Here's how Kevin and I optimized notifications to make 
@airbnb $10+ MILLION in <25 lines of code 

thread 👇 (1/9)

building getroi.app. prev @ airbnb, google

In 2020 I worked at Airbnb as an engineer building the new host inbox. Over 5M hosts use it every day to accept & manage bookings.

Diving further, we saw that most of these retracted RTBs (request to book) were for reservations checking in the same or next day.

Guests would retract in <2 hrs while hosts only saw it 6+ hrs later. Instead of re-booking on Airbnb, guests likely booked elsewhere.

It made total sense.

Imagine going to a new city and trying to find last minute accommodation. You're in a rush- so if the host doesn't confirm your booking quickly, you'll try booking somewhere else.

But why did hosts not even see the request before the guest retracted it?

Turns out, we were sending notifs to the host 3 times:
1. On the initial request
2. 6 hours later
3. 18 hours later

All with the same content regardless of the urgency.

Even though we sent one instantly, hosts likely didn't see it or realize that it was urgent upon reading it.

We wrote <25 lines of code that did two things:

1. If the booking request was for the same/next day, we sent reminders 15 & 30 minutes later by push & SMS

2. Made the content reflect the urgency:
"Brian wants to book your home" ❌ → "Brian wants to book your home for today" ✅

The results? A HUGE conversion increase in booking requests on our platform. Especially amongst professional hosts who manage 5+ listings.

We were now able to capture millions in lost bookings just by optimizing the send time, frequency, and content.

For those of you working on product engagement loops through notifications, I hope this was helpful.

Let me know if you liked this and I'll share more stories about notification optimization at Airbnb and how impactful it was for the messaging product.

While analyzing our messaging data, we found something mind-blowing:

Over 10% of booking requests were retracted by guests BEFORE the host even saw it.