Latest / Marketing Automation with Fexingo: HubSpot, Marketo, and Automated Customer Journeys / Why Your Marketing Automation Needs a Data Diet
Transcript
- Lucas: Let me start with a number. The average B2B marketing database has something like 120 custom fields per contact record. And most of them are empty. Luna: One hundred and twenty? That feels like a hoarder situation. Lucas: It is exactly that. And I think we've quietly created a myth in marketing operations that more data is always better. Capture everything, tag everything, and the algorithm will sort it out. Luna: But the algorithm doesn't sort it out. It just routes it badly. Lucas: Right. So today I want to talk about something I'm calling 'automation entropy' — the idea that every extra field, every unused tag, every untriggered event you add to your automation system actually increases the probability that a campaign will misroute, misfire, or just degrade in performance. Luna: I like that framing. Do you have a specific case? Lucas: I do. There's a mid-market B2B SaaS company — let's call them CloudLogix — they had a lead scoring model that was supposed to send hot leads straight to sales, warm leads into a nurture sequence, and cold leads into a re-engagement drip. Standard stuff. Luna: Standard, but it broke? Lucas: Completely. They had 47 custom fields on their lead object. Fields like 'number of integrations,' 'annual recurring revenue band,' 'tech stack complexity score,' 'decision maker title match,' 'competitor usage flag.' And the scoring model used 31 of them. Luna: That's a lot of weighting to balance. Lucas: It's a nightmare. What happened was, about 22 percent of leads were getting scored as 'hot' — but when sales called, fewer than 5 percent were actually qualified. Meanwhile, a bunch of genuinely high-intent leads were stuck in nurture because they lacked data on one or two of those 31 fields. Luna: So the noise was drowning out the signal. Classic. Lucas: Classic, and totally preventable. The fix was brutal but simple. They cut the number of scoring fields from 31 down to 7. The seven that actually correlated with conversion: company size, industry match, job title seniority, inbound engagement frequency, content download count, demo request history, and referral source. Luna: What happened to accuracy? Lucas: The hot lead qualification rate went from 5 percent to 38 percent in six weeks. And the sales team stopped ignoring the CRM. Luna: That's a massive improvement just from subtracting data. Lucas: Subtracting data, and also subtracting the cognitive load on the automation engine. Every extra rule, every extra field, every extra conditional branch — it doesn't just slow down the system, it multiplies the surface area for error. That's the entropy I'm talking about. Luna: So what's the metric someone should look at first if they suspect they have this problem? Lucas: I'd look at the signal to noise ratio in your automation database. The simplest version: take the total number of captured data points across your lead database in a month — every field update, every event, every page view — and divide it by the number of those data points that actually trigger a meaningful workflow action. Luna: Meaningful as in — sends an email, updates a score, routes to sales? Lucas: Exactly. If your ratio is above, say, 20 to 1, you're probably collecting a lot of noise. A healthy ratio for most B2B marketers is closer to 5 to 1 or 8 to 1. CloudLogix was at 47 to 1 before the cleanup. Luna: That number is shocking but honestly, I bet a lot of listeners just did the math in their head and felt a little sick. Lucas: Yeah, I think a lot of marketing ops people know intuitively that they have a data hoarding problem, but they've never had a simple way to diagnose it. The signal to noise ratio is that diagnostic. Luna: I want to push back on something though. The instinct to capture everything doesn't come from nowhere. It comes from the fear that you'll miss the one data point that explains why a deal closed. How do you balance that? Lucas: That fear is real, and it's not irrational. But the solution isn't to capture everything — it's to be deliberate about what you capture. Before you add a new field, ask: 'If I have this data, what specific action will my automation take that it couldn't take without it?' If you can't answer that in one sentence, don't add the field. Luna: That's a good rule. And you can always add it later if a use case emerges. Lucas: Exactly. The cost of adding a field isn't just storage — it's the cognitive overhead on your workflows, the training burden on your team, and the increased entropy. Every field is a potential