Latest / The Marketing Funnel Podcast with Fexingo: Awareness, Interest, Decision, and Action Stages / How Salesforce Marketing Funnel Uses Data to Drive Action
Transcript
- Lucas: So Salesforce reported last quarter that its marketing cloud revenue crossed the $5 billion annual run rate mark. That's the CRM giant selling marketing tools to marketers. And the way they market their own product tells you a lot about where the funnel is heading. Luna: I'm curious — because Salesforce's own marketing has always felt a bit… corporate-coded. Keynotes, white papers, the Dreamforce spectacle. Are they actually doing something different now? Lucas: Yeah, and that's exactly the angle. For years, Salesforce's funnel was classic enterprise: awareness through events, interest through case studies, decision through a sales demo. But starting in late 2024, they quietly shifted the entire action stage to ai driven prediction. Instead of waiting for a lead to fill out a form, they now score intent in real time across web behavior, email engagement, and even third-party data. Luna: So the action stage used to be 'request a demo.' Now it's 'the algorithm decides you're ready'? Lucas: Pretty much. Salesforce claims that their predictive lead-scoring model, which runs inside Einstein GPT, improved conversion rates by 34 percent in the first half of 2025. And here's the thing — they didn't just apply this to incoming leads. They also stopped doing cold email sequences entirely. No more 'warm-up' drip campaigns. Luna: Wait — one of the biggest B2B marketers on earth just dropped email nurturing? That's a huge shift. Lucas: It is. But their reasoning is interesting. They argue that traditional email nurturing actually trains prospects to ignore you. If you send a five-email sequence and the person never opens, you're just building noise. So they replaced it with what they call 'triggered intelligence' — content only surfaces when the data suggests the person is already in a buying motion. Luna: That sounds great on paper. But how do you even define 'buying motion' for someone who's never heard of you? Isn't that the awareness stage problem? Lucas: Exactly — and that's where they've reconfigured the awareness layer too. Salesforce now treats awareness as a data-collection engine. Every blog visit, every whitepaper download, every page scroll is fed into the scoring model. So the top of the funnel isn't about messaging — it's about signal harvesting. Luna: So the old AIDA model — Awareness, Interest, Decision, Action — gets collapsed. Awareness and interest merge into 'signal.' Decision and action get merged into 'score threshold.' Lucas: Right. And that's a pretty radical departure from how most marketing textbooks still teach the funnel. But Salesforce has the data to back it. They published a case study on a Fortune 500 client — a financial services firm — that used this model and saw a 28 percent reduction in sales cycle length. Luna: I'd love to know what the sales team thinks. Because if the marketing funnel now decides which leads are 'ready,' the sales team might feel like they've lost control of their pipeline. Lucas: That's actually the biggest internal friction point. Salesforce's own sales reps initially pushed back. They were used to getting a list of thousands of names and qualifying them manually. Now the system only hands them maybe 200 high-scoring leads per quarter. Some reps complained they were missing out on potential deals. Luna: And did the data bear that out? Were they actually missing deals? Lucas: Salesforce ran a six-month experiment. They let one region use the old model — full email sequences, manual qualification — and another region use the predictive model. The predictive region closed 19 percent more revenue with the same headcount. So the data supported the change. Luna: That's persuasive. But I wonder — does this only work for a brand as established as Salesforce? If you're a startup with no historical data, you can't train a lead-scoring model. Lucas: That's the catch. Salesforce's model is trained on billions of interactions across their customer base. A new company would have to start with rules-based scoring and gradually switch to predictive. But the principle still applies: don't blast everyone; wait for the signal. Luna: It also raises a broader question about the awareness stage. If you're only capturing data from people who already engage, aren't you just preaching to the converted? Where does new audience growth come from? Lucas: That's where Salesforce still invests in top of funnel brand marketing — but they measure it differently. They don't track impressions or click-through rates. They track 'new-to-brand' contact acquisition cost. If someone lands on the blog and isn't in the CRM yet, that's a measurable signal. They spent about $200 million on brand advertising in 2025, according to their annual report, but they tied every dollar to that metric. Luna: So the funnel isn't gone — it's just become a data feedback loop. Awareness feeds the model, the model feeds the action, action feeds back into the model. That's almost a flywheel. Lucas: Exactly. And it's worth noting that Salesforce isn't alone. HubSpot, Marketo, even smaller players like ActiveCampaign are moving in this direction. But Salesforce's scale makes it a useful case study because the numbers are public and the stakes are high. Luna: Before we go deeper into how other companies might adopt this — Lucas, I want to touch on something you mentioned earlier. You said Salesforce stopped