I've written before that product-led growth only works if the product genuinely leads, without a human propping it up. Building Ovela taught me the deliberate exception to my own rule: there's a category of decision scary enough (medical travel to another country, with real money and real health at stake) where self-serve isn't a growth lever. It's a trust deficit wearing a clean UI.
What the self-serve version actually looked like
Our first instinct, informed by every self-serve product I'd built before, was a filter interface: pick a specialty, compare accredited hospitals, choose a price tier, book. Efficient, scalable, zero marginal cost per user. In early testing, it was also where patients quietly bounced, not because the interface was confusing but because the decision underneath it was too high-stakes for a filtered list to be reassuring. Nobody wants to choose a hospital in another country the same way they choose a hotel.
Self-serve removes friction. For a low-stakes decision, friction is the enemy. For a high-stakes one, a little friction (a human who clearly knows the terrain) isn't friction at all. It's the only thing that makes "trust me" credible.
What the concierge model actually does that self-serve can't
- It absorbs the ambiguity a filter can't. Patients rarely know how to translate "I have $4,000 and I'm scared" into the right combination of specialty, tier, and hospital. A concierge does that translation; a filter asks them to already know the answer.
- It transfers reputation, not just information. A self-serve list of "accredited" hospitals is a claim. A concierge who's personally vetted the specialist and says so by name is a relationship, and relationships are what high-stakes decisions actually run on.
- It scales differently, not less. We didn't abandon scale. We scaled the concierge's judgment through tooling and playbooks, rather than trying to remove the human from the loop entirely.
Starting with one specialty, dental, and a concierge-led match, instead of a broad self-serve filter across every specialty, is what let Ovela compete on trust instead of price in its first year.
How I decide which model a new product needs
I ask one question now before defaulting to self-serve: if this decision went badly, how long would the user privately blame themselves before they'd trust a system to help them again? A wrong restaurant pick costs an evening. A wrong hospital pick, in a foreign country, costs something nobody fully recovers trust in a system for. The scarier that answer, the more a human belongs in the loop, not as a stopgap until the product "matures" into self-serve, but as the permanent, correct design.