A UM-Dearborn student researcher is out to make a ‘better GasBuddy’

September 6, 2021

More powerful fuel searching apps could save consumers money — and have big applications for autonomous vehicles.

A white man’s hand is seen holding a gas pump and filling the tank of a red vehicle.
Credit: MPCA Photos via Flickr/Creative Commons

The goal of fuel price tracking apps like GasBuddy is ostensibly to help folks save money by providing up-to-date price information for local gas stations. That idea sounds straightforward enough. But UM-Dearborn data science and computer science junior Shouryan Nikam says that approach is really too simple if you want to save folks money. Setting aside the issue of GasBuddy’s sponsored results model, which can skew the listings consumers see, Nikam says the main problem with GasBuddy is that it only provides gas station locations and prices; but it stops short of providing consumers an actionable recommendation. Hence, it doesn’t answer the fundamental question of whether traveling a few extra miles to a station with less expensive gas is actually worth it. That decision is left to the consumer, who has to make an on-the-fly choice based more on hunch than number crunching. Moreover, Nikam says if you’re only taking into account fuel price and distance, you’re leaving out relevant parts of the calculation, like maintenance costs for those extra miles traveled, and perhaps more importantly, the value of your time.

All summer long, Nikam wrestled with these complexities as part of his research in the Summer Undergraduate Research Experience (SURE) program. Overseen by Assistant Professor Junaid Farooq, the project had a couple different objectives. First, Nikam set out to answer the question of whether this consumer quest to find the best price is even worth it, and if so, under what circumstances. To do that, he built models for 10 geographically diverse American cities, taking into account variables like fuel price, distance to a gas station, insurance costs, maintenance costs, time spent searching, type of vehicle, and whether a consumer was part of a brand loyalty program. Among the top line findings so far: It’s generally more advantageous for drivers of larger vehicles to travel a little farther for cheaper gas — owing mostly to the larger size of the fuel tanks. “So depending on the fuel price difference between two gas stations, it might be worth it for a large SUV to travel 6-8 miles. But if you have a small car, that range shrinks to 2-4 miles,” Nikam says.

 UM-Dearborn student Shouryan Nikam pumping gas at a gas station.
UM-Dearborn student Shouryan Nikam pumping gas at a gas station.

The data are pointing to some other interesting conclusions too. First, geography matters: The maximum “worth it” distance is larger in less dense areas, like small towns and sprawling suburbs, and smaller in dense urban areas. Nikam says this is likely because cities typically have lots of gas stations packed tightly together, which creates more competition and smaller price differences. Also, loyalty programs seem to offer only modest benefits when it comes to fuel search. “They really only matter a little bit,” Nikam says. “And they matter less for smaller vehicles. If you’re in a dense city, they matter even less.”

Next up, Nikam says he’d like to take these models and develop user-friendly software that does the one big thing GasBuddy doesn’t: provide a data-powered recommendation for which station has the “best” price, all things considered. And while that could be a valuable app for consumers today, Nikam says the really interesting application is a future one. “Once autonomous vehicles become more widespread, an AV’s decision engine may decide to just go out and get gas on its own, based on a change in the price or the time of day. That’s pretty exciting, because this is a framework that could be used by any autonomous vehicle manufacturer.”

Notably, Nikam says their models work best right now for gasoline-powered vehicles, since that’s still the biggest fuel source for cars. But they did build a model that applies to electric vehicles. The more dispersed charging infrastructure for EVs means that, currently, their model is mostly useful for EV drivers taking long road trips. But should vehicle fleets go electric quickly, their app could be just the thing to help handle the dynamic charging environment experts expect to be the norm in an era of renewable energy.

###

Story by Lou Blouin. If you liked this article, you’ll be excited to hear we’re covering 2021 SURE projects all month long. And check out our preview to this series, which explores how SURE is growing the student research culture on the UM-Dearborn campus.