In the first part of my series on local searching, I looked at the poor state of results that appear when someone does a local search on general purpose search engines and Overture's new paid listings system that aims to improve the situation for both users and advertisers.
In this part, I'll focus on the crawler-based methods being used by Google and Mobilemaps to improve local searching when tapping into a web-wide database of content.
Local Search At Google
Called Google Search By Location, the service allows you to query across the entire web and retrieve results that match the subject you're interested in and which are also deemed relevant to a particular geographic area that you specify. Results are shown both in traditional "list style" and mapped visually.
In other words, let's say you want pizza in san francisco. Enter those terms into Google Search By Location, and you'll get ten matching restaurants, all mapped across the city.
Visually, the service is compelling. It's easy to see exactly where each restaurant is located. Zooming out with the "Search Larger Area" option below the map enlarges your search to retrieve matches from surrounding areas. However, though the match count goes up, I found the top listings themselves didn't change, nor did the map.
Legibly, there's more to be desired, especially when compared with a service like CitySearch, which we'll look at more closely next week. Listings don't use the restaurant names as their titles, for example. That forces you to scroll down and read each description to see if the restaurant's name may be present there.
This is a consequence of Google using unstructured data. It doesn't have a list of restaurants, each neatly identified by name and location. Instead, it has a collection of pages it knows should be relevant to San Francisco, because the pages have San Franciscan addresses on them. Google then looks to see which of these San Franciscan pages say "pizza" on them. The resulting set should then be relevant to "pizza" and "San Francisco." However, this doesn't mean each page in the resulting set is necessarily about a particular pizza restaurant. That's why some listing titles may may reflect other topics.
Indeed, this scraping of data and guessing means the service can be prone to error. Go to the second page of localized results for pizza in san francisco, and one of the listings is for a Salon.com article called "Pizza Porn."
What happened? The Salon article is about a pizza place in Ohio where a night manager videotaped himself having an encounter with one of his employees. The article also has Salon's own San Francisco location listed in a footer at the bottom of the page. So, the page is deemed relevant to "pizza" and has a mappable location -- yet clearly, it's not what most people would expect for a search on San Francisco pizza.
On the plus side, if you're looking to scan the whole web for matches and then want to apply a geographic filter, Google's got one of the best ways to go. I certainly like it better than an ordinary Google search for san francisco pizza, if only for filtering out one of those placeholder-style pages that I mentioned last week from the Internet Yellow Pages World Wide that promises information but actually shows nothing on the topic.
Be aware that Google Search By Location only works for US addresses. Google says it hopes to expand the service to other countries, as it is developed further.
The Past Is New
Google's service is treading on trails already blazed by others. Back in April 2000, Northern Light debuted what I then dubbed "geosearching," a first for any major search engine. Through Northern Light's GeoSearch page, as with Google now, you could bring page pages that were deemed relevant to a particular subject and which also were from a particular geographic area.
Northern Light lacked the mapping that Google offers, but another company predates Google with this feature: Lasoo. That company rolled out a service in mid-2001 that looks extremely similar to what Google offers now. Lasoo was never a major search player but instead hoped to partner with others, as I wrote at the time:
Ideally, we'd see Lasoo or a similar service as a complement to a regular search engine. Imagine searching at Google, then selecting a "map results" link to geovisualize your results, when appropriate. It would provide a new, useful view of listings we've not had available before.
Two years later, this type of capability has finally arrived Google. However, it hasn't come through Lasoo. That company was purchased by Overture in mid-2002, and Overture is using the mapping technology for its own local listings service.
Mobilemaps Offers Open Source Local Search
Northern Light's GeoSearch no longer exists, lost when Northern Light closed its public web search service last year. Lasoo is gone, and Overture's local product doesn't involve crawler-based listings. So is Google the only choice, if you want to search the web and have the listings mapped geographically? Pretty much, though High Country Software hopes its Mobilemaps product will change that.
Created by two brothers, Philip and Peter Abrahamson, Mobilemaps was officially launched just last week. As with Google, the service allows you to map results that have been found through web crawling.
Unlike Google, no web-wide demo of the service is available. Instead, the demo only has listings relevant to California. Don't expect the demo to grow, either. That's because Mobilemaps isn't trying to draw users to its service. Instead, it has released its technology as an open source local search solution for others to use.
