Geo-targeting Google AdWords

In the drive toward improving relevance and targeting within the search environment localised search and geo-targeting are close bed partners. The major search networks offer geo-targeted delivery of their paid search programmes, but in the UK their ability to deliver this in a meaningful way is often problematic.

UK IP Detection Limitations for Geo-Location

At present most search engines rely on resolving the user’s IP number to an internet service provider (ISP) - or to be more specific a router belonging to that ISP that the user is using.

The problem is the router ID needs to be registered at a physical location for the location server to be able to return a location for this router - and as router location registration is entirely voluntary for ISPs this data is often absent.

Another issues particular to the UK is the compact nature of our telecommunication structure - which means my unbundled local loop does not equate to user location. My ISP router can be in Portsmouth, Southampton, Reading, or Oxford (all of which I’ve had in the last couple of years) whilst I’m in Bournemouth - so geo-location becomes meaningless other than letting the search engine know which end of the country I’m in when I go online.

• Google current relies on IP number as the default mechanism to determine your location
• IP numbers can have little correlation to physical location in the UK

Why Google Wants Localised Search

Localised search provides Google with three distinct advantages:-

1. For users local results are likely to be more relevant - this will improve their user experience, making them more likely to use the search engine - good for Google.
2. For advertisers locally targeted ads are likely to be more relevant, so higher conversion rates are expected; meaning advertisers are more likely to use the search engines’ paid search programmes - good for Google.
3. For the search networks, local search provides more inventory - impressions can be delivered to different advertisers in different areas for the same keyword search rather than a via country-wide result; this provides significant additional inventory - great for Google.

How is Google Tackling the Problem of Geo-Location?

Like Yahoo! with its Fire Eagle scheme Google has been exploring ways of getting geo-location data more easily. The open source API project that they are sponsoring, Gears is thought to be one route, with the API able to use any geo-location data and provide this to a location server to spit back a longitude and latitude. Gears and Fire Eagle are both opt-in programmes, but the ubiquity of Gears means it stands more chance of becoming a widely used service.

Google also rolled out their SearchWiki recently which provides personalised search for logged in users. Obviously, personalised search is a more refined form of search provision than localisation, but almost all SearchWiki users will experience local results based on geo-location data they have provided Google, or that Google has gleaned based of other data associated with the users’ cookies/log in.

Recently Firefox added the Geode plug-in - this uses the W3C Geolocation spec and is able to use IP Router IDs, mobile phone cell mast IDs, Wi-Fi access point IDs and the built-in GPS chip data from a device - so the methods by which Google will be able to glean location based data is only set to rise.

If one looks at the mobile search market, the coupling of 3G chips into laptops and netbooks, and the increasing demand by users for local results means we can conclude things are changing quickly and the paid search landscape had better get ready for this shift in focus.

Current Limitations - How Can I Make Localised Paid Search Work Now?

When Google setup AdWords geo-targeting, they tested and rolled it out in North America. The IP location issue was not a significant problem in the USA where registered IP ranges are easier to associate with physical locations, but in the UK we are victims of our population density and tightly meshed telecommunications infrastructure.

Obviously Google understands the limitations of IP lookup as the sole basis for geo-location and has made changes to the search query methodology within their AdWords programme. This means that if you add a location name you get localised results both in the natural and paid search results.

• Adwords has been upgraded  to add geo-targeted result when a search query contains the relevant location

However blunt this obvious mechanism might appear it is at least a step in the right direction and means the user can control their result to a degree.

Possible improvements through “network accounts”, APIs and hardware

Google, Microsoft and Yahoo! all offer “local search” but they don’t actually know where you are really unless you say (edit you postcode in your profile). Local search in the UK really means “UK level”. However, it is conceivable we are not far from be presented with much more accurate localised search results tailored to our needs. We will have to see how Yahoo! fair with Fire Eagle and where Google goes with SearchWiki.

