It’s time for that long winter’s nap – or perhaps a long read on a topic of interest to everyone – business listings and Google Maps. Thrill your boss, intrigue your friends and drown the conversation with data – all of this can be found below and it’s free, but only for the next millennium. Why, you might even gift your boss with a copy. I am sure they will thank you. Sorry about the size of the graphics and their extending out of the text frame, but I wanted you to be able to see them.
I have been following Google’s interest in fielding autonomous cars, their use of AI and maps to improve their navigation, and how Google thinks it will be pretty easy to expand their current map and related databases to support these efforts. (See this effort for an example of the coverage). In addition, I recently perused what was a very complimentary review in Wired of Google’s Ground Truth.
Hmmm. Perhaps I have slept through some major revolutions in mapping and map data quality. Well, actually I was in Italy for the month of October and maybe everything new happened then? Actually, I don’t think so because I had reason to use Google and Google Maps to find services I needed in the cities I was visiting. I was amazed. The business listings shown on Google Maps for several cities in Italy appeared to be even more incomplete than the Google Map to which I referred at home. Even more worrying, I was in established urban cores that should be well known to any serious providers of maps. Hmmm.
While in Italy, I had occasion to search for laundries, dry cleaners, stores, and business services, you know, the kinds of things related being away from home for an extended period. Since I was not familiar with the geography of businesses in the cities and regions I visited, I had intended to consult Google Maps to find locations so that I might be able to acquire those services and goods without undue exploration of the city. However, after using Google Maps in Italy this notion appeared to be a cruel hoax.
I had difficulty accepting that the types of services for which I was looking were not available, say in Rome or Florence. As I walked through the cities, I noticed a number of businesses that provided what I was looking for, but they were not shown on Google Maps. In addition, many of these stores had not appeared when I used Google Search to interrogate a category that I thought should contain a specific type of business. Often I could often perform a Google Search on the name of the business that I saw while passing by and found that it would now be displayed on the map as a search result. This process seemed sort of backwards to me, since I was using Google Maps to try and find potential targets in the real world, not using reality to find locations that could be symbolized on Google Maps if I searched for the name I saw on a storefront.
On arriving back in the US I decided to take a closer look at business listings on Google Maps. I was hesitant to once again look into business listings, as it is the scab of the mapping world that never seems to heal. However, as the venerable Yogi Berra is reputed to have said, “You can see a lot just by looking.” So, I headed for two shopping areas near my home after having equipped myself with a camera, GPS, note pad and a trusty sidekick who was willing to work in return for lunch.
About every two of years, for one client or another, I take a close look at the state of business listings on maps or navigation systems, shudder and tell them things were not any better than two years before. Once, I even designed a business listings system for a major player in the electronics space, but when they saw the complexity and expense of becoming a market leader they decided not to pursue their intended strategy. On the other hand, maybe it would be different this time and it sounded like a fun way to spend a few days.
Note: the study reported here was just for my information and does not have the detail of a client report. I think you might find it interesting – at the very least it will save you from having to go out and look for yourself – and if you are interested in business listings it will give you a lot to think about.
Figure 1. The two shopping areas included in the present study are located in Laguna Niguel, California. Basemap courtesy of Google Maps. The areas were chosen based on accessibility to the author and because I had estimated that they would contain close to 100 businesses. Not a large sample, but ample to provide some food for thought on the topic.
Both shopping areas are outdoors and open, rather than an enclosed mall setting. Each consists of a large main building offset from the road, several pads or minor buildings scattered through the shopping area, and acres of parking. In both sites a few shops are located in buildings adjacent to nearby streets. In most cases the businesses in these centers are set-back far enough from the road to render reading business names impossible while driving by. Exceptions to this general rule are the huge signs for the anchor stores in each shopping area (e.g. Home Depot, Hobby Lobby or Wal-Mart).
I started the research by creating an inventory of the businesses operating in each of the shopping centers. I enumerated all businesses in the respective centers noting their individual position, name and general category of business. I entered the names of the businesses in list form with address information and also noted their location on aerial imagery. I, also, had maps of the businesses in each shopping area that I retrieved from the management web sites for the properties and used these to determine details of the geometry of the shops for further comparisons with building outlines and partitions that might be included on Google Maps. Although the maps provided by the management web sites were relatively up-to-date in terms of tenancy, my list of businesses was based on field observations.
