What Is a Good Conversion Rate, and How Do You Know Yours Is Bad?

A good conversion rate is a conversion percentage above your sector median, ranging from 2.35% to 6.6% across industries in the Unbounce Q4 2024 Conversion Benchmark Report, with top-quartile landing pages achieving 11.4% or higher using the formula conversions divided by total sessions times 100.

There is no universal figure that applies to every business model, and the numbers quoted most often are blended averages across verticals that share neither price points nor buying cycles. The question teams actually mean when they ask it is narrower and more practical: is my rate low enough that fixing it is worth money and engineering attention right now? That question has a mathematical answer, and getting to it takes about twenty minutes inside your own analytics.

Conversions divided by sessions, expressed as a percentage. The arithmetic is trivial and almost everyone gets it wrong anyway, because the denominator is chosen badly. Two rules fix most of it, and both are about narrowing what you count rather than adding sophistication. There is a third rule that only matters once you start reporting to other people: fix the window. A rate measured over 28 days and compared to one measured over a calendar month will differ by enough to start an argument that has nothing to do with the website.

Count sessions, not users, unless you have a deliberate reason to do otherwise, and be consistent about it forever. Then split by landing page rather than reporting one number for the site. A site at 2% overall is usually one page at 6% carrying four pages at 0.4%, and the average hides the only page worth working on.

Benchmarks are useful for one thing: telling you whether your number is unusual. They are useless for setting targets, because the top quartile of any category is populated by sites with brand demand you do not have. Read the table below as a range you sit inside, not a goal you are failing to hit. Aggregate studies weight heavily toward large sites with paid traffic, which pulls the medians down for lead generation and up for ecommerce. Treat a gap of half a percentage point as noise.

According to the Unbounce Conversion Benchmark Report, dedicated landing pages achieve an all-industry median conversion rate of 6.6%, but sector dispersion is massive. High-urgency consumer categories such as Events and Entertainment reach a 12.3% median, Legal Services sit at 6.3%, SaaS and B2B technology average 3.8%, and Higher Education lands at 2.6%. The spread reflects commitment cost and buying committee size: signing up for a webinar or ticket requires five seconds of impulse, whereas B2B software contracts ranging from €5,000 to €50,000 per year require multiple stakeholder sign-offs. Your own history over the last twelve months remains a better benchmark than any external aggregate.

Before concluding that the site is broken, rule out the four measurement faults that produce artificially low numbers. In our audits, roughly a third of low rates turn out to be reporting problems rather than page problems, and each one is cheaper to fix than a redesign. Each of the four takes under an hour to verify and none of them requires a developer. Run all four before commissioning any work, because a redesign built on a broken denominator will be judged against the same broken denominator afterwards.

• Conversions that fire on some paths and not others, so phone clicks and PDF downloads never count
• Internal and agency traffic sitting in the same property as customers
• Bot traffic inflating sessions while conversions stay flat
• A thank-you page reachable by refresh, which double counts and hides the opposite problem

A single site-wide rate mixes audiences that behave nothing alike, and comparing that mixed number to a benchmark tells you almost nothing. Branded search converts several times better than cold display, so a shift in traffic mix moves your rate without anything on the site changing. This is the most common reason a rate falls in a month when nobody touched the pages.

Split the report four ways at minimum: branded organic, non-branded organic, paid, and direct. Then compare each stream against itself over time. A campaign that doubled sessions from a cheap source and halved your average rate has not hurt you, and the site-wide number will insist that it has.

Google's own documentation on analysing conversions by channel explains the grouping model, and the default channel groups are good enough for this purpose without custom work.

Conversion rate is a ratio, and ratios can be improved by shrinking the denominator, which is why optimising it in isolation sometimes destroys value. Cutting a qualifying question from a form raises the rate and fills the pipeline with leads the sales team cannot close. Dropping the price raises the rate and takes the margin with it.

The correction is to carry a second number beside the rate on every report: revenue per session, or for lead generation, qualified leads per session. If the rate rises and revenue per session falls, the change was a loss dressed as a win. Teams that track only the ratio eventually optimise their way into worse business, and it takes two quarters to notice.

Weekly reporting on a conversion rate produces noise and panic in roughly equal measure. Below a few hundred conversions a month the week-to-week swing is larger than any change you could make deliberately, and reacting to it means reacting to randomness. Monthly is enough for most sites, with a twelve-month trend line beside it.