The goal is that many different web sites may create their own local search engines. Why give this away? In hopes that those using Mobilemaps will also decide to use the NearbyAds service that High Country is also offering.
Behind The Scenes
So what magic is happing to make crawler-found results geographically relevant? Both Google and Mobilemaps, like Northern Light and Lasoo before them, deem a page to be relevant to a particular location if they find a geographical address on them.
For example, imagine there's a page from legal firm that specializes in personal injury lawsuits. The page uses words like "personal injury" and "lawyer" on it. It also has office locations mentioned, such as a sentence that says:
We have offices in San Francisco, Chicago, New York and London
This page might come up for a search on "personal injury lawyer" localized to "san francisco," since it has both core elements needed, words that match the subject someone is looking for and an address that matches the area someone is seeking.
Unfortunately, the page above can't get mapped, since "San Francisco" is a pretty broad address. That's why pages with more specific information should find favor in the Google and Mobilemaps systems. Imagine the same page saying this:
102 Van Ness Ave, San Francisco, CA, 94102
2600 Cole Ave, Dallas, Texas, 75204
430 Park Avenue South, New York, NY, 10016
Breakleg House, The Strand, London WC2 1XX
The locations above are now specific and in a pattern that can easily be spotted by address-seeking algorithms. Moreover, once you have an address, you can convert that into a latitude and longitude, essential for then mapping locations visually.
"We look for a ZIP code pattern with a US state and ZIP code first on a web page. That indicates there's an address, so we look back a few characters to see if theres an address, then look for pattern match," said Philip Abrahamson, cofounder of Mobilemaps. "Then with data available for free from TIGER [a US Census Bureau mapping system”, we can convert the address into a coordinate."
A similar thing happens with Google, where locations are gathered by looking for them on the pages themselves, then transformed into coordinates that can be mapped, the company confirms.
"We determine the geographic nature of a page by analyzing all the information on a page. As you can imagine this includes, but isn't limited to, information such as mailing addresses and phone numbers," said Google spokesperson Nate Tyler.
If you're trying to be found in these new systems, you need to take a close look at your web pages and make sure that you are being both descriptive about the products and services you provide as well as locations. My past article on the former Northern Light system has some further advice that's directly applicable to the systems that Google and Mobilemap are now using.
Google's system is only for US addresses, while Mobilemaps is for US and UK addresses. What about the rest of the world? That remains to be seen.
"What we hope is that some open source programmers in these other countries will work out the details for those countries," Abrahamson said.
If you are outside these countries, to prepare for this, make use of addresses where appropriate and simply list them in a common format for your country, such as street address, city, postal code, country. That should have you set for the future.
As previously said, Google also expects to expand its system beyond the US, in time.
A few last behind the scenes comments:
Google may return web pages that have multiple addresses yet choose only one of the addresses as relevant, as in this search for san francisco hotels. All the listings in the top results come from the same single page on a hotel discount web site.
Unfortunately, the company won't comment on how the decision is made as to which address is selected when many are shown. Google's FAQ for Search By Location does say that pages with multiple addresses are deemed relevant to multiple locations.
Mobilemaps uses some intelligence to return pages that may not have an address on them. In short, imagine it sees a collection of pages from a particular web site. If one of those pages has an address on it, that address might be associated with the site's home page. As a result, the site home page might turn up in the listings.
Geo Meta Tags?
How about meta tagging? Wouldn't be great if everyone just added their geographical locations to their pages through meta tags? Sure, but for all the usual problems with meta tags, this isn't likely to be trusted by search engines.
Having said this, Gigablast has just rolled out its own set of geo-sensitive meta tags that you can make use of. You can add them, but they currently only work with that particular search engine, which also has nowhere near the traffic of any of the majors. It won't hurt to have them on your pages if they are spidered by the others -- but neither will they help or make up for missing physical locations in body copy.
Earlier this year, Metamend released a new product to place geographical GIS data into special Geomend meta tags on web pages. Similar to the Gigablast tags, these are not supported by any major crawler-based search engine.
gstart is a open source system that predates both Metamend's and Gigablast's moves to add geographical meta tags to web pages. The system works by letting you add your latitude and longitude into meta tags for your web pages. High Country Software, the company behind Mobilemaps, owns the system. It emerged in 2000, but no major search engine ever made use of it. Now High Country Software plans to relaunch the idea in the near future.
