Auctioning Cab Rides: An Alternative to Surge Pricing?
Donned the hat of a consultant for my terminal project on Information Economics during my MBA, along with my study group members. We analysed Uber's current model and recommended a First Price English Open Auction model using inspiration from PageRank’s Markovian algorithm to solve multiple information asymmetry issues. Writing the crux of our proposal here.
Executive Summary:
Cabs are generally hired - be it the standard yellow taxi-cabs of old or the new-age cab aggregator apps like Uber that leverage mobile and technology to map cabs to the customer. However, these regular cab-hiring models come with their fair share of marketplace problems. In this project, we take Uber as an example of present day cab service and look at how the introduction of auctions into the process of cab hiring can solve the standard issues of transparency, matching market demand and ultimately debate if auctioning cab rides can be a viable alternative to the much disliked surge pricing. With just the right amount of gamification, one could end up bidding for cab rather than just hiring it! Uber is a ride hailing platform. Valued over $72 billion with an asset light model, it has heralded a new era of market economics.
The two biggest problems plaguing the Uber of 2019, pre-IPO are:
• Transparency (reducing information asymmetry)
With major PR problems on both rider and driver sides - pricing clarity and revenue clarity respectively, the core value proposition is itself in question. Is Uber a platform or a taxi service? This needs to be clearly elucidated since this has been the center of most lawsuits against Uber.
• Maximising economic (consumer & producer) surplus
Allocation of demand and supply through efficient matching of riders and drivers - maximises economic surplus. This benefits the riders, drivers and Uber (all three stakeholders) by enhancing the aggregate pie. With increasing competition across countries, Uber is facing a real threat as Lyft has already one-upped by announcing an IPO earlier in USA. With Didi acquiring Uber in China, lawsuit woes in the EU, among many more, Uber is really in a quagmire and needs to up its game. We propose auctions as a solution to tackle the above two problems. We analyse the First Price English Open Auction in this report specifically. This will entail a base fare (adjusted for demand) and an incremental bid determining the likelihood of a user matching with a driver. Present Scenario
Surge Pricing Model
Uber’s current model is Economics 101; shifting fares until demand and supply curves match for general equilibrium to be fair to partner drivers and users both. The pricing algorithm pings for demand data on the Uber app every five minutes and updates it to change the pricing surge accordingly. This data is visible to the rider and partner driver as well.
However, surge pricing can work in any/all of the below mentioned methods:
• Reducing demand for cabs (fewer people desire a cab for an increased fare)
• Generating new supply (providing an incentive for new drivers hit the roads)
• Moving supply/drivers to areas with higher demand
Limitations Thereof
• Surge price results in fares altering awfully frequently and is highly localised
• Instead of attracting more drivers on the road in the short-term, Uber’s surge pricing reduces driver supply in adjacent areas[1].
Thus, surge pricing appears to push drivers already on the job toward neighbourhoods with more demand and higher surge pricing, instead of bringing more drivers out on the roads.
Proposed Solution
Alternative Auction Model
This model will function as a First Price ascending bid auction. It will let the users bid in fixed increments on top of the base ride cost, increasing the likelihood of getting a confirmed ride when demand outstrips supply.
Features
In order to keep the user experience as quick, smooth and streamlined as it is now, there will exist certain parameters which a user will be able to define as his/her preferences for the auction model.
Mechanism
Inspired by the Google Ads Auction System, the customer will be matched to a driver, not only based on the relative standing of his bidding price but based on an aggregate rank calculated using the following four attributes
• Bid price: The price which the customer is willing to pay for a ride; basically, higher the bid, the better it is.
• Customer Ratings: A score on the scale of 5 calculated as an average of all the ratings the customer received on his /her previous Uber trips. Again, the higher the rating, the better it is.
• Number of Trips: The total number of trips taken by a customer with Uber worldwide. Again, higher the better.
• Minimum distance to the cabs: The customer’s minimum distance to the available cabs. This factor needs to be as low as possible to ensure proximity of cab. This is considered to give optimal weightage to the proximity of the user to the available cabs.
Service Improvement: How the proposed model makes cab-hiring better
The process of cab-hiring, rather cab bidding is enhanced with the implementation of the proposed first bid auction mechanism due to the following factors:
• Price Discovery: The auction model allocates rides to the users at the prices decided by the value customers associate them with rather than decided by the Uber’s internal surge pricing mechanism. This also empower the customers in assigning their own value to any ride. While in the present scenario, the user has an option only to reject or accept the offered ride at the shown price.
• Transparency: The auction model will make Uber’s ride pricing process much more transparent as customers will now be able to see the aggregate demand at their offered bid price and dynamic probability of securing a ride.
• Increased Market Thickness: As the customer feels more empowered & has an enhanced user experience, a greater number of customers will be attracted to Uber’s platform. As the customers increase, with positive cross-side network effects, number of drivers will also increase. This increase on both sides of platform will lead to increased market thickness with an enhanced experience, which marks impeccable market design.
• Increased Retention: As the customers’ aggregated rank considers the number of past transactions, this will act as incentive for the customer to stay loyal to Uber’s platform & not multi-home with other ride-sharing services.
• Alleviate Congestion: Customers, who are more price sensitive, will try to minimise their cost by participating in the bidding earlier than when they actually need the cab. This will lead to better congestion alleviation than as happens with the current surge pricing model.
Potential Downsides of Proposed Model
Though there are very few potential downsides of the suctioning model, it is worth discussing them in context of the application in Uber’s platform.
• Disutility from booking a cab: As the customer needs to participate in the auction every time he books the cab, there might be a disutility for the customer. To overcome this weakness, we propose adding a feature, wherein the customer can pre-empt and provide his default desired probability of getting a cab in the Profile Settings. Once updated, the system will automatically bid on customer’s behalf according to the preset probability, simplifying the process.
• Increased time for booking: The total booking time might increase due to the introduction of auctions. However, with the enhanced experience, transparency and ultimately satisfaction, it essentially boils down to a trade-off between increased time & more transparency/customer empowerment.
• Emergency Situations: If a customer needs a cab during an emergency, he might not want to wait & participate in the auction. This might lead to attrition. To circumvent this, the customer may be presented an option of choosing a price at which a cab is guaranteed.
Recommendation / Implementation
Uber’s goal is to be as close to a marketplace with perfect competition while maintaining service quality. The latter is a separate problem, targeted through proper on-boarding and star ratings. We have targeted the former problem through an auction model. This should be implemented through a native app-based interface which will receive live inputs from users, calibrate probabilities and match via a ranking algorithm. The ranker, taking inspiration from PageRank will involve four inputs as outlined above. The bidding-based system will be released in subsequent app versions in a phased manner (at select locations where demand greatly exceeds supply) for A/B testing to observe changes until a full-fledged release. There are certain risks as outlined above but the proposed solution is a more economic viable solution in the long term, increasing transparency and efficiencies - near perfect competition in an impeccable market design.
Reference: [1] How Uber surge pricing really works, The Washington Post, April 2015