The Ugly Onboarding That Won: A Conversation with Kuba Tutaj

The best onboarding Kuba Tutaj ever shipped was one he was ashamed to show anyone. Ugly icons, one line of text per screen. It beat the beautiful version and ran untouched for years.

Kuba is a senior growth product manager at Mixmax with over 15 years in product. He spent time at Zendesk, going from PM to principal PM, worked at Base CRM before its acquisition, and earlier grew a company to 400,000 users on a $0 marketing budget. He also co-hosts a Polish product-management podcast. He's a fundamentals guy in an era obsessed with shiny objects.

What a growth PM actually does

Kuba draws a clean line between core product work and growth work. When you build core functionality, you unlock use cases that weren't possible before. The growth PM's job is to bring the people who need those use cases, make sure they discover and activate on them, help them retain, and guide them from simple first moments until they become power users who need more and are ready to buy, with monetization awareness, upsells, and cross-sells visible along the way. Growth PMs, he says, are optimizers of growth loops. They're not creating the product so much as making it available, visible, and ready to try and buy.

And he won't let the role be romanticized as harder than sales. A salesperson runs a hundred conversations simultaneously, each spawning its own touchpoints and tasks, a wide net of cognitive load from low-level detail to high-level strategy and back. Both jobs are hard, complex, and full, and both feel completely normal once you're inside them.

AI changes the level of focus, not the fundamentals

The fundamentals, acquisition, activation, retention, monetization, don't change with AI. What changes is the level of focus each demands, and the speed. Time to value keeps shrinking: at Lovable the activation moment is a prompt that returns an app; at Mixmax, where you once built email sequences from scratch, you now fill in a few fields about yourself and your ideal buyer and AI spins out a performing sequence. So the weight shifts toward distribution, awareness that the capability exists, and upsell.

New metrics emerge too, quality and acceptance rate of AI outputs, accuracy, whether you let users give feedback and actually act on it. Because the risk with prompt-to-result is that the output is shit. And he frames the stakes bluntly: the cost of not learning and testing AI right now is existential. Get it wrong and you learn; ignore it and in three years you're done. Right now every SaaS is a sandbox.

The ugly onboarding

His favorite story is from Base CRM. As a core PM he kept asking leadership whether the features they were building were even used, bringing reports showing people hitting walls. Given a mandate, he looked at onboarding. The existing version was a beautiful product tour, all bells and whistles, with only baseline activation. He brought a version B he was almost too embarrassed to ship, built from around 20 interviews with newly signed-up customers alongside his customer-success partner.

It was "user-driven onboarding": the first screen simply explained what a lead, a contact, and a deal even are, because users think in terms of "reaching out to contacts" and get lost in the product's language. The next screen told them what to do, connect your email, import contacts, invite your team. Ugly model, most basic icons, one clear line of text that drove action. Setup and activation shot through the roof. They never got time to prettify it, and it served for two or three years. High quality matters, he says, but transmitting the right message and creating the fastest path to realized value matters more than feeling good about a shiny product.

Your best salespeople are your GTM team

The mistake product teams make building tools for salespeople is not talking to sales when they should. Your best available salespeople are your own GTM team, and they'll validate or invalidate hundreds of your ideas on the spot. Partner with your VP of Sales, talk about product-qualified leads and revenue numbers, not just the product metrics you care about, so you share a language with the revenue leaders. He runs onboarding calls with new paying customers himself, asking one question: what are you hoping to accomplish with Mixmax? It's empowering, because people think in goals, not features, and the insight flows straight back to the team. "Speak to your customers," he says, and promises to repeat it a hundred times.

From vanity metrics to correlations

Early in his career he watched volumes, emails processed, users hitting some number, and quickly realized those were vanity metrics. He moved to rates: activation, signup-to-trial, trial-to-activation, retention over time. Lately his favorites sit closer to money: correlations. What correlates with purchase, what correlates with higher retention, and modeling initiatives through that lens, if we lift this one rate, what does it mean for the business long term. Less staring at how many users you have, more forecasting on system performance. The connective tissue between revenue, product, and marketing, he adds, is the ideal customer profile, because bringing irrelevant users in defeats the whole point.

Distribution is the bottleneck

Development used to be the bottleneck; now it's distribution, and it feels two-to-five times harder until everyone orients to the new world. Product marketing becomes a different game, less polishing every release, more enabling the whole team to share the distribution load. Finding new users is genuinely hard now that LLMs and AI summaries have turned the game upside down and nobody yet knows how to automate it at high quality. Meanwhile the speed of building isn't a silver bullet, on a green field it's easy, but on ten-year-old code you still need senior humans who understand the system. You can generate a thousand ideas and develop a hundred features, but as Leah Tharin points out, you might still need a hundred months to validate what each does to your business. The full circle always returns to which customer problems to solve first.

Figure out how not to build it

Kuba is candid that he isn't an entrepreneur, he doesn't have the appetite to fail ten times in a row. But advising friends building SaaS, the gap is always the same: talk to ten customers before building. Don't ask whether they'd buy, they'll say yes to be nice and never open their wallet. Ask about their problems, and treat asking for money upfront as a real milestone, because it calibrates whether people trust your solution sight unseen. His reframe of the whole question is the sharpest thing in the episode: instead of "what should I build," ask how you can not build it and still validate it. Design the experiments that prove it without the product, and you'll jump into building better prepared and hungrier to see it live in customers' hands.

Key takeaways

Growth PMs optimize loops, not products. The job is bringing the right users to a capability and guiding them through discovery, activation, retention, and purchase.

AI shifts the focus, not the fundamentals. Acquisition, activation, retention, and monetization still rule. Time-to-value shrinks, so distribution and awareness matter more.

Ugly that converts beats beautiful that doesn't. A basic, clear onboarding that speaks the user's language can crush a polished tour, and last for years.

Your GTM team is your best product research. They'll validate or kill hundreds of ideas on the spot. Talk in revenue, not just product metrics.

Figure out how not to build it. Validate demand with experiments and pre-payment before writing code. Ask about problems, not whether they'd buy.

Frameworks worth stealing

User-driven onboarding

Before showing off your product in onboarding, teach the user the language they're missing. Explain the basic objects in their terms, then tell them the two or three concrete actions that lead to value (connect this, import that, invite them). Clarity that drives action beats a beautiful tour, and you can measure the activation lift immediately.

How can I not build it?

Turn "what should I build" into "how can I validate this without building it." List the experiments, customer conversations, a landing page, a request for money upfront, that would prove or kill the idea. Asking for payment before anything exists is the truest signal of trust, and it sharpens your story before you write a line of code.

Quotes worth keeping

The lines I wrote down.

We ended up with an ugly onboarding, the most basic icons, one line of clear text. Activation shot through the roof.

Your best available salespeople are your GTM team. They'll validate and invalidate hundreds of your ideas on the spot.

Instead of asking what to build, think deeply about how you can not build it and still validate it.

Rapid fire round

Same questions every guest. Quick questions, quick answers.

Best advice you've ever received? To focus.

Advice you ignored and wish you'd listened to? Keep it simple.

What would you tell your younger self? Focus on users even more.

Ongoing challenge that keeps you up at night? Distribution.

Favorite spot? A place in Krakow's Kazimierz district with fresh bread and great wine, a spot he and his wife get to every year or two.

Tool you can't live without? Amplitude. It gives him answers.

---

Kuba (Jakub) Tutaj is a senior growth product manager at Mixmax with over 15 years in product, including a stint at Zendesk from PM to principal PM. He earlier grew a company to 400,000 users on a $0 marketing budget, and co-hosts a Polish product-management podcast. Find him on LinkedIn.