Uber and Lyft’s “Batch-Matching” Markets
https://mashable.com/article/uber-batch-matching-ride-hailing-request/
In class, we extensively discussed the concept of matching markets–situations where people are matched with goods that they have preferences for. Uber and Lyft are really good examples of the matching markets model. Each rider needs to be matched to one driver, and they all want the shortest possible wait time.
The Mashable article discusses how this matching process works in the two different apps. According to the article, Uber’s method for matching riders to drivers has changed over time. They first matched users based on distance, then switched to a time-based matching, and then they rolled out what is currently used: a strategy the company refers to as “batch-matching.” Rather than trying to match individual users to their closest ride, batch-matching evaluates a group of nearby users and drivers and tries to optimize waiting times for the entire group. The article also explains that Lyft uses a similar format, although their system has the added quirk that if you have given a nearby driver a low rating in the past, you will not be matched with that driver, regardless of how close they are to you.
These explanations of the matching systems used by Uber and Lyft are highly relevant to ideas we have discussed in class. We have talked at length about perfect matching, and Uber and Lyft each creates a perfect matching in a unique way. They use “batch matching,” where they analyze multiple drivers and riders at the same time, which creates a matching market like the bipartite graphs we looked at in class. Each company formulates the rider’s valuations in a slightly different way, but overall they both seek to maximize social welfare rather than necessarily provide any single user with their best option.