Exploring Local
Mike Dobson of TeleMapics on Local Search and All Things Geospatial

Business locations, map errors, address errors – Let’s stop the nonsense

January 15th, 2015 by admin

My colleague Dr. Stephen Guptill and I have been noodling about how to apply aspects of the formal theories of map accuracy and uncertainty in a manner that could be used to evaluate and report on the quality of business listing information commonly found on maps. The problems with business listings are not necessarily related to errors generated by the business listing systems side of the equation and may, instead, reflect various inadequacies of the map base or the mapping system used to fuse and present the data. In order to assess the mix of variables and interactions between them that we might encounter while attempting to formally study the issues involved, we have started examining the mapped presentations of businesses located in shopping areas known to us.

Dr. Steve is well-known for his work on issues of map accuracy and has published or presented numerous articles on the topic, some when he was one of the top scientists with the mapping division of the United States Geological Survey and other publications issued after he started his consultancy in GIS and Spatial Epidemiology.

Today’s blog is a co-authored effort by Steve and myself.

In order to get our heads around models, measures, dynamics, and usability related to the topic of map accuracy, we have been examining the mapping systems of online map providers in order to assess whether or not our ideas merit worth further examination. However, every time we look for examples and start examining local business listings, we fall into rabbit holes of extraordinary dimensions and complexity.

In a recent short discussion piece on Data Lakes, Michael Stonebreaker writes on why data lakes are actually data swamps. We recommend his article to you as it is a succinct description of the problems that result when companies misunderstand the complexity of data curation. While Stonebreaker is discussing a simple list, we believe that the curation of spatial data is an enormously complex situation. In the case of business listings, not only do we have quality issues related to the business listing data, but also quality issues related to the map to which the address is geocoded and displayed, as well as the data handling processes involved.

Almost every map provider collects some business listing information and then mixes it with licensed data from companies who profess to create accurate, well-researched business listings. Because some providers appear to be better than others in supplying specific categories of business listings, most map companies license business listing data from several companies, as well as adding in some crowdsourced listings, or perhaps listings that result from their own attempt at creating a business listing registry. The problem with these efforts is not in the provider’s ability to ingest the data collected or licensed, but in the ability to curate spatial data in a manner that results in reliable, accurate geographical information. Read the debacle below to better understand why Dr. Steve and I are nosing around a way to describe and measure the problems with the use and display of business listings.

The blog that follows outlines our search for a single bank and where it is actually located in Northern Virginia. The purpose of this particular blog entry is to provide some background on the complexities of the world of mapping business listings.

Detective Joe Friday of Dragnet fame was known for his monosyllabic summation of almost all investigations. “Just the facts, Ma’am.” He said. So here we go. It’s a long one, but there are lots of pictures.

There is a United Bank located in the area known as Fair Lakes Shopping Centers in Northern Virginia. There are three shopping areas at Fair Lakes, all managed by the same property management company. The United Bank is located in the shopping area known as Fair Lakes Center, shown by the blue map pin on the Google map that is provided on the website of the shopping center. (See Figure 1).

Location of the Fair Lakes Shopping Centers
Figure 1. Location of the Fair Lakes Shopping Centers

To start our search for the bank, we opened Google Maps and panned the map until we had the Fair Lakes Center centered on the map. No bank names or associated symbols were showing on the map so we entered “banks” in the map search bar. The relevant portion of the map that resulted is shown in Figure 2.

Search for
Figure 2. The bank of interest in this blog is the United Bank. It is incorrectly located on this Google Map.

Left clicking on the bank symbol labelled United Bank revealed the information shown in the following image (Figure 3)

Google Maps showing the address information it has for this bank
Figure 3. Google has the location of the United Bank as 13060 Fair Lakes Boulevard. Unfortunately, the location of the map pin is not that address.

Right clicking on the red dot at the bottom of the red map symbol labelled “United Bank” on the display above revealed the information shown on the map below (Figure 4). Note that the address information provided in the next map is not the same as the address information provided above as the location of the bank.

Google's alternative address information?
Figure 4. if you right click on the map pin the address is not 13060 Fair Lakes Blvd.

Right clicking on the location the blue dot displayed on the attached map revealed that Google Maps identified this street-stub as Fair Lakes Parkway and provided an address range for this entity (Figure 5). Note that the address information on the map below is not the same as the address information Google originally provided for the United Bank branch.

