Latest / Marketing Automation with Fexingo: HubSpot, Marketo, and Automated Customer Journeys / Why Your Marketing Automation Needs a Lead Scoring Rebuild
Transcript
- Lucas: Let's talk about lead scoring — but not the way most articles do it. Luna: Okay, I'm interested. Because I feel like every marketing automation platform has a lead scoring module, and most people just turn it on and forget it. Lucas: Exactly. And that's the problem. The default scoring models in HubSpot, Marketo, Pardot — they assign points for things like email opens, form fills, page visits. Those signals have a shelf life. And most teams never ask: how predictive is this signal, really? Luna: So you're saying a form fill from three weeks ago should not be weighted the same as one from today? Lucas: Right. And the data backs that up. I was looking at a case study from a B2B SaaS company — I'll call them CloudSync — who had a 12-signal scoring model. They were scoring things like 'attended a webinar' at 20 points, 'downloaded a whitepaper' at 15, 'visited pricing page' at 25. And their sales team was complaining that leads were cold by the time they got them. Luna: Classic complaint. Sales says marketing leads are trash, marketing says sales doesn't follow up fast enough. Lucas: Exactly. So CloudSync did something radical. They rebuilt their scoring model from scratch. They asked: which signals actually correlate with a lead converting to a paying customer within 90 days? They pulled historical data and ran a regression analysis. Luna: And what did they find? Lucas: They found that only six signals had any real predictive power. Email opens? Zero correlation. Whitepaper downloads? Almost zero. The strongest predictor was 'product demo completed' — not requested, completed. Second was 'visited the integrations page' — that correlated with enterprise deals. Third was 'created a free account and used it more than three times in a week.' Luna: So they cut their signals in half. That must have felt risky. Lucas: It did. But they also introduced something called a signal decay curve. Every scoring event has a half-life. A demo completion might have a half-life of 48 hours — meaning if you don't follow up within two days, the predictive value drops by 40 percent. A pricing page visit might have a half-life of four hours. Luna: So the score actually decreases over time if there's no new engagement? Lucas: Exactly. Instead of a flat score that stays forever, the points decay. So a lead who visited the pricing page three weeks ago but hasn't done anything since — their score drops. It gets deprioritized. And the sales team only sees leads that are actually hot right now. Luna: What was the result for CloudSync? Lucas: Within one quarter, their sales-accepted lead rate went up 34 percent. And the average time to first touch dropped from 72 hours to under 12. Because the scoring model was actually telling sales who to call first. Luna: That's huge. And honestly, I think a lot of teams are afraid to touch their scoring model because it's been in place for years and they don't want to break something. Lucas: Yeah, it's the 'if it ain't broke' trap. But the thing is — it is broke. Just silently. And speaking of things that are quietly running in the background — if these marketing conversations have sparked something you've actually used or thought about trying, there's a small way to keep these episodes coming without any ads or sponsors. Luna: Yeah, it's a couple of dollars a month at most — genuinely makes a difference for us. Lucas: It's buy me a coffee dot com slash fexingo. And if that's not your thing, no pressure at all. But it does let us keep the conversation going without any commercial breaks. Luna: Okay, back to scoring. So after CloudSync rebuilt their model, did they keep it static or do they audit it regularly? Lucas: Great question. They set up a quarterly scoring audit. Every three months, they re-run the regression on the last 90 days of data to see if the predictive power of each signal has shifted. And guess what — it does shift. Luna: Of course. Because buyer behavior changes. What was predictive in Q1 might not be in Q3. Lucas: Exactly. For example, in one audit they found that 'visited the case studies page' had become predictive for a new product line they'd launched. So they added that signal. And 'attended a webinar' had dropped to near zero — because the webinars were getting stale. Luna: So they removed it. That's the discipline most teams lack — they accumulate signals over time but never prune. Lucas: Right. There's a concept in machine learning called feature drift. The same thing happens in lead scoring. The signals that used to predict conversion lose their power. If you're not auditing, you're scoring yesterday's buyers with yesterday's model. Luna: So what's the concrete takeaway for someone listening right now? What's the first step? Lucas: First step: export your last 90 days of leads. For each lead, note whether they converted to a paying customer. Then list every scoring signal you're using — every form fill, every email click, every page visit. Run a simple correlation: which signals are actually present in the converted leads versus the non-converted? You'll probably find that two or three signals account for 80 percent of the predictive power. Luna: And then you can drop the noise. Which also makes the scoring model easier to explain to sales. Lucas: Huge win. Because when sales understands why a lead scored 80 points — 'this person completed a demo yesterday and visited the pricing page' — they trust the score. When it's a black box of 12 signals, they ignore it. Luna: And implement decay curves. That's a technical lift in most platforms, but some like Marketo have native decay functionality. Lucas: Yes. And if your platform doesn't support it natively, you can hack it with a custom field that subtracts points based on days since last engagement. Not elegant, but it works. Luna: I like that. So the big idea: clean signals, decay curves, quarterly audits. That's a scoring model that actually drives revenue instead of just filling the CRM. Lucas: Exactly. The goal isn't to score every lead. It's to surface the ones that are ready to buy right now. And most teams are drowning in 10,000-point leads that aren't actually hot.