point of failure. Luna: You mentioned untriggered events earlier. What about tracking scripts and page view data? That's a huge source of noise for a lot of teams. Lucas: Huge. I've seen instances where a company tracks 200 distinct events per visitor per session — scroll depth, hover time, mouse movement. And they're feeding all of that into their automation platform. The signal to noise ratio on that is astronomical. Luna: But some of that data is genuinely useful for personalization, right? Like, if someone scrolls to the bottom of a pricing page, that's a signal. Lucas: It is, but it's one signal in a sea of noise. The key is to pre-filter at the event level. Don't send every scroll event to your marketing automation platform. Aggregate it client-side and only send the meaningful ones — 'pricing page visited for more than 30 seconds' is a good signal. 'Mouse moved 12 pixels' is not. Luna: So the data diet applies to events too. Lucas: Absolutely. Every event that enters your automation system should have a clear, documented purpose. If you can't write down why that event exists, delete it. Luna: That's hard to do retroactively, though. Most marketing automation platforms don't make it easy to just delete fields or events. There's always the fear you'll break something. Lucas: It is hard. But the cost of not doing it is worse. CloudLogix spent two weeks cleaning up their data model, and they saw results in under two months. The alternative — leaving the entropy in place — means your automation performance degrades slowly over time, and you don't even notice because you're too busy adding more fields. Luna: I want to circle back to the signal to noise ratio as a regular diagnostic. Do you think a marketing ops team should track that monthly, quarterly? Lucas: I'd say quarterly is sufficient. Monthly might be overkill unless you're adding new fields or events frequently. But the real value is in the trend — if your ratio is increasing quarter over quarter, you've got a data hygiene problem that will eventually break your automation. Luna: And what's the biggest red flag number? If someone sees their ratio go from 8 to 1 to 15 to 1 over two quarters, what should they do? Lucas: Immediately audit the last three months of field and event additions. Chances are, someone on the team added a bunch of tracking without thinking about the workflow consequences. Reverse those additions unless they have a clear, documented purpose. Luna: It's such a simple principle — less is more — but it's counterintuitive in a world where we're told to collect more data for AI and personalization. Lucas: It is counterintuitive, and that's exactly why it's worth talking about. The most sophisticated marketing automation in the world won't save you if your data is junk. And the most common source of junk data is not bad data — it's irrelevant data. Luna: You know, if these marketing conversations have sparked something you've actually used — a tweak to your workflow, a cleanup project — I think that's the whole point. We deliberately don't run ads on these shows, and if you want to support that choice, there's a link. It's buy me a coffee dot com slash fexingo. Lucas: Yeah, and honestly, it's a small thing that keeps us free to talk about exactly this kind of practical, no-nonsense stuff. No sponsors to please, no products to plug. Luna: Exactly. So back to the signal to noise ratio — I think one more thing worth touching on is how this ties into the AI hype. A lot of vendors promise that AI will just 'figure out' your messy data. Lucas: Right, and that's dangerous. AI can work with noisy data to some extent, but it's always better with clean data. If your automation entropy is high, feeding it into an AI model just means the AI will learn the wrong patterns faster. Luna: Garbage in, garbage out — but with AI it's like garbage in, garbage amplified. Lucas: Exactly. So before you invest in an AI personalization engine or a predictive lead scoring model, clean up your data model first. It's the single highest roi thing you can do in marketing ops. Luna: Alright, so for someone listening who wants to start tomorrow: what's the one action they should take? Lucas: Run a report of all custom fields on your lead or contact object in your marketing automation platform. Then for each field, ask: 'Is this field used in any active workflow, scoring model, or segmentation list?' If it's not, delete it or archive it. You'll be shocked how many orphan fields you find. Luna: I think that's a great place to start. And if you find 47 fields like CloudLogix did — well, you know what to do. Lucas: Start the diet. Your automation engine will thank you.