Another address for this location?
Figure 5. Right clicking the blue dot seems to indicate that Google thinks this street stub is an address range on a street named Fair Lakes Parkway.

The facts so far are these:
1. Google has located the United Bank at 13060 Fair Lakes Blvd.
2. The bank building is not visible on any of the Street View images on the maps that accompany any version of the address Google provides for the bank.
3. Google has stated that the address of the bank is 13060 Fair Lakes Blvd, but the geocoded location is adjacent to a street that is labeled Fair Lakes Pkwy.
4. The address 13060 is greater than the address range on the block named Fair Lakes Pkwy and the symbol for the bank is placed beyond the end of the street with the highest address, even though this is not the street referenced in the address Google provides for the bank.
5. Further, the location shown by Google for the geocoded location of the bank symbol places it on Federal Systems Park Drive in front of the parking lot for the Northrop Grumman Information Systems building.
6. The street stub the Google labels Fair Lakes Pkwy intersects a street, also, named Fair Lakes Parkway.
7. Route Planner by TomTom indicates that the street-stub labelled Fair Lakes Pkwy by Google is a named continuation of Fair Lakes Blvd and not named Fair Lakes Pkwy. (See Figure 6 below).
8. Neither company is correct, as the street-stub beginning at its intersection with Fair Lakes Pkwy is a private road named Federal Systems Park Drive. We drove by the location and photographed the sign identifying the street as Federal Systems Park Drive. (See Figure7 below.)
9. When queried to find the location at which TomTom geocoded 13060 Fair Lakes Blvd, TomTom located the address several blocks from the Google location. (See figure 8 below.)
10. The bank is not located where TomTom positions it.

TomTom's street address position
Figure 6. TomTom incorrectly labels the street-stub Fair Lakes Blvd.

Actual street sign photograph at the intersection of interest.
Figure 7. Oops. The street is actually named Federal Systems Park Drive, not Fair Lakes Pkwy, or Fair Lakes Blvd.

TonTom's geocode of the bank address
Figure 8. This location is where TomTom thinks the bank is located.

Well, maybe the bank knows its address and where it is located. Let’s look. We “Googled” the corporate website of United Bank to reveal the following information

Bank address according to United Bank website
Figure 9. This is the address of the bank provided by the United Bank corporate website.

Google and the bank agree on the address of the bank, but do they agree where the bank is on the map? The red map pin shown below is where the United Bank’s version of Google Maps shows its location (Figure 10).

Map used by bank to show its location to customers
Figure 10. The bank’s map agrees with TomTom, not Google, even though it uses Google Maps as a base.

However, as noted above, the bank is not located at this position. It appears that the bank’s version of its mapped location is based on lat/lon coordinate pair passed to Google Maps. If you right-click the location of the bank on this map and select “What’s Here?” the query will return a coordinate pair, a Street View image, but no address, nor does Google appear to return addresses for any location along Fair Lakes Blvd. As you might imagine, the Street View image does not include the United Bank (Figure 11).

No address range available using Google's What's Here button
Figure 11. Clicking “What’s Here” on the Google Map reveals a coordinate pair but no address range information. We could not find address range information for any segment of Fair Lakes Pkwy.

As a next step, we called the branch office of the United Bank in question to find out the location from the proverbial “horse’s mouth.” We asked for the address saying that we wanted to enter it into our GPS. The person we spoke to said very slowly that the address for the bank was 13060 Fair Lakes Blvd, and then carefully outlined how we would need to enter the Fair Lakes Center and navigate to the actual location of the bank, since it was not directly accessible from Fair Lakes Blvd.

As you might suspect, we knew the location of the bank from the beginning. In fact, the easiest way to find the United Bank in question is to enter the following address into Google maps: 13060 Fair Lakes Shopping Center (Figure 12). Of course, in order to do that you would have to know this address in advance. In this case, it would be pretty hard to find this address since neither Google nor the Bank provide this clue.

Oh gosh, look at this, a symbol exactly where the bank is actually located (Figure 12)

Houston, we have found the missing bank
Figure 12. Finally, the United Bank in all its glory!