Currently Google, Yahoo! and MSN do not augmented simple IP lookup with their user registration programmes (Google Accounts, .Net Passports and Yahoo! ID) or directly use API/Plug-ins such as Gears or Geode which in theory would allow the search engines to predict location with a much greater degree of accuracy.

• Search engines offer services that require user registration (Google Account, Yahoo! ID and Live Passport) for access to services such as web mail
• Registered users’ locations can be gathered through service offerings and user behaviour (using transient/session cookies and persistent/user cookies) to know who is where
• Search engines can deliver results based on the users’ “known” locations.
• Search engines can receive data from location servers if users utilise geo-location tools (Gears/Geode)

The shortcomings for the search engines of using the registration technique is that the user must first be registered with the appropriate search network and secondly they must be logged in or have an active cookie (persistent/user cookie) in play at the time of the search query for the search engine to “know” where they are located.

Account based solutions are not a silver bullet, but are designed to improve the degree of accuracy for those users wishing to participate in an improved local search offering.

Likewise, plug-in / API based solutions will not be universal - with only a portion users adding this functionality. So unless someone like Microsoft adds this functionality as standard to IE9 and it is set on by default it is not going to solve the issue entirely.

The third element to add is hardware changes: (GPS, 3G and wireless chip data) which again will provide a portion of the total user traffic with geo-location data that the search engines can then access.

But, these three elements combined (registration, APIs and hardware) do offer the potential for a significant number of searches to be accurately geo-targeted.

Let’s take a look at the current Google AdWords platform and how we can get the most out of it for local search:-

Local Search

With Universal Search (blended search results) Google spits out local search results (registered businesses in Google Maps) in the main web search engine results page (SERP). So if you are going for a local audience in your area, the first thing to do before thinking about PPC is make sure your company is properly registered for local search.

Paid search geo-targeting tools

Google AdWords’ geo-targeting tools allow you to include and exclude areas within the area of operation. For example if you had a campaign targeting just London users you can exclude other locations in the UK through Google Adwords campaign location settings.

To activate geo-targeting in Google AdWords go to the campaign you wish to geo-targeting and edit the campaign settings or access the location information directly (as indicated in the screen capture below). If you choose to edit the campaign settings the at the bottom of the page under the “Target audience” heading is an edit link for locations.

Your will then be presented with a window titled “Target customers by location” which offers you four ways to select locations, each with their own tab: Search, Browse, Bundles and Custom.

AdWords will break down locations in a logical hierarchy of Territories, Countries, Regions, and lastly Towns/Cities. So for example we have three ascending levels which would all capture traffic associated with London IP addresses if they were selected as the required geo-location.

• London, ENG, GB (Town/City)
• England, GB (Region)
• United Kingdom (Country)

If you want to exclude locations - say you want to target just London - then the quickest way is to remove all locations and then add the town/city you are targeting. There is also an exclude function that allows you to specifically exclude locations. So it is quite possible to include all of the UK whilst excluding a region such as Northern Ireland or a specific town/city such as Birmingham.

It is really straightforward to set up a geo-targeted campaign in Google AdWords - but the question of IP attribution means this simple AdWords tool is powering a service whose accuracy is at best patchy. Will it actually geo-target? Only on a limited basis.

Google have upgraded AdWords geo-targeting in the last year so now any relevant location based search term such as “widgets manchester” would return geo-targeted results for Manchester. However, this is not passive location matching and requires the user to add the relevant location in their search query.

Does geo-targeting improve your quality score?

There is a perception that local search (geo-targeted) ads will appear higher than non-geo-targeted ads for the same query/bid price metrics. This perception is well founded, but only if a location name is mentioned in the search query, and it has nothing to do with geo-targeting, but rather it can be attributed to the existing “quality score” algorithm which measures the relevance of your ad to the search term. Geo-targeting by its nature tends to be more relevant to search queries containing a location match and will therefore gain a boost in Google’s QS algorithm, pushing up the position.

