The Onboarding Playbook Growth Teams Actually Run

Here's the uncomfortable truth: most product teams obsess over onboarding flows that don't matter. They add product tours, tooltips, and welcome emails while their activation rates stay stubbornly flat. The real work of onboarding—the work that drives retention 6 months later—happens in the first 5 minutes, and it has almost nothing to do with UI walkthroughs.

The growth leaders who've cracked this understand something counterintuitive: great onboarding feels like less onboarding, not more. Laura Schaffer discovered this by accident at Twilio when she ran a test that should have tanked conversion. Lauryn Isford rebuilt Airtable's entire activation strategy around a metric most teams would reject. And Vaibhav Sahgal learned at Zynga that the path to better onboarding started with admitting his own biases were worthless.

What emerges from their experiences isn't a template. It's a philosophy that challenges the prevailing wisdom about user activation.

The signup question that accidentally improved conversion by 5%

Laura Schaffer thought she was about to get fired. As VP of Growth at Amplitude (previously head of growth at Twilio), she'd just violated the cardinal rule of conversion optimization: never add friction to signup. She inserted qualification questions into Twilio's flow—the kind of questions every growth playbook tells you to avoid—and braced for the metrics to crater.

We just asked for forgiveness and kind of put these questions into the signup flow and ran this A/B test with a small group. And I'm fully expecting, okay, this is gonna hurt our numbers, but maybe it won't be so bad. And we start to get the data for this thing. I'm getting an improved conversion. Like we, there's no personalization, nothing past it, just the questions. It improved signups by like 5%. I'm like, what?

— Laura Schaffer, VP of Growth at Amplitude

The insight here cuts against a decade of growth orthodoxy. Schaffer's questions didn't improve conversion because they enabled personalization downstream. They improved conversion because they signaled intent. Users who paused to answer questions were pre-qualifying themselves as serious prospects. The questions created a micro-commitment, a small threshold that filtered out tire-kickers while validating real users.

This finding points to a deeper principle: friction isn't always the enemy. Strategic friction—the kind that helps users self-select—can improve activation precisely because it reduces the number of people who sign up but never activate. The goal isn't maximum signups. It's maximum qualified signups.

Why your activation metric is probably wrong

Lauryn Isford, former Head of Growth at Airtable, makes a claim that sounds absurd on first hearing: a lower activation rate might be better than a higher one. She argues that most teams optimize for activation metrics that are too easy to achieve, which means they're measuring behavior that doesn't actually predict retention.

An activation rate that falls in a lower percentage range, maybe for most companies, 5 to 15%, is better than one that falls in a higher percentage range because it means that there's likely much higher correlation with long-term retention. And you're really working hard to get most of your users to reach a state that they're not reaching today.

— Lauryn Isford, Head of Growth at Airtable

Isford's framework inverts conventional wisdom. Instead of celebrating a 40% activation rate, she'd push teams to find the deeper action—the one only 10% of users complete—that actually correlates with retention 6 months out. At Airtable, this meant moving beyond "created a base" to "invited a collaborator" or "built a relational structure." Those actions were harder to achieve, which meant fewer users activated. But the users who did were 10x more likely to stick around.

This creates a strategic choice for growth teams. Do you optimize for a high activation rate that looks good in a board deck but doesn't predict retention? Or do you stomach a lower activation number while investing in getting the right users to the right moment?

The best teams choose the latter. They accept that onboarding should feel difficult for some users—because those users weren't going to retain anyway.

The Mary from Ohio problem: why your biases wreck activation

Vaibhav Sahgal, now VP of Consumer Product at Reddit, spent years at Zynga building features for users who looked nothing like him. He wanted to add sports integrations to Words with Friends. His users—mostly women in their 50s from the Midwest—wanted something entirely different.

Our users were 50 to 60-year-old women in the Midwest, and they would use this product because they thought it was Fitbit for the brain. And that was just a massive, massive insight for us. In the beginning, I was just littered with all of my biases. What worked for us is putting a picture of this person. Let's call her Mary from Ohio, who is in her 50s and has grandkids. Every time you're making a decision, think about Mary and would Mary want this?

— Vaibhav Sahgal, VP Consumer Product at Reddit

Sahgal's solution was almost embarrassingly simple: a printed photo of Mary in every conference room. But the principle underneath matters more than the tactic. Most onboarding flows fail because they're designed for the product team, not the user. The PM who lives in product analytics assumes everyone else does too. The designer who loves minimalism builds flows that confuse users who need more context.

The fix isn't more user research (though that helps). It's a forcing function that makes the team confront their assumptions every single day. At Zynga, that meant a photo. At other companies, it might mean mandatory user interviews, or recorded sessions playing on loop in the office, or a Slack bot that posts user quotes every morning.

The format doesn't matter. The discipline does. Great onboarding comes from teams who've killed their own biases, not teams with the most sophisticated prototyping tools.

When product-led growth crashes into reality

Pulkit Agarwal, CEO of Chameleon, built his company on the promise of self-serve onboarding tools. Then he nearly went bankrupt because self-serve wasn't the right model for his actual customers.