And look, the Street View image shows the bank, and if you get fiddly with the controls you can position the image and see the address on the door facing the street named…Fair Lakes Shopping Center. Hmmm. (Figure 13)

Wow, you can even see the bank, and its address, in Street View
Figure 13. 13060 shows right on the door. But how about those photos?

Curiously, the Best Buy located next to the bank, provides its address as 13058 Fair Lakes Shopping Center, and not 13058 Fair Lakes Blvd. Perhaps we have discovered some confusion about mailing addresses and location addresses, but you should know that the Shop Fair Lakes website, which leases shops at this mall, lists the address of the Union Bank as “13060 Fair Lakes Center” (figure 14)

Where the Shopping Center Management appears to think the bank is located
Figure 14. It appears the mall management company thinks the bank’s address is on Fair Lakes Center.

In searching further, we found several websites that had the bank’s address at 13060 Fair Lakes Blvd, an equal number that had the address as 13060 Fair Lakes Shopping Center. Several of these sites had the United Bank erroneously located on a Google map in or near the position shown by the bank’s own map of its location, which was erroneous. Of the others that gave its address as 13060 Fair Lakes Shopping Center, two had it located in the shopping center, but not in the correct locations. We also found on Google+ that the address 13060 Fair Lakes Parkway, which does not lead you to the bank, was a verified address for the United Bank, as is shown by the circled shield in the following image (Figure 15). Verified maybe, but Google Maps does not know where to locate this address. Even if it did know the secret code for accurate mapping, it would still not have the bank located correctly!

Wow the address of the bank is a verified location in Google+
Figure 15. Ahh, what does the mall management know about the location of properties it leases anyway?

Of course if you enter Google Search (not Map Search) using 13060 Fair Lakes Shopping Center Fairfax VA, Google will pop a map that shows the exact location of the bank and you can even see it on Street View (Figure 16). Hmmm.

Look what you can find in Google Search, but not map search
Figure 16. Look, if you use Google Search with the location address, you can find the bank.

However, if you combine the bank name with that address and search for “United Bank 13060 Fair Lakes Shopping Center Fairfax VA,” you will see a page of listings, but no map, as this is not the entity relationship in Google’s database (See Figure 17).

Using the location address in normal search does not directly reveal the location of the bank.
Figure 17. Hmm. If you use Google Search with the location address, you still cannot find the bank.

Similarly, if you search for “United Bank Fair Lakes Shopping Center Fairfax VA” you get some search engine results pages and an inset map from Google Maps of the incorrect location. (Figure 18).

More disappointing search for the bank
Figure 18. When you tie the location address to the entity known by Google, it must return an incorrect result.

It appears that even if you knew the correct “finding/location” address for the bank that you could not use Google to find it (Figure 19).

Using the correct finding address for the bank results in the incorrect location when using Google Maps
Figure 19. How frustrating!

Four final notes:

1. If you look at OpenStreetMap for this area in Northern Virginia the United Bank is clearly and correctly labelled.
2. Google (and likely all other mappers) appear to have no idea of the boundaries of shopping centers, although these can easily be found using Google Search. To some degree, the inability to map a business listing within a shopping area reflects the incompleteness of the spatial models used by the companies participating in the local search market. Hmmm!
3. I have seen articles by easily impressed reporters indicating that Google uses machine vision to read the signs in Street View. Maybe Google should spend some time using these capabilities to locate businesses.
4. Google – are you familiar with the concept of “fuzziness?”

Conclusions

The search for the United Bank detailed above is an example of the insanity in the world of mapping business listings. Companies that provide business listings frequently collect, scrape and/or merge incorrect locational information, improperly geocode it, and, in the process, make it impossible for consumers to find the businesses or services that they desire to locate. Why bother to put this stuff on maps if it is erroneous? If a human can’t use your system to find the business, do you think that Android Auto or CarPlay will do any better? Does this time consuming, erroneous process make consumers happy? Do advertising clients appreciate it when a customer cannot find its business due to a listings or map error?

Maybe it’s time to evaluate the scope of this problem once and for all so that companies wishing to publish the locations of business listing data can find better ways to evaluate the accuracy of the data they are offering. While it may be an attractive notion to think that you can provide this type of data without interactive processes and backtracking (see the Stonebreaker article mentioned earlier), those with a serious background in spatial data handling will understand that you do so at a considerable risk. Finally, it is unlikely that anyone can prepare authoritative and accurately mapped business listing data without some human involvement. The United Bank example may be one of those cases. Of course, the degree of the problem with business listings accuracy remains unknown and that is exactly why we are interested in examining it. More as we progress.

