WHAT HAPPENED TO GOOGLE MAPS?
Surprising Changes to Google's Cartography
Browsing Google Maps over the past year or so, I've often thought that there are fewer labels than there used to be. Google's cartography was revamped three years ago—but surely this didn't include a reduction in labels? Rather, the sparser maps appear to be a recent development.
A few days ago, I was looking at some screenshots I used for a post in April 2010. I've posted one of those screenshots below alongside a screenshot of Google Maps today:
⚠️ NOTE: Tap or click any image to see an enlarged version.
Pictured above: The same area and zoom in 2010 and 2016. Notice how many fewer labels are on the 2016 map.
Comparing the screenshots above, the majority of the missing labels are city labels.
Just how many fewer cities are labeled today? Let's count:
(Above, I'm counting the number of city point icons on each map.)
2010 - 46 Cities
2016 - 8 Cities—an 83% reduction in city labels.
46 cities in 2010 vs. 8 cities in 2016—that's pretty significant.
Even more interesting are the kinds of cities that have been dropped from this zoom. For instance, the area's second largest city, Newark, is missing today—even though it was on the 2010 map.
In fact, if you take the area's five largest cities (after New York City), you'll find that none of them are on today's map—even though four were on the 2010 map (Newark, Yonkers, Paterson, and Bridgeport):
The map has become so sparse at this zoom that you have to go 35 miles away from New York to find the next labeled city:
Strange, isn't it? But maybe it's just a New York thing.
Let's check a couple more cities and see if the pattern holds. We'll start with Chicago:
Similar to what we saw in New York, the 2016 map of the Chicago area also has fewer labels.
Just how many fewer?
Let's count the number of cities between the two:
2010 - 44 Cities
2016 - 12 Cities—a 73% reduction.
And similar to what we saw in New York, the Chicago area's second largest city (Aurora) is also missing.
Let's check one more area and see if the pattern continues. This time, the San Francisco Bay Area:
Similar to New York and Chicago, there's once again fewer labels on the 2016 map.
Oakland and Berkeley's omission from the 2016 map is particularly surprising. There's ample room for them, if San Francisco's text is positioned to the left of its icon, as it is on the 2010 map. But that's yet another peculiarity with the 2016 maps: the text is always centered directly above or below a city's point icon. Contrast this to the 2010 maps, where a greater variety of label positions was used.
One last time, let's count the cities on each map:
2010 - 44 Cities
2016 - 10 Cities—a 77% reduction.
Three sets of samples, all showing the same thing: the number of cities has significantly decreased between 2010 and 2016.
While you were looking at the maps, did you notice some of the other changes?
FEWER CITIES... BUT MORE ROADS?
If you look closely at the maps, the cities aren't the only thing that've changed. While the number of cities has decreased, the number of roads has actually increased.
Take another look at the New York maps:
This is the first set of maps we viewed. Doesn't it appear as if there are more roads on the 2016 map than on the 2010 map?
Focusing on Connecticut, there are many more roads on the map today:
And while many roads were added, others that were already on the map appear to have been upgraded. Take these roads on Long Island, for instance:
Many of the Long Island roads were already on the map in 2010, but their appearance is different today: they've been upgraded in importance.
Across the map, in fact, a number of roads are now more prominent than they were in 2010—about 40, in all.
Below, I've highlighted the roads that were upgraded in black:
Interestingly, many of the upgraded roads are shorter segments—and they're generally not as important as the roads that were already prominent in 2010, such as Interstate 95 and Interstate 80.
So many roads have been added and so many others have been upgraded, that the 2016 map is cluttered compared to the 2010 map.
Take the area just north of New York City, near Yonkers, for example:
In 2010, there were plenty of roads in the area, but you could at least follow each one individually. In 2016, however, the area has become a mess. With so many roads so close, they all bleed together, and it's difficult to trace the path of any single road with your eyes.
And look once more at Long Island:
The primary route across Long Island—Interstate 495—is clearly shown as such on the 2010 map. But on the 2016 map, it's suddenly unclear: the newly upgraded roads muddle the map and 495 is lost amongst them. Worse, you can't tell which road the “Interstate 495” icon belongs to.
Regarding the Long Island road network, it's as though the reversal of Tufte's suggestion was implemented between 2010 and 2016. The roads that are dark orange today were all on the 2010 map—but their design has since been changed, causing the map to appear unnecessarily complex. The coherence and clarity shown in 2010 has been lost in 2016.
And consider that none of the upgraded roads are labeled:
If these roads were important enough to be upgraded in appearance, why weren't they also given labels or shield icons? After all, an unlabeled road is only half as useful as a labeled one.
Looking at the maps, there are more roads than there once were—and fewer cities.
I wonder what drove these changes?
One thing's for sure: today's maps look unbalanced. There's too many roads and not enough cities.
What can we do to fix it?
Let's dive deeper...
ORPHAN CITIES & ROADS TO NOWHERE
At the zoom-levels we've looked at, the map is primarily about cities and roads.
Just look at how barren the map looks without them:
Left: Google Maps, 2016
Right: Google Maps, 2016—with all of its cities and roads removed.
