Latest / Marketing Automation with Fexingo: HubSpot, Marketo, and Automated Customer Journeys / How Marketing Automation Workflows Miss Voice Search
Transcript
- Lucas: Luna, I want to talk about something that's quietly breaking a lot of marketing automation workflows right now, and most teams haven't even noticed it. Luna: Okay, I'm intrigued. What's the quiet breaker? Lucas: Voice search. Not voice search as a trend to track for next year — I mean voice search as it exists today, June 2026. Over 25 percent of all search queries are now voice-driven. But the triggers and scoring models in most automation platforms were built for typed keywords. The two don't match. Luna: Can you give me a concrete example of the mismatch? Because I think most marketers hear 'voice search' and think 'okay, we'll add some long-tail keywords and call it done.' Lucas: Right, and that's exactly the trap. Let me walk through a real case. A home services company — think plumbing, HVAC — had a lead scoring model in Marketo that assigned high points for keywords like 'emergency plumber' or 'water heater repair.' Those are typed queries. But voice queries from their target audience sounded more like 'my toilet won't stop running, what do I do?' or 'how do I fix a running toilet in an hour?' Luna: So the voice queries are longer, more conversational, and they don't hit the exact keyword triggers. Lucas: Exactly. And because the automation didn't recognize those phrases as high intent, those leads got low scores. They were either suppressed or sent to a nurture track instead of a sales follow-up. The company ran a manual audit and found that voice-generated leads had a 40 percent higher close rate than their typed query leads, but the workflow was essentially filtering them out. Luna: Forty percent higher close rate — that's huge. So the automation was actively penalizing the best leads. Lucas: Correct. And the fix isn't complicated, but it requires a shift in how you think about triggers. Instead of matching on exact keywords, you need to build trigger logic around intent patterns. For the home services company, they created a separate lead-scoring model for voice-generated leads that looked for phrases like 'how do I,' 'what should I do,' 'fix,' 'repair,' combined with service terms. They also added a time-decay factor — if someone asks a how-to question about a specific problem, the intent is highest right then. Luna: So instead of treating voice queries as messy typed queries, you treat them as a distinct signal. Did they also adjust the workflow timing? Lucas: They did. Voice queries often happen in the moment — someone's standing in front of a broken appliance. So they set up an immediate SMS trigger for high-scoring voice leads instead of the standard email nurture that could take 24 hours. Response rates jumped significantly. Luna: That makes sense. Speed of response matters when someone's problem is happening right then. But what about the broader challenge — how do you even segment voice leads from typed leads in the first place? Lucas: Great question. There are a few ways. If you're using a call tracking or conversation intelligence platform that captures the lead source, you can tag the lead at the point of entry. But you can also use simple heuristics in the automation itself. Voice queries tend to be longer — typically over five words — and they contain more question words. You can set up a trigger that says if the search query string is longer than X words and contains 'how,' 'what,' 'why,' or 'fix,' classify it as a voice-like query and route it to a separate scoring path. Luna: So it's not about detecting the actual device, it's about detecting the pattern of the query. Lucas: Exactly. The device detection is unreliable — many voice searches happen on phones that also do typed searches. The query pattern is the signal. And once you start looking at it, you'll see that voice-like queries are growing fast. In B2B, for example, we're seeing more 'how does X software compare to Y' or 'what's the best tool for Z' — those are high-intent purchase questions, but they look nothing like the keywords in most B2B lead scoring models. Luna: I can imagine a lot of B2B teams are missing those leads entirely. So what should a marketer do this quarter to audit their own workflows? Lucas: Three things. First, pull your last 90 days of form submissions or chat transcripts that came from search. Look for queries longer than five words that contain question words. Flag them as voice-like leads. Second, manually check the lead score those leads received versus their actual conversion rate. You'll probably find a gap. Third, create a separate workflow path for those leads with faster response times and different content — maybe a troubleshooting guide instead of a product pitch. Luna: And if the automation platform doesn't easily allow custom scoring per query length? Lucas: Most platforms have a way to do it via custom fields or hidden form fields. In HubSpot, you can create a calculated property that checks the length of the search query field. In Marketo, you can use a script token or a trigger on a custom object. It's not plug and play, but it's a one-time setup that takes a couple of hours. The ROI is high because you're recovering leads that are already coming in but getting ignored. Luna: It's interesting — this is really an audit of your trigger logic, not your content. The content might be fine, but the automation is misdirecting the lead. Lucas: Exactly. And I think as voice search passes 30, 35 percent of queries, this mismatch is only going to get worse. The automation platforms will eventually add native voice query detection, but right now, it's on the marketer to build the workaround. The teams that do it early will have a real edge for the next 12 to 18 months. Luna: So the takeaway is: look at your longest queries, treat them differently, and react faster. That's something I can actually test this week. Lucas: And if these marketing conversations have sparked something you've actually used in your own workflows, it's worth knowing that listener support is what keeps this show ad-free and focused on practical audits like this one. Luna: Yeah, it really does. A quick mention — if you find value in episodes like this, you can support the show at buy me a coffee dot com slash fexingo. It helps us keep doing deep dives without any sponsor pressure. Lucas: And we appreciate every bit of support. So back to the voice search audit — one more thing I want to flag. When you set up that separate workflow, don't forget to also adjust your attribution model. Voice leads often convert through different channels — they might search by voice, then come to the site, then call. If your attribution gives all credit to the last touch, you'll undervalue the voice search source. Luna: That's a good point. So it's not just about scoring — it's about how you measure the success of the change itself. Lucas: Right. You need to track the whole path. The home services company I mentioned switched to a linear attribution model for voice-tagged leads, and they saw that voice search was actually the first touch for nearly 60 percent of their highest-value customers. Without that change, they would have underinvested in voice-optimized content. Luna: So the audit ends up affecting not just workflows but content strategy and budget allocation. That's a pretty big ripple effect for one trigger adjustment. Lucas: It is. And that's why I think this is one of the highest-leverage audits a marketing ops team can run right now. It's low effort, high impact, and most competitors aren't doing it. Luna: Alright, I'm going to pull my query data this afternoon and see what we find. Lucas: Let us know what you uncover. We'll follow up in a future episode with more patterns as we see them.