Context

Great, let’s establish a common understanding of just what Shared Saving rides is and how users use it. This new feature would give users a new option “Shared Saving” when ordering a ride just like they would choose Lift XL or Lux. This new ride option would let users essentially share a ride and carpool with other riders going in the same direction in exchange for a reduced fare. This means that there may be other passengers in the vehicle when it comes to pick you up or it may pick up and drop off other passengers while enroute to your destination.

To further frame this question, Lyft’s mission statement is to reconnect people through transportation and bring communities together. “Shared Saving” riders helps further this by making rides more accessible at a lower price point and giving riders from the same area a shared common experience.

Factors

“Shared Saving” rides is really positioned to make lift more accessible to more people by offering rides at a lower price point. While it may be more affordable it still may not be right for everyone so let’s list out some factors that may come into play for deciding whether or not riders would use this new feature?

1. Rider’s Budget – This is a cheaper option, but depending on the rider’s financial means they could either just pay for the familiar regular Lyft or in the opposite direction if it wasn’t cheap enough they could just opt for an alternative (subway, bus, bike, or walk)
2. Cost Savings – How much cheaper is this than a regular Lyft ride? If it is only a couple bucks cheaper then riders may not bother dealing with the hassle for just  a few bucks in savings.
3. Sociability of the Rider – Some people are a lot more extroverted than others and I could definitely imagine people who are more introverted being turned off by the fact that they would have to share a ride with a stranger and potentially even have to make small talk. There are memes online of people giving drivers 5 stars for not talking.
4. Time to Pick Up – If there aren’t any “Shared Saving” rides nearby then users are going to be less likely to select this option if they have to wait twice as long to get picked up.
5. Closeness of Other Drop Offs & Pick Ups – If users tried the shared ride but had to go 20 minutes out of their way to pick someone up then they would probably not choose this option again if time was of the essence.

1. Riders Budget – This might be tricky to confirm and quantify as we aren’t going to have exact data. With that said we could approximate the wealth level of riders  based upon a few things such as the assessed value of the property they list as their home destination, the # of rides they’ve taken in the past, the area code they live and ride in. Alternatively, we could focus on Lyft riders in Norway where salary information is public. With Rider Budget approximations in hand, I would break the riders into 10 different buckets each representing a decile and then examine the % of rides in that population that are Shared Saving rides.
2. Cost Savings – In order to help quantify the role price plays in riders decision to use the “Shared Saving” feature, we could perform an A/B/C test where the price of the shared ride is discounted by 10%, 20%, or 30% compared to the regular Lyft price. We could analyze the results of this test by looking at:
   • # of Shared Saving rides taken
   • # of users who have taken a Shared Saving ride
   • # of users who have taken more than 3 Shared Saving rides
3. Sociability of the Rider – I am unsure how much information we have in terms of measuring the rider’s sociability. One of the ideas that came to mind is that if Lyft offers? or previously offered the ability when requesting a ride to select your level of preferred conversation? If Lyft did this then we could make a pass at inferring a rider’s level of sociability then see if there is any pattern in the data confirming our assumption. Alternatively, we could explore the option of offering this “Quiet” ride preference to users choosing Shared Saving rides.
4. Time to Pick Up – There is no need to re-invent the wheel. I’m assuming Lyft has already studied to death how time to pick up affects riders’ choice to ride.
5. Closeness of other Drop Offs & Pick Ups – I don’t believe we would need to A/B test this so much as analyze the data we are already collecting. Due to the nature of carpooling, there should already be a natural variance in the additional  time users must spend en route to their destination. I would suggest we analyze the additional time en route in terms of percentage of what the non-shared ride time would be. 5 additional minutes tacked onto a ride that was supposed to take 5 minutes would be a big deal where as if the ride was supposed to be 25 minutes that wouldn’t be such a big deal.  We could examine the relationship between additional time en route and rider experience by looking at the percentage of users who used shared rides again after a shared ride that took 30% longer than a direct ride.  Additionally, we could look for a relationship between the user satisfaction poll at the end of the ride and the additional time en route. If we don’t see a statistically significant drop in retention or ride quality based upon the additional time en route we could consider expanding the distance we deem as on the way if need be in order to offer more shared rides.

Lyft wants to add Shared savings rides. What factors will you consider to determine riders’ willingness to pay? What experiments will you run to test your assumptions?

