Peril’s of Lyft’s Wait and Save Algorithm

A therapeutic rant

Hi followers. I’m writing this post from the back of an Uber, driving in truly terrible blizzard-like conditions from White Plains to Brooklyn NY. It’s not something I’d recommend doing at 2 AM on a Tuesday morning.

There are a few components to this problem that should be noted (we’ll break down the last one)

• Issue number one: Someone (it was me) tried to get around a red eye and moved a flight up without checking the weather.
• It’s snowing, and despite coming from Colorado, I have not yet bestowed my master plan on the great state of NY for improving snow infrastructure.
• Lyft’s Wait and Save Algorithm doesn’t seem to consider predicted driver density when allowing users to opt for that option.

That last bullet point is poorly written (2 AM brain setting in), but is easily explained. First, a lil backstory.

Lyft offers users with above average patience the option to wait and receive a reduced rate for a trip. In my case, I was provided with the option to:

• Pay $150 for a prioritized booking that would arrive in 1–2 minutes
• Pay $130 for a booking that would arrive in 10(ish) minutes
• Pay $60 for a booking that would arrive in 45–60 minutes.

These options all provide great incentives to riders. I had a Last of Us episode to watch, so the third option was the clear choice for me. I put myself at the back of the line, and in doing so, allowed Lyft to book more expensive rides ahead of mine! It’s a good economic solution, assuming everyone gets a ride.

A slight bump in the road

The issue lies when Lyft claims a user will be able to get a ride, but cannot deliver one. After the waiting period, Lyft tried to match the waiting user with a driver. If there are none available, Lyft cancels the request entirely without warning. This is particularly problematic because it causes a catastrophic incentive loop:

• User waits, expecting to spend $60
• No drivers are available after the waiting period! Unannounced, Lyft cancels the ride request
• User can either book a new ride for (potentially) the same original fee ($150) or wait another hour.

If Lyft can make me wait an hour and go unpunished for a lack of vehicle delivery, there’s no reason for them to actually give me a wait and save vehicle. They should just make me wait and then get me to request a normal ride at a potentially elevated rate.

I don’t actually think any Lyft product person is thinking like this, but given that it goes unchecked, it’s a pretty questionable feature.

How should they solve this?

Lyft ultimately needs a better algorithm to decide whether or not providing a Wait and Save option is a viable choice. Not having it is a great option — no time is wasted and users can make other plans. When it works, it’s also great! Better predicting when it will be able to pair up a driver and a user is the trick here. A few things I’d love them to consider

• Ride request density compared to average
• Driver density compared to average
• Weather conditions, time of day, road closures, and other external conditions that seriously impact a user’s likelihood of being matched with a ride
• All of the above, but considered as a rare of change over the last hour. If there’s a 25% increase in ride requests in the last 30 minutes, Lyft should expect that trend to continue and base their decision to show me options on it.

Outside Checks

Though less than ergonomic, I’m writing this on my iPhone mini in an Uber. If Uber didn’t exist, this would be a far drearier post. That is all to say, I was empowered to choose another option the whole time. I’m not suggesting Lyft is a monopoly at all, nor am I trying to suggest that I was forced to use their service.

Still, much like flight cancellations, I paid for a booking and was unable to use the product I paid for. Perhaps it’s time for Lyft and Uber to start providing some equivalent to Airbnb’s AirCover — charging users an extra 10 cents a ride might help with this. If Lyft fails to book me a ride that they claimed was available, they can give me an Uber or taxi credit. (Much like Airbnb might give a hotel credit!)

Three quick product fixes

1. An improved algorithm that shows / hides the Wait and Save option
2. Messaging that alerts active Wait and Save users of low driver density, perhaps suggesting that they pay more to book right then.
3. AirCover style ride booking insurance that results in transferable credits to Public transportation, Uber, or a local taxi service.

Look, I work in product. I like to build things. So a few quick disclaimers, based on my own career experience

• There’s almost a certain chance that I don’t know how Lyft’s Wait and Save feature works behind the scenes
• It probably works ~99% of the time

Still, it’s important to improve that 1% of rides that users will remember for 100% of rides going forward. Next time, I’ll be Ubering.

I (very occasionally) post product posts like this and sometimes tweet things, @GGarnhart. Say hi!