SafariShare · 14 October 2021 · 3 min read
Every time a passenger confirms a booking on SafariShare, they are agreeing to spend several hours in a private vehicle with someone they have never met. That is a significant act of trust. The platform's job is to make it reasonable.
We think about this constantly. Here is where we have landed.
You cannot add trust the way you add a button. Trust is built through repeated positive experiences and eroded by a single bad one. What a platform can do is create the conditions for trust to form.
Those conditions have layers. At the base is identity: do I know who I am dealing with? Phone verification is the minimum. A real photo, a full name, and a visible history of completed trips build more.
Above identity is track record: has this driver done this before, and what did people say about it? A driver with fifteen completed trips and consistent positive reviews is a meaningfully different proposition from someone who registered this morning.
Above track record is accountability: if something goes wrong, what happens? Does the rating I leave affect this driver's ability to book in future? Is there someone to contact?
SafariShare has all three layers. None of them are finished.
Phone verification has been mandatory since launch. Profile photos became required for drivers early on — we found that passengers are significantly more likely to book with a driver whose face they can see.
Trip history is visible on profiles. The review system lets both sides leave feedback after each trip, visible to future users making decisions about who to travel with.
None of this solves the trust problem. It creates the infrastructure for trust to grow through actual experience.
Review completion rates are lower than we need. Most people, once a trip is over and they have arrived safely, have moved on. The end of a journey does not naturally invite the friction of leaving a review on an app.
We have tried prompts and reminders. None of them have fully moved the needle. The review system is more useful than it would be without it, but it is not yet as rich as it needs to be to help new users make confident decisions.
Safety incidents on the platform have been low. But low is not zero, and each one teaches us something.
The incidents we have seen have been almost entirely logistical: miscommunication about pickup points, departures without notice, a passenger arriving at the wrong location. Not the driver-passenger interaction itself — that is almost always fine — but the coordination around it.
We have made pickup point confirmation more explicit and added a way for drivers and passengers to contact each other through the platform before sharing personal phone numbers.
Trust is the product, not a feature of the product. Everything else — the route matching, the payment flow, the reviews, the support — exists to make the trust between two strangers more reliable.
We are still working on it. The platform today creates better conditions for trust than it did in 2019. It is not yet as good as it needs to be.