In order to determine which businesses were shown on Google Maps I used a desktop computer (several actually) to interrogate Google Maps. The examination involved zooming, scrolling, examining images, examining Street View (when and where it existed), using various “developer’s consoles” associated with several browsers to determine details about the map tiles and Google servers involved, as well as using the Google Index to search for businesses that did not appear on Google Maps. When a Index search resulted in success, I then clicked the inset map that accompanied the search results and examined the location in Google Maps
Note that the use of Google Search and Google Maps are influenced by prior searches, prior Google Maps use and other information that Google knows about you as a user. The research results described here may be specific to my computers (several different machines were used during the process) and not reflect the experiences of others performing similar searches. However, without being more open about why, I have concluded that the results of my map viewing and searching should not be significantly dissimilar to those of most other users of Google Maps who view the Google products using an Internet browser and a desktop computer. Although I checked some of the business listings using mobile devices, this was to ensure that there was a basic compatibility between the business listings on Google Maps using various devices.
Note that after I had finished my original analyses at the end of last week, Google updated its map database and made changes to the businesses in both centers. I incorporated these changes in a second analysis that I will provide as part of this report. Finally, after having received a note from a friend on events at Apple Maps, I decided to include the business listings they had for the areas of interest, and will, also, comment briefly on their efforts in the area of business listings at the end of this blog.
Research Questions and Considerations.
It is my belief that Google or any company providing detailed street maps involving a wide range of scales for presentation should, at the appropriate scale or scales, attempt to name all business that occur with the boundaries of the areas they have mapped. The detail limitations on map content related to the lack of space available on printed maps are irrelevant in digital mapping systems designed to show extreme details in local areas. Users of modern mapping systems should be able to use large-scale maps as surrogates for reality. In the case of business listings this would equate to naming every business in a local shopping area, so that the user could find and navigate to any of the potential targets available and of interest to them.
It is clearly the case that not all businesses in an area are represented on Google Maps. One could postulate several possible reasons for the lack of a complete list of local businesses, although not all are equally likely:
1) Data quality issues – Google might not possess the required data to describe and/or locate businesses accurately within a specific area
2) Geocoding issues – Google might be unable to adequately determine the validity of the address data that must be geocoded to locate businesses.
3) Street View coverage limitations – Some entities, particularly, shopping areas with internal streets and roads may not allow Google vehicles access or, perhaps, Google has instructed drivers to not to capture street view in particular commercial locations (e.g. Google Street View imagery was unavailable for the majority of the internal roads in the shopping areas involved in this study). Alternatively, Google may have the necessary Street View data, but have not yet processed or published them
4) Google wants their users to search for data on unidentified businesses as they make considerable profits from such search activities versus the limited income directly generated by their mapping operations.
5) Google’s map design appear to me to be heavily influenced by their desire to capture more revenues from mobile devices, whose screen real estate could limit the usefulness of burdening the display with the names and positions of all business locations. Alternatively, there could be aesthetic design reason for limiting the number of displayed business tags on Google’s map
I thought that looking more closely at what Google was doing in my local area might be helpful in understanding Google’s display strategy in regards to business listings.I had six general questions that I wanted to answer during the course of my research:
1. How many businesses were in the shopping area and how many of these businesses were shown on Google Maps?
2. What was the accuracy of the locations of business shown on Google Maps?
3. When businesses were located in the correct building, were they correctly located within that building?
4. When businesses were not located on Google Maps, could Google Search map them and what was the accuracy of the locations on Google Maps revealed through Google Search?
5. When “searched” business locations were shown in the correct building, were they located correctly within that building?
6. What could Google’s performance in displaying business names on maps tell me about their strategies and challenges?
Let’s look at the details revealed by the research. There is a lot of data here, so I provide a series of graphics and brief captions, rather than spend a great deal of time in detailed explanation. The graphs generally follow the order of the questions provided above.
