown and the reading cadence are in the conversion rate KPIs.
• Your benchmark is set and the gap identified: what remains is choosing which page to start with and how to measure the effect. The decision window and the order of the work belong to optimizing conversion rate.

To compare your conversion rates by channel, organic, paid and generative, and build your own benchmark instead of borrowing one, the performance reporting module brings these cuts together in a single dashboard.

Frequently asked questions (FAQ)

What is a good conversion rate for a website?

A website converts 1.7% of new visitors' sessions and 2.9% of returning visitors' sessions (Contentsquare, 2026). These benchmarks place an order of magnitude, on an identical denominator, without setting a target. A good rate is one that progresses against your own history, read by channel, and that stays profitable in relation to your acquisition cost.

How is the conversion rate of a website calculated?

A site's conversion rate is calculated by dividing the number of conversions by the number of sessions, then multiplying by 100: 200 conversions for 10,000 sessions give 2%. To track it over time, keep this calculation base unchanged and segment by channel (SEO, SEA, GEO), by device and by landing page.

What average conversion rate should you aim for in B2B?

In B2B, the rate is measured in leads rather than purchases: demo request, form, meeting booked. The benchmark figure is read on a B2B reading by sector, never on an all-markets average, because a long cycle pushes the conversion well past the click. “Solution” and “pricing” pages generally convert more than informational content.

How can the average conversion rate of a site be improved?

The lever that falls to content is the quality of incoming traffic: sharper targeting, a title promise aligned with what the page delivers, dedicated pages by search intent. The other levers have other owners: speed belongs to development, the mobile journey and payment to the product team, A/B tests to UX.

Does the average conversion rate vary by device?

Yes, and that is one more reason not to compare two averages taken across all devices. The share of global web traffic coming from mobile reaches 60% (Webnyxt, 2026), but this mix varies from one site to another. Two sites with different device splits will therefore show different rates at equal performance: segment by device before concluding.

Why can a site's average conversion rate fall while traffic is rising?

The rate falls when the quality of the traffic deteriorates: more informational keywords, broader campaigns, less qualified audiences. The fall can also come from a change of offer, or from points that belong to product and development, such as loading time or forms. To diagnose it, segment the rate by channel, intent, device and entry page.

What is the difference between a micro-conversion and a macro-conversion?

The macro-conversion is the objective tied to revenue: purchase, quote request, meeting booked. Micro-conversions are intermediate steps, such as an add to basket or a download: they are tracked separately and never enter the rate. A published average does not always say which ones it counts, which is enough to make it incomparable.