Your onboarding isn't a UX problem. It's a psychology problem.

Last week, product and engineering leaders gathered over dinner to hear from three behavioral scientists working at the intersection of psychology and product design. The panel featured Kristen Berman, CEO and founder of Irrational Labs, Cassie Yount, Product Marketing Leader at LinkedIn Premium, and Kelvin Kwong, CPO at Mars Pet Care. The through-line? Most onboarding problems aren’t design problems, they’re human behavior problems. And solving them requires thinking less like an engineer and more like a psychologist.

1. Numbers beat words. Always.

LinkedIn ran the test. Generic qualitative messaging like “apply faster” and “get noticed” lost to quantified proof points like “3× more profile views” or “apply 60% faster” every time.

Behavioral scientists call this the quantification dominance principle. When a number is present, humans anchor to it. A stat cuts through the noise, and if you can measure the value your product delivers, lead with the number and not a description of the product. This matters especially in onboarding, when you have seconds to make a case for someone’s attention.

2. Personalization is mostly theater, and that’s fine.

“The perception of being understood matters more than the technical sophistication of your recommendation engine. Show your work.”

LinkedIn added a “we’re personalizing your plan” interstitial to their onboarding, despite already having 98% of the information they needed. The message was almost entirely performative. It still drove a dramatic conversion lift.

Behavioral scientists call this the Barnum effect. People respond powerfully to the feeling of being understood, even when the “personalization” is mostly generic. A health app reframed a clinical assessment as a shareable personality quiz and saw 23% more therapeutic engagement. A fitness company found users preferred simple peer comparisons over sophisticated regression models correlating caffeine and sleep patterns.

3. Some friction is actually good for you.

“Good questions signal that you understand the user’s problem better than they expected. That’s a selling point, not a speed bump.”

The conventional wisdom is to remove every step that stands between a user and activation. The data from this panel suggests that’s wrong, at least partially.

LinkedIn added questions before their final conversion moment and saw a 20% lift. A mental health product added questions that demonstrated genuine empathy for anxiety sufferers, and saw onboarding completion go up.

The distinction is between questions that serve the user (demonstrating empathy, enabling real personalization) and questions that serve only the company (name, zip code, demographic data you don’t use). Cut the latter. Keep the former.

4. If you want someone to take an action, put it in front of them.

Headspace surfaced a specific course at the end of their onboarding quiz, which for most users  was just the basics course. No personalization required, and enrollment doubled.

One Medical added appointment booking to their onboarding flow. Bookings increased 20%. The feature existed before; it just hadn't surfaced at the moment of highest intent.

Most product teams bury their most important actions behind menus and “explore” screens,  optimizing for flexibility rather than direction. The research suggests the opposite, that people want to be guided. The default behavior is inaction, so make the right action the path of least resistance.

5. Onboarding sets the mental model. Full stop.

Apple Watch requires users to set ring goals before they can do anything else. This seems annoying, but it’s actually brilliant. It gets users thinking about future usage before they’ve started, which drives habit formation. 

An AdWords call center found that reps who opened with “this is a 3-month program” increased retention by 14% just by setting a realistic time horizon upfront.

What you teach users to expect during onboarding is what they’ll bring to every subsequent session. If onboarding communicates “this is fast and easy,” users will churn the moment it gets complicated. If it communicates “results take time,” they’ll stay.

6. For enterprise, the job is reducing fear, not demonstrating features.

B2B buyers aren’t purely rational. They’re humans who worry about looking bad or making a decision that blows up. The panel identified three core psychological levers:

Uncertainty aversion.  People want to know the product will work. Amazon Prime’s “cancel anytime” messaging was its single biggest growth driver, as it collapsed the perceived risk of trying. Money-back guarantees and heavy early discounts work through the same mechanism.

Regret aversion.  When close to a decision, buyers vividly imagine the ways it could go wrong. One cybersecurity company closed deals by saying: “If we fail to protect you, we pay the $5M claim.” That’s a guarantee that directly addresses the nightmare scenario.

Present bias.  The decision-maker is a person, not a company. They want to leave work on time. They want to achieve that professional win. Your messaging needs to speak to what they personally get, not just what their company gets.

7. Inaction is the default. Design against it.

A medical device company sent an email with a 50% open rate asking users to mail back a device they weren’t using. Almost nobody did. The insight: they weren’t rejecting the product, they just weren’t acting. The fix was a forced choice, “activate now or ship it back.” Suddenly, activation was the easier option.

A fintech company changed “continue or close” to “accept or decline.” Declining feels like giving something up. That reframe had identical functionally but was different psychologically, and it outperformed a $10 incentive to activate. Deadlines and closing-door language work through the same mechanism. They give people a reason to act today rather than later (which usually means never).

8. Showing immediate value can backfire.

Tech products love the “aha moment” and getting users to value as fast as possible. The research suggests this can go badly wrong. If a user imports their data and it doesn’t work well, or spends 45 minutes setting up a profile with no payoff, that’s a stronger negative signal than never having tried at all.

Sunk cost can create commitment, but only when the payoff materializes. A bad first experience with your core feature is worse than a slower path to a good one. Fix the underlying quality before you optimize for speed.