Plaza de la Paz
Figure 2. Plaza de la Paz is comprised of 52 businesses. Google has mapped a portion of these and included 2 additional businesses that once operated here but are now out-of-business. Approximately half of the businesses at Plaza de la Paz were shown on the Google Map of this area.
Figure 3. While most of the mapped locations were relatively correct (near where they should be located on the map or easily findable from that position if at the shopping area) those that were misplaced or significantly misplaced appeared to me to be errors resulting from some aspect of the geocoding process. Frequently, these locations were symbolized along external roads bounding the shopping areas, but generally not symbolized within the buildings that actually house these businesses and which Google shows on its map.
Figure 4. As noted previously most of the businesses that were mapped were located in the correct building, although some of these were generously misplaced within the building that they occupied. In some instances this was not a major problem, but in other cases, it put the business on a side of the building that was not visible if you were looking for the business near the location where Google had located it on its map. I moved a business into the misplaced category when it was unlikely that most people could figure out where the business was based on the symbol location. Note – the main building in this shopping area is several blocks long, so the measure of “occurring within the correct building” is a very liberal measurement.
Figure 5. One-hundred percent of the businesses not shown on Google Maps had locations within the shopping area that were known to Google. In order to find these businesses I entered each business name (known to me only because I had observed the name in the field) in Google Search (not Map Search) and then clicked on the map that appears to the right of the listing on the search engine response page. In most cases, these businesses were located in the correct building, although several were shown along streets or in parking lots, rather than within the building outlines that Google had extracted from satellite imagery.
Figure 6. Searching the Google Index revealed eighteen businesses that were mapped located in the correct building. A modest number were misplaced within the building and shoppers would likely experience difficulty finding these businesses using the locations provided on Google Maps.
Next, let’s see the results to the same questions for the second study area.
Figure 7. Google had a much better grasp of the businesses that were located at the Marketplace as compared to Plaza de la Paz. In part, this may be that the layout of the shopping center is “open” and the buildings surround central parking rather than being embedded within buildings surrounded by parking areas, as at the Plaza de la Paz. Google did not map any out-of-business locations at Marketplace.
Figure 8. Not only did Google list more of the businesses in the Marketplace, but it was able to locate them more accurately than in Plaza de la Paz.
Figure 9. Unfortunately, while able to identify the correct building containing the business, Google was less able to accurately locate the businesses within the building. My conclusion was that these were geocoding errors resulting from attempting to geocode the address data of the businesses without detailed reference information for the units involved. For example, these businesses have addresses such as 27150 A, 27150C, 27150F. There are no units labelled B, D, or E. Without knowledge of the omitted unit identifiers, it would be difficult to properly geocode the actual addresses for units that do exist or to even know where to start geocoding within a building.
Figure 10. Only one of the existing Marketplace businesses omitted from Google Maps could not be found and mapped using Google Search. Most of these results were correctly located.
Figure 11. In general, location accuracy of the results from Google Search was commendable. Although several locations were misplaced within the correct building, it should be noted that the sample size was small.
Thoughts so far
Google Maps clearly did not represent all businesses in the study areas. Approximately one-third of the businesses in the areas studied were not included on Google Maps. Moreover, when Google Maps was able to locate the business in the correct building structure, it frequently mislocated some of these businesses within those confines. In addition, approximately one-quarter of the businesses shown on Google Maps were either shown in the wrong building, symbolized in the parking lot or along an adjacent road, or were, in a small number of cases, no longer in business at the shopping areas examined. The businesses not on the map could be searched and mapped if you knew the name of the business from some other source. In general, the accuracy of the locations of businesses found through search was slightly inferior to that of the businesses appearing on Google Maps but not requiring search to symbolize them. However, when these “searched” businesses were located in the correct building, they were as likely to be accurately located within that building, as those businesses initially shown on Google Maps in the correct building. The last finding was unexpected. If the locations that required search can be shown on the map with approximately the same accuracy as those that are normally shown, why not show all of these data on the map to begin with?
In order to get a better understanding of Google’s strategies in naming businesses, I needed to see a little more data.
I decided to investigate if there were categories of businesses that appeared to be favored by Google and other categories that seem disadvantaged by Google’s approach. As is well known, attempting to categorize almost anything results in instant argument. And so it goes. I decided on the categories to use and feel that they adequately represent the variety of businesses that were discovered in the field.
