Why Google Maps Keeps Sending You Down the Wrong Street
The tools that once shaved minutes off your commute have quietly turned against you, and the reason is math, not malice.
When everyone runs the same routing algorithm, the shortcuts stop being shortcuts, and the quiet residential streets those algorithms discover fill up until they jam like the freeways drivers were trying to escape.
Here is what is actually going on. Services like Google Maps hand your GPS location to an algorithm that hunts for the quickest way to wherever you are headed, and that algorithm chases exactly one goal: get you there fast. It does not care one bit what that does to everyone else on the road (and it never claimed it would). Multiply that selfishness across an estimated one billion drivers worldwide and the whole system degrades. The apps have, in fact, gotten worse over time.
How Navigation Went From Static Maps to Real-Time Rerouting
Understanding why they broke means going back to how they grew up. Navigation lived in dedicated gadgets and dashboard units long before it moved to your phone. The earliest digital-map makers — Navteq (now Here Technologies) and TomTom — got started roughly three decades ago, and their whole obsession was the underlying data, shipped as fresh maps only every few months. That was it. No live updates, no rerouting, just a disc and a promise.
That slow release schedule had a real consequence: between updates, nothing moved. Everyone with the same disc got the same directions, and those directions stayed put. As historical GPS trace records grew and cellular coverage and bandwidth improved, developers began serving traffic information in near real time. Providers began gathering live speeds and positions from users who opted into data sharing, then integrated those speed estimates with maps, painting routes red, yellow, and green — red for likely congestion, green for unrestricted flow.
The turn came when the apps stopped merely flagging jams. Here Technologies, TomTom, Waze, and Google began offering real-time rerouting suggestions that stack live traffic on top of the road network's own characteristics. Searching every possible route proved too heavy for a driver already moving, so developers wrote algorithms that pick out a handful of candidate routes, size up each one's travel time, and pick the best.
The Selfish Routing Problem
Selfish routing — a concept economists have described since the early twentieth century — is what happens when every driver grabs the fastest path for themselves and nobody weighs the cost to anyone else. That is exactly what your phone does on your behalf. And when a network of drivers each chases the individually fastest path, the system settles into a state where no single driver can switch to a faster road, what traffic engineers call user equilibrium.
The catch is that this equilibrium is a lousy deal for the group: it is generally worse than the system optimum, the arrangement that would minimize total travel time for everyone. The gap between the two has a name: the price of anarchy, the ratio of user equilibrium to system optimum. Local optimization by individual drivers does not achieve global optimality, and the result is decreased network performance and a higher price of anarchy. Bittihn and Schadschneider have argued that app use has produced a new form of Braess's paradox, the counterintuitive result where adding capacity or choices can make everyone slower. A toy experiment by Cabannes and colleagues found that navigation apps increased overall travel time for all users, and follow-up work concluded that the apps never cured Braess's paradox in the first place.
Why Nobody Can Fix It
The structural flaw is worse than a bad algorithm, because no company has the information needed to do better. Navigation apps do not track how many users they have sent onto a given route, and just as damaging, they cannot know where competing services are routing drivers. One app's decision to divert a hundred cars down a side street is invisible to the other apps diverting their own hundreds down the same street.
The apps are also blind to consequences that fall outside their travel-time math. They do not account for whether residential streets can absorb the redirected traffic, nor for whether motorists appearing in unexpected places compromise safety. Shaqfeh and colleagues have made the deeper point: sharing real-time traffic information with every user of a network is itself sub-optimal. The very feature that makes the apps useful — everyone sees the same live data — is what breaks them at scale.
Neighborhoods Under Siege
Navigation apps have peeled growing numbers of vehicles off congested freeways and boulevards and funneled them onto quieter streets with fewer stoplights — roads never meant to carry that kind of load. Traffic jams now appear unexpectedly in previously quiet neighborhoods around the world. The effects were noticed as early as 2015.
Consider Miguel Street, a winding, narrow route through San Francisco's Glen Park neighborhood. Until a few years ago, only residents traveled it; now it is jammed with drivers cutting through from clogged Mission Street to Market Street. Residents struggle to reach their own homes, and accidents happen daily. The list is long: Sherman Oaks in Los Angeles, the San Jose Mission District in Fremont, and Boston's Dorchester, where locals gripe about cars speeding through at rush hour and drivers glued to their phones between turns.
Israel was among the first countries to feel these effects because Waze was founded there. Someone living in the Herzliya Bet neighborhood took the company to court after the app wrecked local traffic. London shortcuts once known only to the city's black-cab drivers now teem with app users.
Is There a Way Out?
The research offers a narrow and frustrating answer: the problem is worst when compliance is either very low or very high, and best somewhere in the middle. Using a susceptible–infected–recovered model borrowed from disease-spread studies, researchers found that traffic system performance improves when 30 to 60 percent of users follow dynamic routing, with the smoothest balance between congestion building and clearing landing at 40 percent adoption. In other words, if fewer people obeyed their phones, everyone might move faster.
Getting there means persuading drivers to accept routes that are not selfishly optimal, and the evidence on that is mixed. Van Essen and colleagues measured 88 percent compliance when the app advised the usual route but only 31 percent when it advised a social, less selfish detour. Perceived reliability and ease of use shape whether drivers comply at all.
Become an AutoGuide insider. Get the latest from the automotive world first by subscribing to our newsletterhere.
AutoGuide's daily news hound. I report on the latest breaking news in the auto industry across the US and around the world.
More by Marcus Lewinsky
Comments
Join the conversation
As soon as a few people get sent down the side streets and get stuck at busy intersections the algorithms adjust in real time. It's not nearly as bad as the authors make it out to be.