B2B SaaS lead generation playbook

Marketing hits its lead number three quarters running and sales still says pipeline is thin. That argument is running inside most B2B software companies right now, and more leads has never once ended it. What ends it is changing which leads count and where they go.

The position this page takes is simple to state and annoying to act on. Rank every source by cost per closed-won opportunity instead of cost per lead, and the ranking you get back is close to the reverse of the one currently justifying your budget. The rest of this page is the table, the definitions and the routing changes that follow from it.

Which lead sources actually produce closed-won deals

Cost per lead flatters cheap channels and punishes the ones that close. The table below scores eight sources on the three numbers that matter together: what a lead costs, what share of those leads become an opportunity, and how long the money takes to come back.

Source	Typical CPL	Lead to opportunity	Observed payback
Branded search	$8 to $40	25% to 45%	2 to 5 months
Category and non-brand search	$60 to $250	8% to 15%	9 to 18 months
Review marketplaces (G2, Capterra)	$40 to $180	12% to 25%	5 to 12 months
LinkedIn lead gen forms	$60 to $200	3% to 8%	12 to 24 months
Webinars and virtual events	$35 to $120	4% to 10%	9 to 18 months
Partner and customer referral	$0 to $60	30% to 60%	1 to 4 months
Cold outbound (fully loaded SDR)	$180 to $600	10% to 20%	12 to 20 months
Community and dark social	Hard to price	20% to 40% self-reported	Unclear, usually fast

Read the second column before the first. LinkedIn lead gen forms often produce the cheapest looking number on a slide and the worst opportunity rate on the list, because a one-tap form prefilled by the platform asks nothing of the person tapping it. Branded search costs more per click in absolute terms and converts at five to ten times the rate, for the obvious reason that someone typing your company name has already done the work.

Referral sits at the top of every version of this table I have ever built. It is also the source you cannot buy more of this quarter, which is why it gets underinvested. You can increase it, just not on a media buying timeline: partner enablement, customer marketing and a referral mechanic inside the product all take two to three quarters to show up.

The wider set of tactics, including the ones that do not fit neatly in a cost table, sits in our 19 B2B SaaS lead generation strategies guide. This page is about how to choose between them.

What G2, Capterra and the review marketplaces actually cost

Review marketplaces are the most reliably underrated paid channel in B2B software, and the pricing is more variable than vendors admit. In most categories you are buying clicks at somewhere between 2 and 15 dollars. In CRM, project management, HR and anything else with forty vendors bidding, the effective cost climbs above 20 and occasionally past 30.

Capterra enforces a minimum monthly budget around 500 dollars. That number matters more than it sounds, because it kills the popular plan of testing the channel for 200 dollars and deciding from noise. At a 6 dollar click and a 500 dollar floor you are buying roughly 83 clicks a month, which is not enough traffic to read a conversion rate from. Plan on three months at 1,500 to 3,000 dollars monthly before the data means anything.

Pay per lead deals, where a marketplace sells you a contact rather than a click, typically run 30 to 100 dollars and higher in enterprise categories. The arithmetic looks attractive until you audit what arrives. A meaningful share of pay per lead volume is students, consultants and people researching for a report, and the marketplace has limited incentive to filter because it is paid on delivery. If you buy this way, negotiate a rejection window in writing and actually use it.

Your volume problem is probably a definition problem

Here is the diagnostic. Pull the last 500 leads marketing sent to sales. Count how many matched your stated ideal customer profile on company size, industry and geography, before any behavioural scoring. In most companies that number lands between 30 and 55 percent.

If yours does, you do not have a volume problem. You have a definition problem that presents as a volume problem, because sales is drowning in records and starving for buyers, and the only visible symptom is them asking for more. The marketing qualified lead definition is where this gets fixed, and in most companies it has not been rewritten since it was copied off a HubSpot template in 2021.

A workable definition has three gates, applied in this order:

• Firmographic fit, which is binary. Employee count, industry, geography and tech stack either match the profile or they do not, and no amount of engagement overrides a miss.
• Buying group evidence, meaning two or more people from the same account have engaged inside 60 days, or one person with a title in the decision set has.
• Intent, which is the action they took. A pricing page visit, a demo request, a comparison page read, a trial start. Not an ebook download.

Notice what is missing. Content consumption on its own does not qualify anybody, because in 2026 a large share of it is a competitor, a job applicant or a research tool. If you want the longer argument for why demand and leads are different things that get measured differently, our demand generation vs lead generation comparison lays out both sides.

A scoring model built on fit and buying group signals

Point scoring fell out of favour for good reason: most implementations awarded five points for a webinar attendance and five for a pricing page visit, which is a claim that those two behaviours carry the same information. They do not.

The model below is the one I would rebuild with. It separates fit from intent rather than adding them together, because a 95 point score built entirely from behaviour at a 12 person company is a trap.

Two details make this work in practice. First, run the fit gate before scoring, so an off fit account can never reach the threshold no matter how much it reads. Second, score the account rather than the contact. A VP of Engineering reading integration docs while a procurement manager at the same company opens a security page is a live deal, and contact level scoring shows you two mediocre leads.