Selling Local Search: Google's Catch Up
Overture's local search program, which I described in the first part of this series, only lists people who've paid to be in the program. Akin to real world yellow pages, if you don't pay in some way, you don't appear.
In contrast, the Google and Mobilemaps products I've covered in this article involve free listings. The listings are found by crawling the web and included without charging anyone to take part.
The downside to such a system is that, as mentioned, it relies on unstructured data. That can make for false matches and not lend itself to greater details that you may find from systems like CitySearch and online yellow pages. The other downside, from a publisher's viewpoint, is that that it produces no revenue to keep you in business.
So where's Google's local paid listings product, to complement its local search service in the same way that web-wide paid listings complement its web-wide search service?
When I spoke with Google about this in August, the response was that there are no plans for anything special, when it comes to local. The existing AdWords program was seen as working well already for those who wanted to aim locally. Google indicated that those who wish to target specific localities can do so via matching options already offered. In other words, if you want to target those seeking "san francisco dentists," then be sure you've configured your ads to target those words.
Two months later, the tune sounds slightly different: "We do not have any specific product plans to announce today," said spokesperson Nate Tyler, suggesting perhaps Google is now thinking that they need to develop something special.
NOTE: Three days after this was written, Google did unveil a form of local advertising, through a new Regional Targeting feature. This feature lets advertisers choose their ads to be seen by those only in certain areas of the United States.
Indeed they might. Google Search By Location is only a research project right now, so Google can be forgiven for not having the ad component ready. But at some point, you'd expect Google will make it a regular service. Google's existing advertisers will want to appear alongside the editorial listings offered -- and if the map continues to be a centerpiece of local search results, they'll want to be mapped on it, as well.
To do that, Google can't get by with keyword matching. It will need some type of greater structured system similar to what Overture is doing.
Earlier this month, InterActiveCorp CEO Barry Diller, whose company owns CitySearch, said that his company has talked with Google about partnership opportunities. When I spoke with CitySearch about this in September, it seemed to be about ways that CitySearch's own web site pages might be integrated into Google's regular results. However, it could be possible that Google might want to make use of the local paid listings product that CitySearch recently launched, which would mesh well with Google's editorial results. We'll explore these CitySearch paid listings more in part three of this series.
Overall, I'd expect to see Google take some action soon, especially given that the company never likes to be seen as falling behind its competitors. Google's has had local search technology of its own compliments of a programming contest it ran last year (FYI , the programmer that won now works for Google). Yet it was only after many headlines last month about Overture's local search demo going live that Google finally posted its own service, presumably to look competitive in the local space.
That demo may relieve some press pressure from reporters who now see Google having a "local" product. However, it does nothing for the advertiser pressure to come. Once Overture's program goes live later this year, advertisers will see a new way to buy local ads. They'll want the same from Google.
In addition, as we'll see in the next part, there's a strong case for paid listings producing good relevancy when searching goes local.
Selling Local Search: Mobilemaps' Dream
Unlike Google, Mobilemaps does have a paid listings program designed to complement its editorially generated local search results. Called NearbyAds, these are sold on a cost-per-click basis to advertisers. All listings are associated with a geographical location, making it possible to map them alongside editorial listings that are displayed.
Up to two ads will be shown. To appear, your ad will have to be geographically relevant to the search results being displayed. If there are more than two ads, then those paying the most per click get selected, Mobilemaps said.
NearbyAds remains more a business plan than a reality. Currently, the service has no distribution and no advertisers. Abrahamson hopes this will change if people begin making use of its Mobilemaps software. Since the launch last week, there have been 50 downloads, he said.
The company also wants to proactively approach possible distribution partners.
"We hope eventually to attract local portals like local newspapers, so we'll make some effort to bring on board people like that," Abrahamson said.
Coming Up Next
Next week, I'll be spending time looking at the local search product being rolled out by local leader CitySearch. I'll also be recapping online yellow pages resources that are available. Many major search engines offer good, yellow page style listings already, but users may be unaware of these. I'll guide you to these options plus highlight how the search engines themselves are looking to make local search magically happen when it should, without the user doing anything special.