The vibe coding trap marketers risk falling into

Welcome to Evidently—the bi-weekly newsletter where a first-time VP of Marketing shares the (sometimes) ugly truth of what it’s like to do this job. If you’re just joining, I’ve talked a lot about reflecting, leading, and running so far. Hop into the archives if any of those verbs speak to you. Today, though, is all about vibe coding.

"Can we just build this ourselves" is a question more and more marketers are coming up against with the rise of vibe coding. If I’m being honest, it’s less of a question, more of a mandate. “Before you buy it, try building it.”

I get the instinct. I've felt it myself. And for a narrow set of things, the answer is genuinely yes, and it's great. But I think a lot of marketing leaders are about to learn the same lesson the hard way: Building a tool that looks good on the front end, and building a tool that works consistently, is secure, and maintained over time are two very different projects. And if you're the one who said yes to building the latter, you're now the engineer supporting it.

That distinction is worth unpacking.

Where vibe coding’s actually worked for us

The vibe coding wins on our team all share something in common. They're small, use case specific, and don't require a huge lift to actively maintain them post-build.

Our Product Marketing Manager Mary used Replit to create an internal repository that houses our best customer examples. It's easy to add to and update, the content is accessible and shareable across the entire team, and it looks good.

Our Director of Growth Luke built a Temple Run-style game a while back in promotion of our annual The Evidence Gap report, aptly named: Jump the Evidence Gap. We're currently working on another interactive experience that’ll serve as a digital companion to our Pass the Proof direct mail campaign. Think: a digital Pass the Pigs-inspired game that adds delight to a physical touchpoint. (Keep this on the down low because we haven’t launched it yet…)

On the enablement side, Jillian, our Director of Content Marketing, built an interactive prompter that helps the team turn our research report findings into LinkedIn post ideas, which made it a lot easier to get more people posting creatively around our launch. 

In overseeing Highline this year, she also built an automated event status checker that pings her in Slack with new hotel room bookings so she can make updates in HubSpot. It’s the kind of unglamorous, recurring task that vibe coding excels at.

Dashboards are a clear frontrunner use case, but come with a caveat.  Luke's built marketing analytics dashboards, Jillian's built an event budget tracker. Both are useful for pulling data into one place quickly and displaying it in the ways we want. You need to be willing to double-check the numbers though, as the hallucination factor is still very real. I’m wary and push the team to gut check everything coming out of these. But as it becomes more trustworthy, I think this is a great use case.

Where it falls apart

The mistake with vibe coding is assuming that small wins scale into a build-your-own-everything strategy.

The moment you're trying to build something with real logic, real users, and real uptime expectations, you've quietly signed up to build and maintain a product. And most marketers, myself included, are not equipped to do that alone unless we're unusually technical. Case in point, we tried pushing something onto our own website that required more than a vibe coded front end, and it became clear fast that we'd need real engineering resources to get it live and keep it that way. These things aren’t always as plug and play as everyone likes to claim they are on LinkedIn.

Because of this, the question with vibe coding for marketers quickly morphs from "can I build this?" into "am I signing up to be this thing's engineer for as long as it exists?" If the answer is yes, that's a real cost, and it's worth being clear-eyed about it before you commit.

Some numbers behind the gut feeling

We just wrapped a research project with Scott Brinker of chiefmartec, surveying 300+ marketing leaders on what they think about vibe coding and how teams are actually using it. The full report lives here if you’re interested. Turns out, the use case pattern I'm describing applies beyond just my team.

The most common things marketing orgs are vibe coding right now are workflow automations, integrations, and data pipelines. Behind-the-scenes stuff. Customer-facing apps sit way down the list, only 17% of respondents have built one. I’m not surprised by that at all.

There’s also an understandable skills gap at play that contributes to what marketers can feasibly build. Teams cite it as the number one barrier to expanding vibe coding further across their teams. The part that’s probably not too surprising: Only 21% of orgs include training as a guardrail, even though the orgs that do outperform everyone else on confidence and speed to ROI. In my mind, there’s got to be a connection between engineering and marketing to make this work well, which adds workload outside the marketing org. Is it a useful long-term bet to make? I’m not sure yet.

For now, we're telling people to build without teaching them how to build well. And if most orgs won't invest upfront to train people properly, I wouldn't bet on them investing in the harder, less visible work of maintaining what gets built.

Where we go from here

We're all in the phase right now where we're mandated to use these AI code generation tools. And because of that, along with a lack of training, marketers are running into the gotchas of the maintenance of it. The folks who have more resources to truly build and maintain will either do that or they'll find it's too much and they'll go back to vendors for the things they need. Because those vendors are always updating their products. They’ve secured things on the backend. There's a lot of benefit to that in value, time, and cost when compared to the DIY alternative. Especially as AI gets way more expensive. 

The skills of building a product are way different than marketing one. This begs the question: are we asking marketers to become something different? Will we all have to become technical marketers to a certain degree? Will team sizes continue to shrink? We’re seeing some regression to the mean already with AI costs ballooning, but it’ll likely settle somewhere in the middle. Point solution and homegrown AI toolsets will become more of the norm, especially in larger companies, just like point solution SaaS tools have found their place. Same same but different?

In my opinion, I think it’s a miss to have marketers not market. Don’t get so distracted by shiny UI’s you can spin up in seconds that you forget your team needs to market your product. Is what you’re vibe coding helping do that?

I don't think vibe coding is going anywhere. But the current hype around it has to either ground itself in realism or recalibrate away from the shiny tools to the operational norms and people using them. Those who test and stay savvy will be in a good spot. Those who forget the fundamentals of marketing will get found out. And hopefully this whole thing will end up with some sort of positive impact on humanity… but I’m not gonna get into that here.

—Al

3 things I’m digging this week:

• Vibe Code Check - 300+ marketers weigh in on how they’re actually vibe coding today.

• Mary, our creative powerhouse PMM, published our first few customer use cases this week. Not to be confused with regular case studies. I’m super bullish on use case marketing and can’t wait to fill out this library. Would love to hear who y’all think are nailing use cases in SaaS right now.

• “The greatest hour in ultrarunning” - The Western States 100 is the premier 100-mile ultra in the US. It went down this past weekend and beyond the broken records, the final hour before cutoff (called “golden hour”) didn’t disappoint. Highly recommend watching the moment when the final runners came in. It’s always the most inspiring part of a race to me. Pure grit.

UserEvidence, who?

UserEvidence makes it easy to earn and keep trust from first touch through expansion.

We turn customer evidence, references, and advocacy into one system that builds credibility with buyers and strengthens relationships with customers.

• Not buried in Slack.

• Not stuck in decks.

• Not dependent on a hero effort from one person.