Product signals deserve their own line in the model because they are the strongest predictor available to a company with a trial, and they are routinely left in the product analytics tool where sales never sees them. Wiring Amplitude or PostHog events into the CRM as scoring inputs is a two week job that outperforms most quarter long campaign programmes. The tooling choices around that sit in our demand generation software stack guide.

What to gate and what to set free in 2026

Gate things that require your data or your product to produce, and set free everything that reads like an argument. A benchmark dataset, a calculator output, an assessment, a template with your logic baked into it, a private demo recording: those are worth a real email address to a buyer mid evaluation. A 14 page PDF restating four blog posts is not, and hiding it behind a form in 2026 mostly stops language models and AI search surfaces from citing you.

That last point has changed the calculus more than most teams have absorbed. When a buyer asks an assistant to compare four vendors in your category, your gated PDF contributes nothing to the answer. Ungating your best thinking costs you a small number of low intent form fills and buys you presence in the place where category shortlists are now formed.

The middle path that works: publish the argument, gate the instrument. Write the full methodology on the page, then offer the spreadsheet, the interactive version or the segment specific dataset behind a form. Our pipeline coverage calculator works this way, and the pattern converts roughly two to four times better than a content gate because the thing behind the form is useful rather than merely longer.

How unqualified volume raises your CAC

This is the part that gets skipped, so here is the arithmetic with real numbers.

An AE working mid-market deals can run roughly 40 to 55 first meetings a quarter before quality drops. Say 45. Each unqualified meeting costs 25 to 40 minutes of live time plus research, notes and CRM hygiene, so call it an hour fully loaded. If 40 percent of routed leads are off fit, that AE spends 18 hours a quarter on conversations that could never close, and more importantly occupies 18 of their 45 meeting slots.

Now price it. A mid-market AE on 180,000 dollars on target earnings costs roughly 15,000 dollars a month loaded. Forty percent waste is 6,000 dollars a month of sales capacity per rep, or 216,000 a year across a three person team, spent producing nothing. That cost lands in your CAC and gets blamed on marketing spend, which is the wrong line item.

Here is the honest cost of doing this. Reported MQL volume will fall 40 to 60 percent in the first month, and it will look like marketing broke something. Pipeline does not improve for six to ten weeks because the deals that benefit have to work through the cycle first. If you cannot get executive agreement to hold the line through two bad looking reviews, do not start, because half executed this is worse than not doing it.

Routing and response time mechanics

Speed and accuracy fight each other, and most teams pick speed and then wonder why win rates did not move. Route on fit first, then optimise the clock inside each lane.

Three lanes cover nearly everything. High intent and high fit goes straight to a calendar via instant round robin, ideally with the meeting booked before the thank you page loads. High fit and low intent goes to an SDR sequence with a 24 hour first touch. Off fit goes to nurture and never touches a human, regardless of how many whitepapers they open.

Response time inside the first lane is where the classic data lives. The finding that a lead contacted at five minutes is around 21 times more likely to qualify than one contacted at 30 minutes comes from James Oldroyd’s Lead Response Management study, published in 2007. Treat the multiple as directionally right and numerically stale: buyer behaviour, form volume and scheduling tools have all changed since. What has not changed is that the shape of the curve is brutal in the first hour and nearly flat after day one.

The mechanical fixes are boring and they work. Put a scheduling tool on the demo form so the buyer books themselves. Alert in Slack, not email. Set a written first touch target per lane and report the 90th percentile rather than the average, because the average hides the leads that sat for three days. Our sales and marketing SLA template has the lane definitions and the escalation path already written.

We did not fix pipeline by spending more. We fixed it by admitting that half of what we sent sales was never going to buy, and then having the argument about it.

Reporting that survives a board meeting

Report three numbers per source and nothing else: cost per opportunity, opportunity to closed-won rate, and blended payback in months. Cost per lead can live in the operating dashboard where the team tunes campaigns, and it should stay out of the board deck entirely, because it invites the question of why you do not just buy more of the cheap thing.

Add one qualitative input that most teams still refuse to collect: a self-reported source field on the demo form, worded as an open question about how they first heard of you. It disagrees with your attribution platform constantly, and the disagreement is the useful part. Podcasts, communities, Slack groups and a colleague at a previous job show up there and nowhere else, which is precisely the demand your last-touch model is crediting to branded search.

For the coverage math that connects all of this to a number, run your current opportunity rate through the pipeline coverage calculator and look at what a 10 point improvement in lead to opportunity does compared with a 30 percent budget increase. In most models the routing change wins, and it is free.

What to change first

Do the 500 lead audit this week. It takes an analyst two days and it tells you whether the rest of this page applies to you, which is a better use of two days than another channel test.

If off fit share comes back above 35 percent, fix the definition and the routing before you touch budget, and warn your CEO that the reported numbers will get worse before they get better. If it comes back under 20 percent, your problem genuinely is volume, and the source table at the top tells you where to add it: branded search and referral first, marketplaces second, paid social last.

Either way, the strategic frame sits one level up in B2B SaaS marketing, the channel mechanics live in SaaS lead generation, and the programme that feeds all of this is covered in the B2B SaaS demand generation playbook. Start with the audit, then pick the next page based on what it tells you.