Conversion rate: what good actually looks like

Somebody in the room will say the number. "Average B2B conversion rate is two point three five percent" — cited from a report, attached to no source anyone in the meeting could name, and treated as a bar the company has either cleared or not. Nobody asks what a "conversion" was in that report. Nobody asks whether the seven thousand websites it was averaged across sold anything like what this company sells, to anything like the audience this company sells to. The number gets written into next quarter's target anyway, because it is the only number in the room, and having a target feels better than admitting nobody has one.

This happens constantly, and the instinct behind it is not wrong. Wanting an external reference point is reasonable — how else would you know if two percent is bad? The failure is not the wanting, it is settling for the first number that shows up, because that number was built by averaging together things that should never share an average.

The denominator is doing all the damage

Every conversion rate is a fraction, and the entire argument in every benchmark report is about the numerator — what counts as a conversion, a form fill versus a booked call versus a newsletter signup, and reasonable people can argue about which of those is worth counting. That argument is a distraction from the bigger problem, which sits in the denominator and gets almost no scrutiny at all.

The denominator in a published benchmark is "traffic." Not a defined kind of traffic — all of it, whatever arrived, from whatever companies contributed their analytics to whatever report aggregated them. A visitor who typed the company's name into a search box because they already had a signed contract in their inbox counts the same as a visitor who clicked a display ad they will never think about again. A person who has been reading the company's newsletter for two years and finally clicked through counts the same as someone who landed on the blog from an unrelated Google search and left in four seconds. These are not similar events wearing different disguises. They are different populations with wildly different intent, blended into one rate, and the blend ratio is different at every company that contributed data to the report. A company running heavy brand campaigns and one running none will land at different benchmark positions with identical product-market fit, purely because their traffic mix differs, and the published number cannot tell you which is which.

This is why two companies can both report a two percent conversion rate and be in completely different situations — one converting warm, already-convinced visitors at a rate that would embarrass them if they knew it, the other converting cold strangers at a rate that is genuinely impressive. The published average erases exactly the distinction that matters, and then gets handed back to both companies as if it applied to each of them.

Segment by intent before you look at a single rate

The fix is not a better benchmark. It is refusing to compute one number across a traffic mix you have not defined, and instead running the calculation separately for populations that actually behave alike.

Direct traffic — someone typed the URL or used a saved bookmark — is close to the warmest population you have; they already know the company exists and chose to come back. Branded search, someone who searched the company's own name, sits right next to it: they were already looking for this specific company, not evaluating it against alternatives. Referral traffic, arriving from another site's link, carries whatever intent that other site's audience carries, which varies enormously depending on whether the referrer is a partner's resource page or a random forum thread. Non-brand organic — someone searching a problem or a category, not a company name, and landing on the site without having heard of it before — is the coldest traffic that still shows any intent at all, and it is the segment that answers the question people actually mean when they ask about conversion rate: how good is this site at turning a stranger into an enquiry.

Compute a rate for each of these separately and the branded and direct numbers will usually look flattering, sometimes ten times the non-brand figure, and that gap is not something to explain away. It is the whole point of doing the segmentation. A high blended rate that is mostly brand and direct traffic converting is a healthy sales relationship expressing itself through the website, not evidence the website itself is persuasive to someone who has never heard of the company. The number worth tracking, worth reporting, and worth trying to move is non-brand organic to enquiry, because it is the only one of the four that isolates what the site does rather than what the company's existing reputation does for it. It is also usually the smallest of the four, often by an order of magnitude, which is worth saying plainly before anyone panics at the drop from the blended figure they were used to quoting.

An enquiry is not the same thing as a qualified enquiry

Segmenting the traffic side fixes half the fraction. The other half — what counts as a conversion at all — needs the same discipline, because "enquiry" quietly expands to include whatever the form captured, and a rising enquiry count gets reported as success even when sales is complaining that none of the recent leads go anywhere.

Write the definition down before measuring anything. Not as a vague sense of "good leads" but as a checklist a sales rep could run against a submission in ten seconds: company size within the range the product is built for, a problem the product actually solves, a role with some ability to buy or recommend, sometimes a stated timeline. Anyone who submits the form and clears that checklist is a qualified enquiry. Anyone who submits the form and does not is still an enquiry — worth counting, worth understanding, but not the number that goes in front of the leadership team as the site's conversion rate.

The two numbers move independently, and that independence is the entire reason to track both. Removing a field from a contact form, adding a chatbot with a low bar to open a conversation, running a broad awareness campaign — all of these reliably raise the enquiry count and just as reliably dilute the qualified share, because the friction being removed was doing screening work whether anyone designed it to or not. A team that reports enquiries alone will watch the top-line number climb for a quarter while sales quietly disengages from the website as a source, and nobody in the reporting chain will be able to explain the disconnect, because the report never separated the two things that were moving in opposite directions. This is the mechanism behind the lead quality problem, and it is worth reading in full if a marketing team and a sales team are currently blaming each other for numbers that are both technically true.

Build the target from your own history, not somebody else's report

Once the traffic is segmented and the outcome is defined, the benchmark question mostly dissolves, because the honest target was never going to come from outside the company anyway. It comes from the trailing twelve months of the company's own non-brand organic to qualified enquiry rate — enough data to smooth out a bad month and a good month, short enough to still reflect the current site rather than a version from three redesigns ago.

Take that trailing rate as the floor, not the ceiling. A target that is simply "beat the last twelve months" is boring and correct, and it has a property no external benchmark can offer: it is comparable to itself. Next December, this December's number is the honest comparison, adjusted for whatever changed in the traffic mix in between — a new campaign that shifted the ratio of branded to non-brand traffic will move the blended rate without the site having gotten better or worse at anything, which is exactly why the segmentation has to happen before the comparison, not after.

Give a rate change enough time before believing it

The last mistake is impatience, and it compounds the first two. A visitor who lands on the site this month does not necessarily convert this month. In a B2B sales cycle that runs weeks or months, this month's non-brand visitors are still working through evaluation when next month's report gets written, and this month's conversion count is provisional in a way a monthly dashboard cannot show. Comparing week over week, or even month over month, against a sales cycle measured in weeks is comparing a number to a version of itself that has not finished happening yet.

The practical rule is to size the waiting period to the sales cycle, not to the reporting calendar. A company with a two-week cycle can trust a monthly comparison. A company with a ninety-day cycle needs at least one full quarter before treating a change in the rate as real, and ideally two, because the first quarter after a site change is exactly the quarter still full of visitors who arrived under the old version and are only now reaching a decision. The same patience that applies to traffic volume generally — covered in how long before SEO shows up in the numbers — applies here with extra force, because a conversion rate is a ratio of two things that are each still settling, and a ratio of two unsettled numbers is the least stable statistic available to anyone running a monthly report.

None of this produces a single number to put in a slide next to a competitor's. It produces something more useful: a rate you defined yourself, on traffic you segmented yourself, compared honestly against the only history that was ever actually comparable to it, which is measuring what a website does applied specifically to the one metric everyone reaches for first and understands least.