How would you design surge pricing for Uber or Lyft?

Context :  State , current problems , why ?

High demand during evening hours , Less supply of drivers .

How does uber works ?

 

 

Risks ?  many users will find it expensive

 

Goals ?

- Maximize revenue.

- Incentivize Drivers to be available

- Supply and demand .

 

Who ?

Drivers

- get notification on app -- opens app - check price and drop location,   

Customers

- Enters drop location -- check price -- book - wait -- onboard -- journey -- reached -- feedback .

Platform

 Earns money - percent of commission , More rides or hiigher AOV = more money .

 

Problems :

Users book in peak hours but dont get rides . No availability of drivers .

 

- Surge pricing

Total no of ride request receive - given point of time

Total no of drivers available at any given point time .

Base fare + Surge fair = Total Fare .

Total fare = Function ( delta of ride request to driver available )

Function = Combination of Demand and Supply

 

 

Metrics to measure :

- Overall Revenue during surge

- Drivers  and customers not matched

-Users dropped due to surge

- Incremental revenue increase during surge.

 

Users NPS

Driver NPS

Uber? mobility marketplace to move people and goods from one place to another.
Which place are thinking to design the surge pricing ? Worldwide

Why do we want to do this? Because during peak hours there is too much demand that we are not able to serve this demand. We see this opportunity to balance the supply and demand through surge pricing.

Do we want to design the algorith or user experience ? Both

Goals of surge pricing :
1. To incentivize drive to come online so as to increase supply
2. To reduce demand by increasing the prices

Algorithm :
 surge pricing might depend on?
1. No of user demanding for cab (cab demand) [Directly proportional]
2. No. of cab drivers ( Cab availability) [ Indirectly proportional]
3. Time of the day ( Day / Night) [Day less , Night more]
4. Traffic conditions ( Long time of the journey) [ Peak traffic more, otherwise less]

First and foremost, I will define the baseline supply and demand using the hisorical data. This baseline will be then calculated realtime during each scenario to maintain its relevance based on data new data.  The demand supply curve will keep on getting calcualted based on real time. Based on the difference between baseline and current, a multiplication factor will be added to balance the equation.

User journey-
1. User puts an intent to ride
2. System calculated the baseline and realtime supply-demand. Based on these calculations, system will rflag the ride as surge ride.
3. User will be trasparently shown about surge pricing.

Driver -
1. Whenever there will be surge applicable, drivers will be notified.

2. They will see the earnings that they can earn through surge clearly

How would you design surge pricing for Uber or Lyft?

 

Clarifying the scope of the question.

 

• Are we talking about any specific part of Lyft’s business (carsharing, e-scooters, rentals) - carsharing)? Carsharing (hailing)

• What exactly do you mean by surge pricing? ( Is this something called Primetime?) I assume it’s a price change during busy hours.

• Do you have any information about how Lyft has priced rides recently?

• Do you have any information about how the competitors are pricing their rides?

• Is there any particular objective you’d like me to keep in mind?

• Is there a particular user segment you’d like me to price the product for?

Framework:

1. Describe the product and industry

2. Define the objective

3. Choose the strategy

4. Suggest the price for the product

5. Summary

Describe the product and industry.

• Lyft is a carsharing platform that allows commuters to get from point A to point B. On the other hand, drivers can earn money by utilizing their vehicles and providing driving services to commuters. So as I recently mention, there are two main user groups: Drivers and riders.

  • Drivers can be grouped by vehicle type and also by usage frequency:

    • Frequency:

      • Frequent drivers - working for Lyft as a main source of finances ~ 40 h/week

      • Sometimes drivers - working as a driver additional income (weekends and few drives during the week)

      • Rare drivers - do it for extra cash, not regularly

    • Type of vehicle

      • Regular cars (sedans, budget)

      • Mini Van

      • SUV

      • Premium

      • Luxury

      • Trucks

  • General commuters - can be anyone in the range of 16 -....., but I think the ICP would be the person in the age of 21-60, who need regular commuting between home and work or school, or anything else. 

