No two published reports price a B2B lead the same way

Where three publishers agree

This is the rare marketing term where the definitions line up, and they line up right to the moment anyone has to count something or pay for it. After that the same four words stop being useful.

B2B lead generation is identifying the companies that could buy from you and getting them interested enough to raise a hand. Two definition pages write it as identifying ideal buyers and then enticing or attracting them.

The encyclopedia entry keeps it shorter, as "attracting and capturing consumer interest in a product or service", and the shorter version is the one that travels into ad copy without the qualifying half.

That entry is also alone among the three in defining the thing being counted. A lead, it says, is "the contact information and, in some cases, demographic information of a customer who is interested in a specific product or service." Contact details, plus interest. That is the whole object.

The definition most of the web repeats, in full

1. 1The whole object is two things: contact details, and interest. Nothing about a budget, a role or a timeframe.
2. 2The footnote marker sits on the definition. The pages selling you leads carry no equivalent.

That width is also why cost per lead is a metric before it is an amount: the divisor is whatever you decided to count.

Two published answers to that width exist, and they point in opposite directions. The peer-reviewed literature keeps the object and adds a step.

Migao Wu, Pavel Andreev and Morad Benyoucef reviewed 44 studies and describe qualification as ranking leads against a scale so you can prioritise them.

Forrester drops the object instead. Its B2B Revenue Waterfall guide asks you to move beyond marketing-qualified leads and to advance opportunities with connected buying groups.

So the thing this page counts is the thing Forrester wants you to stop counting. Say you win a deal after four people at one company each downloaded something: on Forrester’s reading that is one opportunity, and on yours it is four leads.

Hold onto how wide that is. Contact details plus interest covers a newsletter signup and a webinar registration. It also covers a demo request from a company that has already picked you, and a badge scan from someone who wanted the tote bag. The definition does not separate them. It asks for contact details and interest, and it leaves who decides that interest open. They are not worth the same. They do not cost the same. No published definition separates them.

The platforms selling you leads do not narrow it either. Google Ads documentation for lead form assets describes what the form does, that people "submit their contact information (such as an email address or phone number)", without saying what makes a submission worth counting. LinkedIn's own guide skips the definition too. It goes straight to the part that matters to it: "one hundred unqualified leads are far less valuable than 10 qualified ones."

Four reports and four definitions

Google suggests a question next to this term asking what a good cost per lead is for B2B. That is the practical question. Four of the definition pages take the term on without printing an amount. So here are the reports that do, one at a time.

FirstPageSage publishes cost per lead by industry. B2B SaaS comes out at $310 paid and $164 organic. Cybersecurity is $411 and $404, IT and managed services $617 and $385, software development $680 and $510, manufacturing $691 and $415. Thirty industries, two columns each.

A second report publishes by channel instead, as quartiles. Google Search Ads runs a $60 bottom quartile, a $140 median and a $280 top. LinkedIn Ads runs $90, $220 and $380. Content marketing and SEO run $30, $85 and $180. In person events run $400, $650 and $1,200. Account based marketing aimed at one enterprise account runs $1,200, $2,800 and $5,000 and up. That top band is the price of naming the buyer before you market to them, and it only pays under conditions you can check. For example, say you sell to 240 companies at six figures a deal: at that deal size $2,800 a lead survives a conversion rate well under one in ten. Say you sell to 20,000 companies at $4,000 a deal, and it does not survive any rate you could plausibly hit. That report publishes the price and not the rate, so the rate is the input you supply. The fit test is in our guide to account-based marketing.

A third report publishes ranges: paid search $60 to $220, LinkedIn $90 to $320, organic and SEO $50 to $150 fully loaded, trade shows $200 to $600 and up. A fourth publishes averages: Google Search $45, Meta $35, LinkedIn $125, TikTok $28.

Read those four paragraphs again and notice that you cannot put them on one chart. FirstPageSage prices "paid" with no channel split. The second prices "Google Search Ads". The third prices "Paid Search", which may or may not include Microsoft. The fourth prices "Google Search". Four labels, four scopes, and not one defines its scope.

The statistics do not match either. A $45 average, a $140 median and a $60 to $220 range are three different kinds of number. Subtracting them gives you a figure, and the figure means nothing.

A fifth report is worth naming for the opposite reason. LocaliQ, with WordStream, publishes search advertising benchmarks across more than twenty industries. It says whose campaigns it counted: "thousands of our customers' advertising campaigns across Google Ads and Microsoft Ads." Its overall figure is $66.69, with Arts and Entertainment lowest at $26.84 and Attorneys and Legal Services highest at $131.63.

That figure is not the B2B answer. It prices a conversion, not a lead, across a customer base spanning more than twenty industries, of which B2B software is one. The figure earns its place here twice over. It answers the one question the other four skip: whose campaigns produced the number. It also shows what happens when a method is held still. The spread above is what one industry choice costs, with nothing else changed. Variation this wide is normal, and the reading only breaks down when what varied goes unstated.

For instance, pricing your own lead before buying against it is what our B2B PPC service sets up first.

How many campaigns produced these numbers

Four things decide whether any of those five numbers can be set beside another: the sample size, the attribution window, double counting, and what made a lead qualified.