First, I categorized the businesses in each of the two study areas and, then, lumped them into one pool of businesses representing both locations. The results were as follows.
Figure 12. It appears that there is a good mix of business types in the two areas studied. Dining dominates, followed by Specialty Retailing (mainly shopping for general stuff), Services (banking, FedEx, etc.), Salon Services (beauty shops, nails, etc.,) and Home Furnishing, Decorating and Supplies.
Figure 13 Wow; guess Google feels that people in “South County” like to dine out since Google Maps really emphasize the Food and Restaurants category. It appears that the businesses names published by Google on its maps reflect the general distribution/occurrence of these categories within the two shopping areas. The segments emphasized by Google are those commonly thought to be of most interest to mobile users.
Figure 14. This is probably the most interesting of the graphs as it shows the performance of Google Maps in terms of businesses mapped in each of the business categories. Google provides good coverage of the most popular gategory Food and Restaurants. In the two shopping areas studied it does a relatively poor job of representing Pet Services and Supplies and Pharmacy & Medical Services.
Indeed, in several categories it seems to be almost a fifty-fifty chance that an individual business will appear. Oh wait, it’s generally worse than that since, in most categories, national or regional chains/brands are well represented on Google Maps, at the expense of …small businesses. However, my experience is that this problem really exists at the level of business listing aggregators who supply their lists to Google and other providers. It is quite easy for these companies (or Google) to contact major retailers and solicit the location of their stores and outlets (or scrape them from the web). Finding the small business owners is much harder and that appears to be reflected by the lack of representation of small businesses on Google Maps.
Well, now that I had a better grasp of how Google Maps was performing, I had one more issue to deal with before concluding –
What about those Google Maps updates?
Before I could post this blog last weekend, Google decided to push out new map updates early Saturday morning. So, realizing that this might be an interesting opportunity, I processed the new data and made some new graphs. It appears that Google Maps really is trying to improve its business listings, but still under represents business listings. The update made only one minor adjustment to the Marketplace shopping area and this was to indicated a Redbox existed in the parking lot (not where it is located). Since the businesses mapped at the Marketplace were unchanged I did not update the graphs for the Marketplace. Plaza de la Paz, however, received a significant upgrade to the number of businesses listed.
Plaza de la Paz (after Google Update)
Figure 15. Mapped locations in Plaza de le Paz increased by seventeen percent in last weekend’s Google Maps update.
Figure 16. The number of mapped locations at Plaza de la Paz increased from 28 to 39 in the most recent Google Maps update, but some of the additions were misplaced and the number of erroneous businesses entries increased from two to three.
Figure 17. While the number of businesses mapped in the correct building increased by ten, the number of businesses locationally misplaced in the correct buildings increased as well.
Next, I wondered about how these updates were applied across the business segments?”
Figure 18. The recent Google Map updates was focused on the Food and Restaurants segment. With this update Google has identified all of the businesses in this category that exist in the two shopping centers! Other changes were the addition businesses to the Service category, as well as single business additions to Salons, Home Furnishing and Pharmacy.
As can be seen on the graph, a number of segments are still under-served and in most of these the businesses that are not represented are small businesses. One interesting exception is that Google continues to snub a sizable AT&T shop in Plaza de la Paz, although you can see the unnamed individual aisles in the Home Depot store instead, if that would be of interest to you. All in all, Google is making good progress in putting businesses on the map, but one has to wonder how soon they will accomplish the goal of complete representation and accurate location of local businesses.
Some thoughts about the location accuracy of listings.
The accuracy of business location on Google Maps is generally good for major chains (which often are in a stand-alone unit that makes it easier to identify the business), but less accurate for small businesses that are, often, clustered along the length, or sometimes the perimeter of a buildings.
In most cases in this study Google appears to have relied on geocoding to locate the addresses of their business listings (since they appear not to have Street View Data for these areas) and in many cases the quality of the address data, the map match, or the geocoding algorithm produced, or combined to produce, insufficient accuracy in the name placement results. In the previous sentence, the “map match” refers to the detail available to place addresses within a building in which all of the units may have similar addresses. For example, in both shopping areas there was one large, elongated main building that ran almost the length of the entire shopping area. Unless you have access to the building plans that detail the size and location of the units that house individual businesses, you will have difficulty matching the address to a reasonable location within the larger unit.
