We faced this problem of dynamic pricing as students in Pittsburgh quite often. Thus, we thought to dig deep into it and took up this project!
Examined data and API's made publicly available from Uber and collected data from events in the Pittsburgh area for analysis.
Analysed the data using technologies such as R, Python and .looked at various trends and assessed causality
‘Surge Pricing’, the idea of raising the taxi fares for riders when the demand for rides exceeds the supply of drivers in the area, at least that’s what is claimed by Uber and hence commonly believed. A common pattern of shifting the market prices observed in several industries, such as, flight tickets for a long weekend trip or even a Super Bowl final ticket. Surprisingly, not only Uber riders, but also Uber drivers/partners are concerned about the surge pricing. According to a questionnaire, we came across on Quora and our interviews with Uber drivers . In spite of a surge in certain areas, some drivers prefer not to drive to the location with a higher surge unless the new location has a significantly higher surge. One of the plausible reasons could be more than frequent changes in surge, making the trip to a higher surge area, more than often, not worth it. Another perspective could be the miscommunications between Uber and riders about surge prices. Riders are sometime unware of a surge price when requesting an Uber. Drivers feel embarrassed when they charge customers higher multiplier.
The motive of this project is to:
1) Examine the pattern of Uber’s surge pricing strategy by performing various types of analyses on
Uber, metrological and city-events data. We will be using R to perform the exploratory analyses
(in addition to Tableau) and build the prediction mode using R.
2) Build a predictive model that predicts the Uber surge at a given time and location.
3) Provide recommendations for an alternative pricing strategy that can help mitigate the negative customer/partner sentiment.
We created a short video to better address the motivation for this project as well, do check it out!
Because this research paper is based heavily on both qualitative and quantitative insights, data
collection methods played a very important role throughout the process. The data that feeds into the
predictive model is extracted for three purposes:
1) Analysis of Uber Surge Pricing
2) Analysis of effect of weather on the Uber Surge Pricing
3) Analysis of effect of city events on the Uber Surge Pricing
The figure below displays the several layers of operation of our data collection methodology
We checked for multiple locations and cities and noticed that the surge price for the product lines in a neighborhood which have negative net supply, have the same surge level. Cannot be a coincidence that they are the same all the time. Below is the plot for a Monday night in Pittsburgh.
We believe that Uber does this so that riders do not take advantage of the different surge prices across different product lines at the same point of time. Let us take a scenario where we have a very high demand for UberX, say a surge 5x and very less demand on UberXL which would bring the surge to 1.5X only. Now, the customer can take the advantage of using UberXL for a much cheaper price than paying 5X on UberX. However, this becomes unfair to the riders who want an UberXL. Now they have to pay 5X instead of 1.5X surge which is the actual demand of UberXL. How far is this strategy ethical? We do not know yet.