Location Targeting in Local Search
Maps are a useful contributor to improving the functionality of local search. Well, at least they should be! Unfortunately, problems related to the accuracy of the data used to create the maps often make their utility less than beneficial to users of local search.
Last time we spoke about geocoding and address errors and I wanted to take geocoding problems just a little further down the road (no pun intended). Most people use local search to find a location that meets their need for some sort of commercial transaction. Whether the searchers are looking for products, services, vendors, leisure activities or anything else, the end result is that the customer is likely to spend money if the local search functionality can correctly target the appropriate location.
Where many systems seem to fail is in mapping a viable location for shops in strip malls and shopping centers. Curiously, the performance accuracy of many mapping solutions is non-linear and locational accuracy varies across space both within and between the results of the various providers of local search.
It is easy to understand why the spatial accuracy varies between providers: some simply use better data from suppliers who can provide accurate positioning information. In other cases, the variance within the results of a specific provider may result from method induced errors or the fact that it is difficult to uniformly update large, spatial data bases. But enough of this spatial-temporal philosophy stuff, what does this all really mean for users of local search?
Let’s look at an example
SuperPages has a great interface, nice human factors and a clean design. Their maps from Microsoft are good looking and they have oblique aerial imagery from Pictometry to help people travel the last mile. So, let’s see how they do in providing search results. I have used a local neighborhood near my home as the testbed.
The search below was for coffee in a specific zip code. The results were disappointing. The two symbols placed on the map to show the location of the two coffee shops are incorrectly located and incorrectly shown as located in a condominium complex. In fact, the locations provided for the coffee shops cannot be accessed without driving east (off the image) and wandering through a spiderweb of streets only to find out there is no coffee available at these locations. I have noted the correct locations on the image.
Let’s look at another common problem – It’s in the shopping center but where?
As you can see, the Starbucks shop is not located where the provider positioned a symbol indicating its location. Unfortunately, the surrounding area is a “planned” community, so there is no signage visible from the street to save you from wandering the mall until you find the Starbucks, way up there in the corner. The problem here is that the vendor supplying the map data was unable the provide the specific location of the store, only an address for the shopping center in general.
Can the problem be solved? Sure. As we mentioned last time, you need parcel data and a system to integrate parcel information with aerial photos and addressing information to more accurately position addresses. Below is an example using Yahoo’s new mapping package that shows the two Starbucks discussed above located exactly where they occur in reality.
Unfortunately, the data that Yahoo is not uniformly correct, as the location for another Starbucks in the same area is incorrectly positioned.
Finally, two more examples. Yahoo gets a rural location in the middle of nowhere absolutely correct. Pretty good. Unfortunately, it misses my house in an urban environment.
The lack of accurate address positioning is a major shortcoming in the functionality of many vendors local search offerings. Data accuracy varies by vendor and it appears that too many local search companies may be evaluating the cost of mapping rather than its accuracy. From my perspective, if local search cannot connect the user to the spatial location of the opportunity, it fails the litmus test that useful local search systems must pass.
As I mentioned last time, there are solutions to these problems. Yesterday, I learned of a new one from DMTI Spatial. Its product MDU (Multiple Dwelling Unit) now includes verified unit floor counts, actual suite address numbers, building type, number of floors, etc. that the company feels are necessary to provided precision decision making. Unfortunately, the dataset is only for Canada.
Overkill? Not really. What we haven’t discussed here is the results of local search when the targeted stores are located in multi-level malls. Most local search users would appreciate knowing which mall entrance they should park near in order to visit the store they have targeted. Not an unreasonable request, but one that most vendors can’t meet today.
Vendors should be spending more time quality-assuring the cartographic data and spatial software that they plan to use to show the results of local searches. It’s clear that users cannot be happy about the quality of results and likely that advertisers are equally dissatisfied.