In both shopping areas, the quality of location data, as shown by business placement on Google Maps was quite unimpressive and less advanced than I had hoped. While Google was able to identify the building outlines within both shopping areas, it was often unable to accurately determine where the business was within the unit. If Google actually desires to field autonomous cars that can deliver customers to a commercial destination, it is my belief that the accuracy of the business location placement on Google Maps will need to be vastly improved.
Street View is probably a portion of the required elixir, but being up-to-date would require the vehicles to explore all shopping centers on a recurring update cycle. To be of benefit, the update cycle should be based on a “change” index calibrated to economic indicators and other local and regional trends that effect the provision of services and retailing.
While Street View would provide some of the required geometry, I find myself wondering why Google does not appear to gather the details concerning store placement that are available from most property management companies. People who make their money leasing property know the units, their details, and even their clients. Of course, they may be pursuing these data and may not yet have compiled it for the areas studied here.
Before I end this already overly long blog, I feel compelled to comment on the utterly unhelpful “what’s here” tab that shows up when you right click on or near a location on Google Maps. In my experience the result is an address and a Street View image, if one is available. Oh, that’s really helpful. Once in a while you will see an address and a business name, but that happens infrequently. Given the details about locations that Google has amassed, why is this button’s functionality so lame?
Over a decade ago I gave a presentation about the utility of the GIS-powered “What’s Here” functionality. When you click a spot on the map the system responds with map details showing what is around the point you have selected and links allowing further exploration. This is a trivial spatial search that would add a world of utility to maps such as Google’s. Why hasn’t Google or anyone else invested the time and effort to generate a space filling functionality that would make those empty maps incredibly useful to anyone searching for anything close by the point of interest?
Now back to the topic
One final note, a colleague mentioned some news about Apple Maps and the charge it is making in displaying business listings on its maps. Well, take a look at the next figure.
Figure 19. Apple critically lags Google in providing business names on its maps. Worse yet, Apple does not appear to provide building outlines, nor have access to any specific referential data that might improve its geocoding. As a result, it relies on street-face geocoding to place most of the businesses it names. The result is that the majority of the businesses are named along the surrounding city streets, even though this is not where the businesses are located. In the shopping areas examined in this study, street face geocoding and mislocating the business is better than not locating the business at all, but not very valuable to the consumer. Unless Apple Maps changes its current strategy for naming and locating businesses, it will remain severely behind Google in the Names Race. One ray of hope, Apple showed some of the small businesses that Google is missing. How about that?
Near the start of this blog I asked what Google’s performance in displaying business names on maps tell me about their strategies and challenges? My conclusions are quite simple:
1. Google is clearly working on improving the quality and abundance of the business listings shown on the basic Google Maps.
2. Google Maps continues to suffer from incomplete coverage of business listings and complicates this with less than accurate locations used to show the businesses that do appear on their maps.
3. Google Maps clearly under-serves the small business community. Although it has worked on strategies for improving this coverage it does not appear to have a significant advantage in this segment.
4. While the businesses in an area that are not shown on their maps appear to be known to Google, for some reason the company appears not to have enough confidence in these listings to present them as included content on their basic maps.
5. It remains unclear to me whether the incomplete listings reflect a quality control issue or is an indication that Google does not feel a need to provide complete coverage of businesses on its maps.
6. The completeness of business listings on Google Maps is a critical issue with implications for the utility of Google maps, as well as their ability to support navigation and autonomous vehicles in the future.
7. Google is obviously spending cycles trying to improve the quality and comprehensiveness of their business listing data and likely spending a ton of money to do so. What are its competitors going to do to compete if Google creates a highly accurate business listings database to go with its highly accurate maps? Either pay the toll, be shut out of the market, or continue to wonder why inferior products don’t excel in the marketplace.
There were several additional interesting items in the data collected for this study, but this blog is already too long and I have some Christmas shopping waiting for me. So with that, I send my warmest wishes for a Happy Holiday Season and may 2015 be your best year yet!