There are three moments when a check is genuinely warranted regardless of schedule. After a release that touched templates or forms, because tracking breaks silently. After a change in traffic mix, because the mix moves the number by itself. And eight weeks after a redesign, which is the earliest the data means anything.

According to Nielsen Norman Group's guidance on measuring UX, most teams over-collect metrics and under-decide on them, which describes conversion reporting exactly.

Evaluating whether a conversion rate change is real requires calculating sample size and Minimum Detectable Effect (MDE) before launching any experiment. Calling a test early because one variant leads after four days is the most common mathematical error in marketing, producing false positives over 40% of the time. Rigorous testing locks in a 95% confidence threshold and 80% statistical power before a single visitor enters the split.

Sample size depends on your baseline conversion rate and the relative lift you need to detect. If a landing page converts at 2.0% and you want to detect a subtle 10% relative improvement (moving the rate from 2.0% to 2.2%), you need roughly 38,000 sessions per variant, or 76,000 sessions total. By contrast, if you test a structural overhaul designed to produce a 30% relative lift (moving from 2.0% to 2.6%), the required sample drops to roughly 4,400 sessions per variant. Sites with fewer than 5,000 monthly sessions per page should skip micro-copy tweaks and test bold layout, offer, and form-length changes through disciplined A/B testing.

Page speed and form complexity provide the largest reliable lifts when you need high-impact hypotheses. According to Portent's conversion speed research, pages loading in 1 second convert 2.5x to 3x higher than pages taking 5 seconds, and keeping Largest Contentful Paint (LCP) below 2.5 seconds and Interaction to Next Paint (INP) below 200 milliseconds prevents mobile drop-off before the form even renders.

Not every low number deserves budget, and knowing which ones do is the whole decision. The test is arithmetic rather than judgement: multiply your monthly sessions on that page by the gap between your rate and the median for your type, then by your average order or lead value.

If the answer is smaller than the cost of the work, leave it alone and spend on traffic instead. If it is several times larger, you have found your next project, and it is almost never the homepage.

The changes that move conversion rates are duller than the ones that get written about. Across our client work the same four appear at the top of nearly every audit, in roughly this order of return per hour spent, and none of them require a redesign.

Cutting form fields to the minimum that lets you follow up. Putting proof, whether that is logos, numbers or a named client, directly beside the thing you are asking for. Making the price or the price range visible somewhere. And removing the second call to action that competes with the first.

Across the audits we run, one change outperforms the rest often enough to be worth naming: shortening the form. It is unglamorous, it takes hours rather than weeks, and it works because every field is a small decision the visitor has to make while already unsure.

The rule we apply is to keep only the fields you would act on within a week. A company size dropdown that nobody filters by is not qualification, it is friction with a business justification attached. Phone number on a first-contact form costs conversions in every test we have run and is almost never used before the second call.

For ecommerce the equivalent lever sits in the checkout rather than the form. Baymard Institute's checkout research documents an average cart abandonment rate near 70%, with a large share traced to forced account creation and to costs appearing late. Both are single changes with measurable effects, and neither requires touching the product pages everyone wants to redesign first.

A benchmark article that ends in agreement rather than action is worth nothing, so here is the shortest sequence that turns your number into a decision. It fits in two weeks alongside other work, and it costs nothing but attention.

Week one is measurement. Verify the four faults above, then build the funnel table: every landing page with sessions, conversions, rate and value per session for the last twelve months. Sort by sessions. Do not interpret anything yet, because interpretation before the table is where opinion enters and stays.

Week two is a single decision. Take the top five rows, calculate for each the gap between its rate and the median for your type, multiply by sessions and by value. One row will be several times larger than the others. That row is your project. Write down what you expect the fix to return and by when, before you start, because that sentence is the only thing that will let you judge the result honestly in three months.

Compare it to your own last twelve months first, to your site type second, and to the top quartile never. If it sits below the median for your type and the arithmetic above says the gap is worth real money, the next step is a structured CRO audit and hypothesis-driven A/B testing rather than a guess about button colors.

If you would rather have that done properly, our conversion optimization work and dedicated landing page builds start from your funnel table and end with a ranked list of changes, with clear scope and pricing before work begins.