    • Regular users 

    • Sometimes users

    • Rare users

    • Another user group is business users, those who do business rides with   

  • Currently, Lyft has around 30% of the market share in the US. Along with the company’s mission “Improve people’s lives with the best transportation ” the main objective for the company is to expand market share and become a profitable company

• On the carsharing side, the cost structure consists of:

  • Incentives for drivers

  • Operating costs

    • Product development and support, infrastructure

    • Marketing and PR

    • Administrative personel

    • Office and procurement

    • Utilizing 3rd party providers (Google maps, payment systems, etc.)

    • Agencies and partnerships

    • R&D

  • Lyft is on IPO since 2019 and it shows solid growth starting in autumn and this indicates the overall good health in company growth and good image for investors showing $58 per stock.

  • Lyft is a commercial organization and as far as I know from the annual report it is not profitable, but if the goal is to become profitable by 2021. First of all by an increasing margin, secondly - optimize costs. Moreover, Lyft provides not just services, but overall simplifies the experience for commuters, for example, the launch of Lyft Pink subscription and connecting routes public transit and all possible transportation. This all proves the aim to simplify commuters life

• The competitive landscape. At this moment the main competitor on the car-hailing side is Uber with a 70% market share in the US. Uber operates globally and has coverage in a bigger amount of cities in the US. Lyft focuses on North America ( On the other hand surveys show that Lyft driver satisfaction is higher than Uber’s. Also, it’s important to mention there are smaller carsharing services like Via and others, but currently, they don’t play a huge role in the competition.

• Any competitive advantages (e.g. cost structure, distribution channel, talent, market share, partnerships) Lyft has a higher satisfaction rate among drivers, which gives Lyft an advantage among drivers to stick to them. Also, Lyft provides not just services, but overall simplifies the experience for commuters, for example, the launch of Lyft Pink subscription and connecting routes public transit and all possible transportation. This all proves the aim to simplify commuters’ life. With a 30% market share, Lyft also operates an e-scooter business, which also helps commuters with short distances and places without public transport. Another advantage is a partnership with Sixt, which gives a chance to commuters to rent a car via the app for long-distance drives, road-trips via a wide network of Sixt rental services. 

 

Define the objective

Considering the information above I think the objective for Lyft should be maximizing the profit for the ride-hailing business during surge hours.

For Lyft, it might be a solution to the strategic goal to become profitable by the end of 2021. On the other hand, increasing the price during Primetime allows Lyft to increase the incentives for the drivers, so they could be more interested in switching to Lyft, or using it during prime time. This also gives Lyft users larger demand for cars and possibly less time to wait. 

Choose a strategy

For the chosen strategy I would focus on the Frequent riders, that commute from work-home, which usually happens during peak hours (morning and evenings)

For this type of user and the company objective, I think Premium pricing would be a good choice because it gives users higher demand for cars and less time to wait, and for this use-case, they need to pay more. 

Also to support the objective I would suggest using an additional promo of Lyft Pink that gives discounts for the rides for the users. On the other hand, it may increase the revenue from subscriptions. For this we need more deep investigation and consider the following factors:

• The discount for the users

• Will this initiative profitable with Lyft Pink subscription?

Suggest a price for the product

The regular price per mile is $1.6, however, the ratio for the busy hours is an unknown value, because it depends on the area and demand at the pickup areas. The simplest solution would be to increase this ratio by 10-20%. I can think of the following coefficient tiers:

• 1.2 somehow busy + 10% = 1.32 ( $2.11 per mile) 

• 1.4 busy +15% = 1.55 ( $2.48 per mile) 

• 1.5 extremely busy +20% = 1.7 ($2.72 per mile)

Summary

• For designing surge pricing for Lyft I would suggest focusing on regular commuters that usually commute during busy hours.