Quite a difference, isn't it?
Now let's see what the map looks like with just cities and roads:
Surprising. With everything removed, other than cities and roads, the map still looks like a map.
In many ways, the map above is a simple network map: The cities are the nodes. And the roads are the paths between the nodes.
If you live in a city, you're likely quite familiar with network maps. You might even use them pretty regularly.
Here's a good example of one:
And here's an even more famous one:
One thing you'll notice about these maps: they rarely have lines that don't have any stations. And they rarely show stations that aren't connected to lines.
After all, what would be the point of a line that didn't stop anywhere? Or a station that didn't get stopped at?
And yet this is exactly what we've seen on Google Maps, both in 2010 and in 2016.
Take another look at our Bay Area maps:
Notice the pattern?
2010 - Lots of cities, but very few roads.
2016 - Lots of roads, but very few cities.
...or put another way:
2010 - Lots of stations, but very few lines. (And many stations aren't connected to lines.)
2016 - Lots of lines, but very few stations. (And most lines don't have any stations.)
If this was a transit map, it wouldn't be very useful.
Is it useful as a road map?
Let's take a closer look at a couple of areas within the Bay Area.
First, the Pittsburg / Antioch area:
2010 - Cities, but No Roads. Pittsburg and Antioch are shown—but how to get there? No roads are shown that go to Pittsburg and Antioch.
2016 - Roads, but No Cities. Roads leading to Pittsburg and Antioch are shown—but Pittsburg and Antioch aren’t labeled. Why travel on those roads? Where do they go?
On the 2010 map, Pittsburg and Antioch are what cartographers call "Orphan Cities". That is, they're cities that lack connections to the rest of the road network.
A similar situation exists with Santa Cruz:
2010 - Santa Cruz, but No Roads. Santa Cruz is shown, but it's orphaned (i.e., there are no roads going to it).
2016 - Roads, but No Santa Cruz. Four different roads leading into Santa Cruz are shown—but Santa Cruz isn’t.
On either map, it's not immediately clear how to travel between San Francisco (or any other Bay Area city) and Santa Cruz.
See the problem?
Both maps, the one from 2010 and the one from 2016, have a similar issue: a lack of balance.
Google Maps of 2010 had plenty of cities—but not enough roads. It was unbalanced.
Google Maps of 2016 has a surplus of roads—but not enough cities. It's also out of balance.
So what is the ideal?
* * *
When I was living in Chicago, I once saw an incredible map in a stationery store. Someone had taken the map and turned it into a cover for a photo album, the kind of thing you'd typically find on Etsy and other online craft sites.
I snapped a few pictures of it because it's one of the best examples of this concept I've ever seen. (The photos were taken with an iPhone 4 on a dark, winter day—so please forgive the quality.)
Now compare the map in that photo to a comparable zoom on Google Maps today:
Even though it's from the early 1960s, the old map is more balanced than the Google map.
Notice that there are no orphan cities on the old map. Every city is connected to the road network:
And not only are there no orphan cities, but there are no unnecessary roads. Nearly every road is labeled—and nearly every road has a city along it. We don't have the cluttered nest of unlabeled roads that's on the Google map. Instead, every road serves a purpose.
The balance is spot on.
And the map is incredibly efficient.
I took that picture more than five years ago, but I've always wondered what a digital version of that old map would look like. Would it be better than Google Maps today?
Let's try a quick experiment.
First, we'll take Google Maps as it is today:
Now, let's add all of the cities from the 1960s road map:
The added cities certainly improve the map's balance.
Next, let's discard all of the roads that weren't on the 1960s map. We'll keep the freeways, since many of them weren't built yet—but let's get rid of everything else:
It doesn't feel as though we're missing much with all those unlabeled roads gone. And the individual roads are now so much easier to follow, especially above Chicago. You can actually trace each one with your eyes.
Put another way: every line now has a station. And every station has a line.
There's one problem, though: we still don't have labels on most of the remaining roads.
Let's rectify that by adding most of the shield icons from the 1960s map and see what we get:
Nearly every road is now labeled, and the map is more useful.
Let's take one last look at the map we started with and compare it to the map we ended up with:
If I were lost in this area, I know which map I would want to use.
* * *
All things considered, I suspect that Google Maps's city reduction was an optimization for reading the maps on mobile devices, and that the new roads were added to make the maps look less empty (once the cities were removed). After all, a map with fewer labels is a map that's faster to read.
Consider that during the period between April 2010 and April 2016 (when all the screenshots were taken), sales of mobile devices exploded. This chart from Benedict Evans captures the explosion in sales:
Also consider that during the same period, mobile usage of Google Maps surpassed desktop usage.
Given these trends, it's likely that Google Maps was optimized for mobile—and this explains some of the changes we observed earlier.
Unfortunately, these “optimizations” have only exacerbated the longstanding imbalances already in the maps. And as is often the case with cartography: less isn't more. Less is just less.
Google should restore its map's balance by adding back a few of the city labels it removed. As someone who's admired and used the product for many years, I hope that they do.