Five pages, and not one gives a number. LocaliQ names a population without counting it: thousands of its own customer campaigns. FirstPageSage is the only one that gives a data window: January 2022 through June 2025. The same page says it was last updated on 8 May 2025, before the window it claims to cover ends. Nobody else says how many campaigns, companies or leads sit behind the figure. The same page carries the line "Last updated: May 8, 2025" above a report labelled 2026.

The most transparent of the four, in full

1. 1A data window is given, and it ends fifteen months before the report's own label.
2. 2No sample size. Not how many companies, not how many campaigns, not how many leads.

You cannot tell whether a lead credited to search arrived from search that day or first saw an ad two months earlier, because the attribution window is missing everywhere. You cannot tell whether the same person filling in two forms counted once or twice. And you cannot tell what made a lead qualified enough to enter the count in the first place.

That absence is about the published page, not about the work behind it. There may be a methodology sitting in a document somewhere. A reader comparing their own number to these has the page, and the page does not say.

One report names the problem and publishes averages anyway. Its own sentence about scope is worth more than its averages.

The numerator moves too

1. 1Which costs count is a decision, not a fact. Creative, landing pages and tools are commonly left out.
2. 2The 35 to 60 percent gap is ConversionStudio's attribution to HubSpot's 2025 State of Marketing Report. It was not verified in that report.

On that attribution, companies including full campaign costs report figures 35 to 60 percent higher than those counting ad spend alone. That compares reporting conventions, not one campaign counted two ways. Nobody publishing a benchmark tells you which convention theirs used.

One table with a registrant and a qualified lead

The scope problem shows up when you compare publishers to each other. A second problem needs no comparison at all. It sits inside a single report, in two lines of the same table, under one heading.

One report prices webinars and virtual events at $35 to $160 per registrant. Two lines later it prices content syndication at $45 to $130 per verified marketing qualified lead. A registrant is a person who typed an email address into a form. A verified marketing qualified lead has been checked against a definition of fit. Both appear in the same table, under the same heading, as cost per lead.

Another report prices the webinar channel per lead, at a $150 median, against that $35 to $160 per registrant. Whether those two describe similar campaigns is not knowable from either page: neither says what the offer was, who the audience was, or how a registrant became a lead.

This is why a low benchmark is not automatically good news. For instance, suppose two teams each spend $10,000 on webinars. One counts registrants and reports 200 leads at $50. The other counts only registrants who passed a fit check, finds 40, and reports $250. Same campaign, same spend, five times the cost per lead. The cheaper figure describes a looser filter, not a better campaign.

It runs the other way at the sales end too. One vendor published a customer's numbers on this: Alt21 reported a close rate "around 0.2%" from content leads against "nearly 20%" from direct inbound enquiries. Those are the source's own approximations, from one company, published by a vendor. Cohort sizes, period and the definition of a close are not given. Read it as a reason to check your own two rates separately, not as a ratio to carry anywhere.

Build your own number before you compare it

A benchmark you cannot reproduce is trivia. Four decisions turn cost per lead into something you can act on, and all four are yours to make.

1. Write the lead test

   Which form, which fields, and what has to be true about the company. Not a category name, a test. For example: a demo request, from a company with more than fifty staff, where the work email resolves to that company. If two people on your team would classify the same submission differently, the definition is not finished.

2. Fix the cost convention

   Media spend is the easy part. Creative production, landing page work, the tools and the share of a salary that runs the programme are the parts commonly dropped. Whichever convention you pick, keep it fixed, because a number that changes definition halfway through a year cannot be compared to itself.

3. Set the window

   If somebody first arrives in March and fills in a form in May, which month carries the cost and which carries the lead? Decide before you need it, or every awkward case gets settled in the direction that flatters whichever channel someone is defending.

4. Split by channel

   A blended figure hides the thing you would act on. LinkedIn costing more than search is only bad news if the leads convert at the same rate, and separating them is how you find out that they do not.

Google Analytics Help, attribution settings, read 10 September 2026Whether March still gets the credit for a form filled in May

The two lookback rows are Google’s published defaults and alternatives. The first two rows are settings this figure does not fill in.

Only after those four are fixed does an outside benchmark become useful, and even then it works as a sanity check, not a target. For instance: your search cost per lead comes out at $400, and the publisher whose scope most closely matches yours prints something near $140. That gap is worth an afternoon. If yours lands at $150 instead, you have learned nothing, because you still do not know what they counted.

What cost per lead does not decide

Everything above is about making the metric honest. This part is about how far an honest version of it carries you. Not as far as the attention it gets would suggest, and there is a better number waiting behind it.

Cost per lead is a cost, not an outcome. It answers what you paid for whatever you decided to count, and the Alt21 figures show how far that can sit from whether anyone buys.

A cost per lead is priced at the moment of capture, so it cannot know whether that lead qualified. The number worth watching sits one step down: what each channel costs once its leads have passed the qualification stage you defined. Both need a defined handoff between marketing and sales. Our guide to the difference between an MQL and an SQL covers where the line sits and who moves it.

A practical rule: publish the definition next to the number, every time. Three lines are enough. It takes a paragraph, and it separates a metric your board can question from one your board has to take on faith.

That paragraph is missing from every benchmark quoted on this page, and its absence is the whole reason a reader cannot use them to answer the question they came with.