Bookmark and Share

Posted in Authority and mapping, Bing maps, business listings, Categorization, Data Sources, Geotargeting, Google, Google maps, Local Search, mapping business listings, TomTom

4 Responses

  1. Fred

    Interesting read and indeed show how complex and difficult this industry is if you want to do this (globally) right.

    But curious why you didn`t include here.com as they fuel 99% of the car navigation with their maps and seem to have the largest work force of geo-specialist?

    Do you find them too focused on the car industry, less on the business listing?

    “Hi, Fred:

    Thanks for your comment.

    As we noted in the blog’s introduction, we have been looking at maps from various suppliers to inventory the types of errors that we might find in business listings. The example reported was used as we thought it particularly illustrative of some of the problems involved in the business listings universe. We purposefully did not intend to create a comparison between various online providers of maps. We did refer to TomTom when we were looking for the street-stub name, but then decided to simply go into the field and photograph the street sign to establish the real name of the street rather than canvass other providers, since without going into the field we would not know the truth of the matter.

    Make no mistake, we hold HERE in high regard and will be looking through their products if we decide to launch our examination of business listings.

    You mention that HERE may employ more “geo-specialists” than others in the industry. We think that claim may be related to how you define the term. HERE, Google, Apple and TomTom certainly have large teams focused on geography, but, at this point, we do not have enough reliable information to assess and categorize work force sizes.

    Thanks again for your comment,

    Dr. Mike and Dr. Steve”

  2. Thom K

    “The United Bank example may be one of those cases”….perhaps we’ll never know as this research used one example out of millions that could potentially be correct. Small sample size studies reveal biased opinions seemingly fuel the authors agenda.

    “Hi, Thom:

    Thanks for your comment.

    If we did not make it clear enough in the introduction to the blog to which you refer, let us reassure you that this was only an example of the “rabbit holes” that one finds when attempting to make sense of business listings. However, we thought it was an example that was worthy of further commentary. As we detailed in the article, we are considering a substantial endeavor in this area, but are taking our time to make sure that the measures we are considering would be useful in attempting to provide a framework for analyzing the wide-ranging problems with business listings.

    Thanks again, and keep up the blogging at disturbed geographer.com .

    Dr. Mike and Dr. Steve”

  3. Philip

    Mike, you use Google and TomTom, and what about HERE (NAVTEQ)?
    Full disclosure: I work for HERE.
    I’m sure you know HERE has geo-analysts driving the roads and human based QA processes on top of automated ones. Getting POI data right all the time is a challenge. Especially with fast turn-over POI like restaurants or locations that use vanity addresses or descriptive addresses that are challenging to geocode.

    I totally agree with your conclusion and can’t wait to read your further findings.

    Following a link to the United Bank on HERE.com: https://www.here.com/usa/_/atm-bank-exchange/united-bank–840dqbvn-3afa98a05d374ae594dce08e6cef5808?msg=United%20Bank&map=38.85616,-77.39612,17,normal

    “Hi, Philip:

    Thanks for your comment.

    Yes, we have been following HERE and its corporate predecessors since the mid-1980s. As noted in the response to another comment posted below, we hold HERE in high regards and the single example we presented was meant only to be an example of the complexity of the issues surrounding creating accurate business listings. We hope to examine the offerings of all companies in this space, if we decide to move forward with the effort noted in the blog.

    We wish you and Team at HERE the best of success.

    Thanks again,

    Dr. Mike and Dr. Steve”

  4. Jim L.

    Drs. Mike & Steve,

    Enjoy reading your blog. Seems like the online maps are getting worse instead of better! Toss in the number of businesses who decide their cell phone is enough and land line is no longer needed. They release the number before making any corrections to their datasets and then wonder why the customers, reading the phone number from a refrigerator magnet, are confused.
    On the recent side, Bing’s decision to add non-validated Yelp listing to their maps has really added to the confusion. Cartographers of old are rolling in their graves to see these modern maps.

    Keep up the good reads.
    Jim L.; (Smart DataMapping in Maine)

    Hi, Jim:

    Thanks for your comment. We appreciated hearing from you.

    Dr. Mike & Dr. Steve