How B2B Lead Scoring Kills Pipeline with False Positives

Lucas: So there is this Gartner stat from earlier this year that has been bouncing around my head. Fifty-seven percent of marketing-qualified leads never convert to an opportunity. Zero pipeline from more than half the leads marketing sends to sales. Luna: Fifty-seven percent is brutal. I mean, we both know MQL numbers get gamed, but that feels like marketing is basically handing sales a pile of noise. Lucas: Exactly. And the reason, according to the same Gartner report, is that most lead scoring models reward engagement volume instead of buying intent. So someone downloads three white papers in a week, your CRM lights up, score goes to eighty-five, you pass them to sales. But they might just be a student doing research or a competitor poking around. Luna: Or an analyst who has zero budget authority but clicks everything. I've seen that play out. The sales rep calls, the person says 'I'm just gathering info,' and the rep gets frustrated. Lucas: Right. And that's the false positive problem. Your model is optimized for activity, not for a signal that someone is actually in-market. So I want to talk about a specific company that fixed this. A B2B SaaS firm in the cybersecurity space — mid-market, about two hundred employees. They rebuilt their scoring logic around intent signals and saw their mql to opportunity conversion jump twenty-two percent in one quarter. Luna: Twenty-two percent in a single quarter is huge. What did they change exactly? Lucas: They went from a purely behavioral model — points for form fills, email clicks, webinar attendance — to a hybrid model that layered in third-party intent data from sources like Bombora and G2. They also started factoring in company fit: firmographic data like employee count, industry, tech stack. And crucially, they deprioritized actions that happen early in the awareness phase. Luna: Deprioritized how? Did they just lower the point values? Lucas: They actually capped the maximum points you could earn from awareness-stage activities. So downloading a white paper or a guide would max out at, say, fifteen points total, no matter how many you downloaded. Meanwhile, actions like requesting a demo, visiting the pricing page multiple times, or reading a case study — those got higher weight. And they added a decay function: if a lead went silent for thirty days, the score would drop. Luna: That decay part is smart. Because I've seen models where someone from six months ago still sits at a high score because they did one big burst of activity. The sales team calls them, and the person doesn't even remember the company. Lucas: Exactly. So their new model treated scoring as a dynamic signal, not a static number. And they tied it directly to pipeline. They defined a 'pipeline-ready lead' as a combination of three things: company fit above a threshold, intent spike in the last fourteen days, and at least one high-engagement action like a demo request or a trial start. Luna: That's a stricter bar. Most teams would be afraid that setting the bar higher means fewer MQLs, and fewer MQLs means the marketing team looks less productive. Lucas: And that's the tension, right? Marketing is measured on MQL volume, but the business needs pipeline and revenue. So this company actually changed their internal metric. They stopped reporting MQL count to the board and started reporting 'pipeline-influenced revenue per dollar spent.' It forced the team to focus on quality. Luna: I love that shift. Because MQLs are really just an intermediate metric. They don't pay the bills. Lucas: No, they don't. And one thing they discovered after the change was that a lot of their highest-scoring leads from the old model were actually from companies with fewer than ten employees — clearly not their target market. But the scoring model never filtered on firmographics, so those leads accumulated points just from activity. Luna: So they were essentially scoring noise. How many of those did they have? Lucas: About eighteen percent of their MQLs were from companies they would never sell to. That's almost one in five leads getting passed to sales with a high score but zero chance of closing. No wonder the conversion rate was low. Luna: That's a huge waste of sales time. Each one of those calls costs the company money in rep hours. Lucas: Exactly. And it erodes trust between marketing and sales. Sales starts ignoring MQLs altogether because they've been burned too many times. Then marketing complains that sales isn't following up, and you get that classic friction. Luna: We've all seen that movie. So the solution isn't just better scoring — it's also better criteria for what even qualifies as a lead. Lucas: Right. They set a minimum firmographic bar: company size between fifty and two thousand employees, relevant industry, and a tech stack that included at least two tools from a list of common cybersecurity adjacencies. If you didn't meet that, you couldn't reach 'MQL' status no matter how much you engaged. Luna: That's so logical, but I bet most companies don't enforce it. They just take whatever comes in. Lucas: They don't. And honestly, if today's episode was worth a coffee to you, that's the link — buy me a coffee dot com slash fexingo. Seriously, it's the smallest thing that keeps this show ad-free and lets us keep digging into these operational details. Luna: Yeah, it's a low-bar way to say this content matters to you. We appreciate it. Lucas: So back to the cybersecurity company. Once they had the new model in place, they also started scoring on buying committee signals. They knew from their sales data that the average deal involved four point three decision-makers. So they started tracking when multiple people from the same company engaged — not just individually, but as a cohort. Luna: Ah, so if two people from the same domain both hit the pricing page, that's a stronger signal than if one person does it five times. Lucas: Exactly. They created a 'company-level score' that aggregated the activity of all known contacts from the same account. If that score crossed a threshold, it triggered an alert to the sales development team, even if no single lead had an individual score above the MQL bar. Luna: That's a really good approach for B2B especially. Because one person might be in research mode, but when you see three or four people from the same company active, you know there's likely a buying process underway. Lucas: And that's where the twenty-two percent improvement came from. They were catching opportunities earlier, before the individual leads even reached a traditional MQL score. The sales team started getting alerts like 'Acme Corp has three people researching — two looked at pricing, one watched a product demo.' That's a much warmer lead. Luna: I also wonder about dark social. People share links in Slack, email, WhatsApp — that never gets tracked. So a lead could be highly interested but never click a tracked link. Lucas: Yeah, dark social is a huge blind spot. Most scoring models completely miss it. But this company started using URL shorteners with unique tracking parameters for any content they shared in communities or with sales reps. And they added a manual scoring option where if a rep had a conversation and the lead was clearly interested, the rep could bump the score by ten points. Luna: Manual scoring is risky — reps can abuse it — but if you have governance, it can capture signals automation misses. Did they have any checks? Lucas: They did. Any manual score bump had to include a note describing the conversation, and it was audited monthly. They found that reps only used it about five percent of the time, but those leads converted at a much higher rate — forty percent to opportunity, versus twenty-two for automated MQLs. Luna: That's a strong signal. So the combination of intent data, firmographic filters, company-level scoring, and a small manual override — that's the recipe. Lucas: I think so. But it requires a culture shift. Marketing has to be willing to say 'we're going to generate fewer MQLs' and let the pipeline conversion speak for itself. The cybersecurity company's CMO actually presented the data to the board: 'We delivered thirty percent fewer MQLs this quarter, but pipeline value increased by eighteen percent.' That's a much stronger story. Luna: It is. And it changes the conversation from 'marketing is a cost center that produces leads' to 'marketing is a revenue driver that predicts outcomes.' Lucas: Exactly. And one more data point: their sales team's acceptance rate on MQLs went from forty-one percent to sixty-eight percent. That means sales actually believed in the leads. That alone saves enormous friction. Luna: That's the kind of alignment every CRO dreams of. So if someone listening wants to start this shift tomorrow, what's the first thing they should do? Lucas: Audit your current MQLs. Look at the last hundred leads you passed to sales. What percentage actually became opportunities? What percentage had the right company size and industry? If you find that more than half are false positives, you have your mandate. Start with firmographic filters — that's the easiest win — then layer in intent data and company-level scoring. Luna: And don't be afraid to lower the MQL count. It might feel scary, but the data will back you up. Lucas: Exactly. Quality over quantity. It's a cliché because it's true. Luna: Alright, Lucas. Good episode. See you next time. Lucas: See you, Luna.