email warm-up sequences entirely. That's a pretty bold claim. Can you walk through what that actually looked like? Lucas: Sure. So a typical warm-up sequence is: day one, 'Hi, check out this ebook.' Day three, 'Here's a case study.' Day seven, 'Free trial?' Day fourteen, 'Last chance.' Salesforce eliminated all of that. Instead, if a prospect visits a pricing page twice in a week, the system triggers a personalized video from a sales rep within 24 hours. No email sequence. Just a direct human touch when intent is high. Luna: But what if the prospect is never ready? You just never contact them? That seems like leaving money on the table. Lucas: Salesforce's data suggests that most of those 'maybe later' leads never convert anyway. Their model found that only 2 percent of leads who entered a nurture sequence and didn't engage in the first 30 days ever became customers. So they're comfortable ignoring them. It's a pretty ruthless prioritization. Luna: Speaking of ruthless — if that kind of data-driven focus is useful to you in your own marketing, it's exactly the kind of thing we talk about on this show. And we keep these conversations free of ads because we think the topic deserves that. If you find value in that, you can support the show at buy me a coffee dot com slash fexingo. Just a simple way to say 'keep going.' Lucas: Yeah, and it genuinely helps us stay independent. We don't have to tailor content to sponsors. So if that's something you appreciate, that link is there. Otherwise, let's get back to the funnel. Luna: So, Lucas — one critique I've seen of Salesforce's approach is that it might work for big-ticket B2B, but what about consumer goods? Or lower-priced SaaS? Can you really apply predictive scoring to a $10-a-month subscription? Lucas: Good question. For low-price products, the cost of the scoring model might exceed the customer lifetime value. But the principle — trigger-based action — still applies. Think about Netflix: they don't send you 'warm-up' emails to watch Stranger Things. They just show you the trailer when you're already on the platform. That's triggered intelligence. Luna: So the funnel becomes more about timing than volume. It's less 'how many people can we reach?' and more 'when is the right moment to reach them?' Lucas: Exactly. And that shift has huge implications for how you budget. If you're spending $500,000 on a trade show booth to generate 1,000 business cards, but only 50 of those cards ever show buying intent, you might be better off spending that $500,000 on a predictive model that finds 200 high-intent prospects from web data alone. Luna: But trade shows also build brand awareness — which can't be measured by immediate intent, can it? Lucas: That's the classic tension. Salesforce still attends Dreamforce — huge event. But they measure it differently now. They track how many of the attendees become 'new-to-brand' contacts in their CRM within 90 days. If the event doesn't generate that, they'd reconsider the investment. That's the level of discipline they're applying. Luna: So the entire marketing funnel is being rebuilt around data science. The creative side — the copy, the design, the story — does that become secondary? Lucas: Not secondary, but more targeted. Salesforce's emails now have subject lines that are A/B tested by Einstein GPT. The creative is still important, but it's optimized for the specific person at the specific stage. So the copy might change based on whether you're a CTO or a marketing manager. Luna: That sounds like a win for efficiency. But I worry about the loss of serendipity — the customer who discovers you randomly and falls in love. Does the data-first funnel kill that? Lucas: It might if you over-optimize. Salesforce's CMO, Sarah Franklin, said in an interview that they reserve 10 percent of their marketing budget for 'unmodeled experiments' — things that don't fit the predictive framework. So they're aware of the risk. But the core engine is data-driven. Luna: That 10 percent figure is telling. It acknowledges the limits of the model. Because no algorithm can predict a Black Swan — or a viral moment. Lucas: Right. And that's where the human marketer still matters. The model can tell you when to act, but it can't tell you what to say. That's still the job of the creative team. Luna: So what's the takeaway for someone listening who manages a marketing funnel today? Should they start building a predictive model? Lucas: Start small. If you have a CRM, look at your historical data: which behaviors preceded a sale? Maybe it's visiting the pricing page three times. Maybe it's downloading a specific whitepaper. Build a simple rules-based trigger first. Then, as you collect data, you can graduate to a machine-learning model. Luna: And if you don't have any data? What's the first step? Lucas: Define your ideal customer profile manually. Then focus on getting that person into your funnel — even if it's through a simple newsletter. Then start tracking what they do. The data will come. Luna: It's interesting — we started this episode talking about a $260 billion company's AI model, and we're ending with 'start a newsletter.' That's the funnel in a nutshell. Lucas: Yeah. The tools change, but the fundamentals — know your customer, time your message — they don't. Salesforce just happens to have billions of data points to refine the timing. Luna: Alright — next episode, let's look at a company that does the exact opposite. Maybe one that relies entirely on serendipity. Lucas: I know just the one. See you next time.