• For the objective is to maximize the company’s profit I suggest the premium pricing strategy during busy hours - increasing the coefficients by 10-20%.

Tradeoffs:

• Price increase during busy hours may affect customer churn, because they may choose Uber for cheaper rides or switch to ridesharing. 

• On the other hand, it will allow attracting more drivers (by increasing incentives for them), but if demand will decrease this may affect drivers and they will churn too. 

It’s important to launch the A/B test with surge pricing comparing it to the control group before releasing new pricing.

 

1. Are you looking at the UI design & user flow for the feature or a high level algorithm to decide surge ? both
2. In terms of target segments, are we targeting all users in all our markets or are we leaving a certain segment ? Yes for all users of Uber & Lyft
3. Is it okay to ignore difference between uber for business & uber for personal use for this analysis ? yes
4. Also I would ignore Shuttle pricing for this analysis. Okay

1. Maximise revenue for Uber 
2. Maximise revenue for drivers
3. Bring more drivers on road if capacity is constrained
4. Provide service to people who intend & can pay higher for the service

1. First & foremost, surge pricing is available only when demand is high & supply is low. So we first need to measure demand & supply. I would capture data like this : Date | Day | Time | Sector Name | Demand in Sector | Supply in Sector | Destination Sector | Price. I would keep on populating this data automatically. Ideally destination & supply sector should just be big enough to impact the demand.
2. Second step here is to figure out the demand curve for each sector & destination. Demand would be function of date, day, time & price. To create the equation for demand curve, we will need to alter the prices periodically to get enough data. We can take help of data scientists & statisticians to give us the demand equation. The capability we will need at backend is that we should be able to change prices for each sector to get enough data to create a demand curve. Ideally this demand curve won’t be static, so if we can create an automated way pf creating demand curve it should help us in scaling faster. The idea is that we should be able to predict the demand with a sufficient accuracy.
3. The third step is to create the supply curve. This would again be a function of date, day, time & price. The idea is that we should be able to predict the supply with a sufficient accuracy.
4. Once we have the demand supply curve for a sector, we can now vary the price to bring the demand & supply under control. If we draw a graph of demand & supply for a sector for a time, the area under these curves would be revenue.
5. We will then define thresholds for starting & stooping surge. Say for eg if Supply > Demand by 20 cars remove surge and vice versa. These limits have to be exprimented with to come up at an optimal number. 
6. Surge pricing has lots of ethical aspects to it but I would make sure that the surge is limited to say a maximum of 2x. Very high surge can lead to customer backlash, legal issues & can spoil the market
7. The idea here is to get good at predicting demand & supply basis price & we will need to use some sort of predictive analytic techniques here. Equation view mentioned above is a simplified view in the sense that equations may change over time. So we need a system which creates a prediction of demand and supply may be by using ML and then accounts for external factors like say for eg weather etc using AI.

1. We need a notification to be send to drivers once surge is going to be applicable so as they can come online. Notification needs to be localized to the source sector. 
2. We need to send a notification a little ahead & not just when the supply has come to zero. So a feature like if 10 cars remaining, send the surge notification.
3. To prevent drivers from gaming the system, we need to put some rules for eg that a driver can’t turn on or off the availability more than X times a day else drivers will make themselves unavailable to get surge pricing
4. Driver needs to be informed once surge pricing is removed.
5. From user perspective, once the user puts source & destination, we should inform the user of surge.
6. For the users who can wait, we should inform the users who tried to make a booking but didn’t do because of surge that surge is now removed so as they can book if they need.

1. Instead of perfecting demand & supply curves, we can just play with surge & keep on experimenting with values for each sector. This is a quicker but not so accurate approach.
2. This model can be enhanced for a user basis users Willingness to Pay. Say for eg user A has paid 2x surge in the past but user B only pays 1.2X surge, so for user A we always show 2x surge & for user B always show 1.2X. This model will maximise revenues but ethically may not be the right thing to do.