The mistake was logical: Agarwal and his co-founders were the kind of people who bought software with a credit card for $25/month. So they built Chameleon for people like themselves. The problem? Their real customers were enterprise teams with dedicated activation owners, evaluation processes, and budgets that required POCs and implementation support. The self-serve model Agarwal built wasn't just inefficient—it was preventing the people who actually needed Chameleon from buying it.

Leah Tharin, PLG advisor at GotPhoto, draws a distinction that would have saved Agarwal months of pain: product-led growth isn't a business model, it's an enabler. And it only works when the product naturally lends itself to self-serve adoption.

The lesson cuts both ways. Teams that force PLG onto products that require high-touch onboarding waste time building self-serve flows that frustrate users. And teams that avoid PLG because "our product is complex" often miss opportunities to reduce friction for a subset of users who could activate without help.

The question isn't "Should we do PLG?" It's "Which users can activate themselves, and which need our help?" Great onboarding answers that question honestly, then builds different paths for different segments.

The first 5 minutes are the whole game

Ron Schneidermann, CEO of AllTrails, and Maria Thomas, former Chief Product Officer at Buffer, arrive at the same conclusion from different angles: if users don't hit value in the first session, onboarding has already failed. Not "struggled." Failed.

Thomas points out that most teams treat onboarding as a sequence of screens users move through before they get to the "real product." That's backwards. Onboarding is the real product. It's the first impression, the first value delivery, the first moment users decide whether this thing is worth coming back to.

Schneidermann's experience at AllTrails reinforces this. The app's entire onboarding philosophy centers on getting users to their first trail recommendation as fast as possible. Not explaining features. Not collecting preferences. Not showing a product tour. Just: here's a trail you'll probably like, based on where you are right now.

That's activation. One clear action that delivers immediate value. Everything else—the account creation, the preference settings, the feature discovery—can happen later, once the user has a reason to care.

The implication is uncomfortable for teams that have invested heavily in onboarding flows: most of those flows are solving problems users don't have yet. They're answering questions users haven't asked. The first 5 minutes should do one thing well, not 10 things poorly.

AI won't save your broken onboarding

Marily Nika, GenAI Product Lead at Google, has watched countless teams try to fix onboarding problems by adding AI. It rarely works, because they're using AI to automate a bad experience instead of fixing the underlying experience.

AI isn't the product. The experience is the product. AI tools, AI models, AI agents—these are not the product. It's gotta be about how AI enhances the customer experience and delivers real value. The foundation of growth is still about value creation and experience. AI is a new tool we can use.

— Marily Nika, GenAI Product Lead at Google

Nika's point echoes what Pulkit Agarwal learned at Chameleon: product tours and onboarding tools can't fix a product that doesn't deliver value. Adding an AI chatbot to a confusing onboarding flow doesn't make the flow less confusing. It just adds another layer of complexity.

The teams that use AI well in onboarding treat it as an accelerator, not a replacement. They use it to personalize paths users are already successfully taking. They use it to surface the right feature at the right time for users who've already activated. They use it to reduce time-to-value for qualified users.

What they don't do is use AI to paper over product-market fit problems, or to automate onboarding for users who shouldn't be onboarding in the first place.

What actually works: the anti-playbook

The growth leaders in this piece disagree on plenty. Schaffer adds friction; conventional wisdom removes it. Isford wants lower activation rates; most teams want higher. Sahgal needs photos in conference rooms; others rely on data dashboards. Agarwal abandoned self-serve; Tharin champions it for the right products.

But they converge on a few non-negotiable principles:

First, stop optimizing for vanity metrics. Signups don't matter if they don't retain. Activation rates don't matter if they don't correlate with long-term engagement. The only metric that matters is the behavior that predicts a user will still be active in 6 months. Find that behavior. Build onboarding around getting users to that behavior. Ignore everything else.

Second, kill your biases or they'll kill your product. You are not your user. Your preferences don't matter. Your workflow doesn't matter. What matters is understanding why Mary from Ohio opens your product and what she's trying to accomplish in the first 5 minutes. If you can't answer that question with specificity, you're not ready to build onboarding.

Third, less is more, but only if "less" is the right thing. Removing steps from your onboarding flow is pointless if users still don't reach activation. The goal isn't fewer screens. It's faster time to value. Sometimes that means more friction (Schaffer's questions). Sometimes it means more handholding (Agarwal's sales-assisted onboarding). Sometimes it means one hyper-focused action (Schneidermann's trail recommendation). Context matters more than tactics.

Finally, onboarding never ends. The teams that win don't build an onboarding flow and call it done. They test, iterate, rebuild, and test again. Schaffer ran her test on a Friday night because she knew it might break things. Isford rebuilt Airtable's activation metric after years of using the wrong one. Sahgal kept that photo of Mary in the conference room for 4 years because he knew his biases would creep back.

Great onboarding isn't a project. It's a discipline. And the leaders who've mastered it don't have a playbook to share. They have a mindset: relentless focus on the moment a user decides your product is worth keeping.