Conversion Rate Benchmarks by Industry and Funnel Stage

A.The ranges, and what actually generates them

Conversion rates depend on variables that differ enormously between two businesses in the same industry. Traffic source mix is the largest of them. A site taking 80% of its traffic from branded search converts several times higher than one taking 80% from cold social, with an identical product and an identical page. Price point matters almost as much: a product at $9 a month converts higher than one at $900, all else equal. Geography, device mix, brand awareness and product complexity follow.

Then there is the definitional layer, which is invisible in every published table. Some teams report conversions over sessions and others over unique visitors, and the two can differ by 20-30% on the same data. Some count a lead as a form submission, others as a qualified lead. A rate at the top of its range may be strength, or it may be a looser definition than the benchmark used.

B.Which leaves one valid reading

Because source mix dominates, a benchmark comparison is mostly a comparison of audiences rather than of pages. Visitor-to-trial for SaaS illustrates the size of the effect cleanly: direct and branded search convert in the 8-12% band while cold paid social can sit at 0.5-1%. Both figures come from sites that would report a blended rate inside the published 2-5% range, and the blend tells you only what their media buying looked like last month.

So the honest uses are narrow. A benchmark detects an order-of-magnitude problem: a B2B landing page at 0.3% against a 2-5% range has something structurally wrong, usually traffic quality or a form asking for a phone number too early, and that is worth knowing. It can also set a first directional goal when you have no history at all, where 2% is a reasonable next step from 1% and 10% is not. Both uses are one-time. Past that first read the benchmark stops carrying information, because every subsequent question is about your own movement, and the benchmark does not move.

What it cannot do is rank you. Sitting above the range is not evidence of a good page, and sitting below it is not evidence of a bad one. If you want the arithmetic itself, our note on how to calculate conversion rate covers the definitional choices that decide which number you end up with.

A.Each stage isolates a different mechanism

A headline rate is the product of several independent machines, so it cannot diagnose any of them. Stage rates can, because each one is sensitive to a different cause. Add-to-cart at 8-12% reads demand and product page quality with checkout friction held out of it. Cart-to-purchase at 45-65% reads checkout friction alone. A healthy add-to-cart with a weak cart-to-purchase puts the problem past the cart, and the inverse puts it on the product pages, which is a conclusion a 2.4% blended rate can never support.

The SaaS stages split the same way. Pricing page to trial at 3-5% (top performers 7-10%) reads price acceptance, and a weak rate there means confusing packaging, a price above expectation, or value that was never established on the pages before, each with a different fix. Trial to paid at 15-25% for a 14-day trial, slightly lower at 10-20% for 30 days, reads activation rather than marketing. The card requirement is the single largest distortion in this row: card-required trials cut signup volume sharply and land at 40-60% trial to paid, some of which is people who did not cancel. Comparing a card-required 45% against an opt-in 18% is not a comparison. Our guide to SaaS product analytics covers what happens after signup.

B2B splits by intent depth rather than by machine. Landing pages convert 2-5% with top performers at 8-12%, and the gap between average and top is the widest in any category here, which usually reflects traffic targeting rather than page craft. Demo requests run 0.5-2% of total traffic, but 30-50% of visitors who reach the request form complete it, so a weak demo number is nearly always an upstream problem. Gated content converts 3-7% at the download page, and field count is the reliable lever: cutting a form from six fields to three can double it. B2C outside e-commerce reflects lower commitment throughout: email signup at 1-3% unincentivised and 5-10% with a real incentive, account creation at 2-5% with social login typically adding 20-40%, free-to-paid at 2-5% of active free users, where the definition of active is doing most of the work.

B.And then they multiply

Stage rates buy comparability and cost you the headline. Sitting at the midpoint of every published range does not produce a mid-range business, it produces a specific number that nobody publishes, because it depends on a stage mix no two companies share. Run the arithmetic and a SaaS funnel at the midpoints lands around 0.7% visitor to paid, and that is the rate your cost per acquisition has to clear.

Multiplication is also why incremental work pays disproportionately. A 10% improvement in add-to-cart combined with a 10% improvement in cart-to-purchase yields 21% overall, and a third stage takes it past 33%. Teams that improve most are not the ones that find a single large win, they are the ones holding several small ones at once. The corollary is less comfortable: a stage that silently regresses cancels two that improved, and a blended rate will show you the net without telling you which stage moved.

This is the point at which benchmarks stop being able to help. There is no published range for the product of your stages, no published range for your source mix, and no way to tell from a blended number whether last month was a page improvement or a campaign that bought cheaper traffic. Answering that requires a series rather than a comparison, which means it requires your own history.

A.Building one that survives a change in traffic mix

The mechanism is straightforward. Record the rate at every funnel stage, over at least 30 days and ideally 90 so a seasonal swing is inside the window, segmented by traffic source, device, and landing template. Every confound that makes an industry comparison imprecise is now held fixed, because it is the same site, the same price, and the same audience on both sides of the comparison.

The segmentation is not a refinement, it is the point. A blended rate moves whenever the mix moves, even when every segment is flat, so an unsegmented baseline reproduces the exact failure of the benchmark on a smaller scale. Buy a month of cheap paid social and your blended conversion rate falls while nothing about the site got worse. Segment-level rates plus the mix weights are the honest pair; either alone misleads.

The practical barrier used to be instrumentation, and it is smaller than it was. Autocapture records interactions without an engineer defining each event first, a setup pass scans the site and proposes the funnel, and you can describe the conversion you care about in plain language and have the Funnel report built around it, then hold the segments as Populations so the same split is reused rather than rebuilt. What that removes is the tracking plan gate, not the thinking: you still have to decide which conversion counts.

B.What a baseline still cannot do

It cannot tell you the ceiling. A baseline says where you have been, and no history distinguishes a page performing near its limit from one that has never been tested properly. That is the residual job a benchmark does, once, at the order-of-magnitude level described in part one, and it is why the two are complements rather than rivals.

It also cannot separate a real improvement from a lucky month. Rates regress, small samples swing, and a stage with 200 conversions a month will produce apparent 15% movements from nothing at all. The discipline that fixes it is the same one that fixes everything else here: diagnose before prescribing, and confirm with a properly powered experiment rather than a before-and-after. Pair the quantitative drop with qualitative evidence, because a funnel tells you where people leave and never why. Our note on CTA design and the landing page guide cover the usual first hypotheses.

What the baseline changes is which question counts as answered. Being above the industry average while declining month over month is a worse position than being below it while improving, and only one of those two facts is visible in a benchmark comparison. Trajectory outranks position, and trajectory is the one thing an industry table structurally cannot show you. For e-commerce-specific ranges by industry, device and source, see our guide to e-commerce conversion rates.