How I understand Shared Saving rides:

– Shared Saving rides – is the feature integrated into the Lyft main app. Whenever a user selects this feature, the app gives the opportunity to save money by traveling with another passenger that is heading in the same direction or to the same destination point.

First, let’s clarify the scope of the question:

– Do we know anything about the reason why Lyft wants to add this feature? Is there any specific goal that Lyft wants to achieve? For example, increasing the revenue or increasing the adoption of this feature. Increase adoption

– By saving rides do you mean an existing option or new functionality? Existing

– I assume we are talking about the mobile app, right? Yes

– Is there any known type of data in terms of willingness to pay? No

Okay, Lyft wants to increase the adoption of shared saving rides. 

User group. The main user group of this feature would be people who are sensitive to the cost of rides and want to decrease their spending on transportation. The main user group would be users of the shared drive feature.

On the other hand, there is a driver who is interested to earn extra money from each ride. But we will be focusing on the passengers mainly.

Before considering factors that determine willingness to pay let’s talk about factors that determine the cost for Shared Saving rides.

• Existence of the riders that are willing to share rides

• Existence of the rides with matching routes

• Number of riders (2-3)

• Off route to pick and drop off shared riders

• Length of the shared ride

• Time of shared ride

• The high demand for shared rides (busy area, peak hours)

These factors will affect the cost of the ride, so we can turn them into hypotheses:

• Build the route for 2nd passengers by offering passengers to walk, that’s why the cost may be cheaper.

• Schedule ride in advance. Scheduled rides (requests) in advance will create more chances for shared rides

• Repeated rides. If there is a match, let’s say min two passengers should drive a similar route (Mo-Fri) at 8 am.

I would prioritize the first hypothesis: Build the route for the second passenger by offering this passenger to walk to the preferable pickup spot, that’s how the cost may be cheaper.

Experiment: 

Run A/B test to determine WTP. I would suggest running it in one city with high demand for shared drives. On the other hand, we need to have a significant piece of data before running the experiment, so I would dig into existing data to make sure we will have enough information from one city. If not, I would consider adding more cities to the experiment.

Group A: Offer extra the walk route to the pickup point with visible price discount (economy) versus the route with off-route pickup.

Group B (control group): No extra walk time with off-route pickup (assuming that this is standard behavior)

Success metrics:

• # of completed shared rides (with more than one passenger) (D/W/M)

• Breakdown of completed shared saving rides vs shared rides in group A

• Breakdown of completed rides Group A vs Group B

• Generated revenue Group A vs Group B

• Canceled rides Group A vs Group B

• Satisfaction Group A vs Group B

Tradeoffs:

• For this experiment, we need to have a significant piece of data to make a correct conclusion (high demand for riders and supply of riders)

• We need to see if higher adoption with shared saving drives generates more revenue for the Lyft

• On the other hand, walking distance should be significant to affect the cost of the ride. The saving for the ride may be too low or on the other hand, the walking distance may be too long

• Pre-planned or scheduled rides may increase low demand and make shared drives happen.

Define shared saving rides are?

Assuming it is a way for the riders to save money by sharing rides with other passengers. The riders will be added as the ride begins and may or may not include additional passengers.

Factors that determine the customer’s interest in using shared saver:

1. The price of the rideshare compared to the full ride
2. The expected time they will lose
3. The drop in average ride rating and thus impact on customer retention
4. Average wait time increase for a ride
5. Priority order of the drop-off. Do people get dropped off in the order they were picked up?

In order to understand the impact of each of these, I will have to come up with a set of hypothesis:

1. Hypothesis 1: Users will retain same or better by x% if we can reduce the price per ride Y% and not increase the time to drop a user by Z%.
2. Hypothesis 2: Users will retain same or better by x% if we can reduce the price per ride Y% and not increase the time to drop a user by Z% and ensure their order of dropoff doesn’t matter.
3. Hypothesis 3: Users will retain same or better by x% if we can reduce the price per ride Y% and not increase the time to drop a user by Z% and ensure their average wait time doesn’t matter.
4. A few more hypothesis that I will define with the help of a data scientist to define to validate the independent and dependent variables in the factors mentioned above

Validating hypothesis

Once I have defined such a hypothesis, I will work with the data scientists to design several multivariate experiments and start running these experiments on a market by market basis.

I will wait to get statistical significance and then understand what are the actual factors that affect a rider’s preference and whehter this is a good idea or not?