IP locations verses real world locations - working around the problem

Both Microsoft AdCenter and Yahoo! Paid Search fall foul of the same IP attribution issue and are thus equally flawed as Google AdWords. So what else can we do to try and geo-target paid search?

Keyword Matching

Make sure your ad groups contain the relevant location names as part of the keyword phrases so you’ll have matching keywords trigger whenever anyone searches using a relevant location in their search.

As well as using all the keyword combinations related to the targeted geographic locations also be aware that you can further optimise using the matching types such as phrase match and exact match to improve targeting and bid price performance.

Where you’re not - negative match filter (exclusion)

You can use negative matching to exclude searches e.g. scotland would exclude searches for ‘widgets scotland’. Negative matching is a very powerful way of improving geo-targeting where the searcher is looking for a geographically based service/product and uses the location in the search query. However, again this has its limitations and will only work to exclude irrelevant searches if the user actively searches using a location based word you are excluding.

Filtering with you advert copy

Remember that the human viewing you advert is a sentient being and is probably somewhat literate as they have used a search engine to initiate a query!

This means you can tell them where you operate in the advert copy. So give the searcher a geographic/location context for the products/services you are offering if it needs to be geo-targeted. This will help with targeting and click-through-rates. It will also act as an exclude filter of sorts, backing up the message of where you are and where you are not.

This is important, because the user might not use a location in their search, so negative matching will not have the opportunity to trigger and prevent the advert displaying. In this case your adverts’ copy is going to try and preclude the user clicking on the link when it is not relevant to them.

The downside of this tactic is that to work it will reduce click-through-rates (CTR) and thus impact your quality score. However, reducing unwanted irrelevant clicks and their associated costs will always take precedence over any impact on your quality score which might be marginal.

Quality Score Benefits

When a location used in the copy matches the search query you will receive a boost in your AdWords quality score for that ad. This provides a three fold benefit:-

1. the traffic is more qualified traffic
2. your cost-per-click should be reduced
3. your budget will go further.

It is all obvious stuff, but it is amazing how many AdWords accounts we’ve taken over from reputable “Search Specialists” that could not get this basic marrying up of Ad Group keywords and copy right.

Granularity means better performance. So optimise early to receive the long-term cost and position benefits.

Landing page & website optimisation

Further improvements to localised targeting

You can further improve performance and relevance by including the location data in the landing page and within the website’s information architecture to improve you adverts’ quality score. An advert’s bid price and positioning performance will thus be enhanced by association with the landing assets’/website’s where keyword matching occurs. So if the user searches for “widgets Birmingham” and the advert’s landing page is called “Widgets Birmingham” in the HTML title and heading, and the website’s navigation also has “Widgets Birmingham” as a link to this page the advert will receive a boost in the quality score. This will all help improve geo-targeting by improving your relevance for any users searching using a location based query.

Be explicit in your landing page copy about where you operate. Think about a landing page for each town/city if they are important to your commercial operation. If you operate in Hampshire, Southampton, Portsmouth, Eastleigh and Winchester say so in your copy. Have a page specifically about each of these locations and the products/services offered there.

If you can link to these landing pages with navigation links, all the better. This will allow your website’s information architecture to help contribute to improving you quality score, so when someone searches for a location based search term you will be better placed in the AdWords results for a lower cost.

Conclusion

We are probably on the cusp of true local search offerings from the search engines - everything they have said and done indicates we are on a gradual path to this end, and we’re not far off.

Registered users can currently volunteer their locations to the search engines in order to receive more locally relevant results. Outside of “logged in” registered users we have Google Maps providing universal search results offering something approaching a local search, but it is on a very limited basis - local businesses have to register their address and offering - and the results are tied to a location in the search query - so a bit limited.

In the meantime we must work with the existing way web search operates, and use the tools of the existing AdWords programme to fulfil the matching and filtering process.

• Make sure your keyword list contains target location names
• Make sure your ad group copy (Titles, description and display URL) mention relevant location information
• Make sure your negative matching has any location names you wish to exclude
• Turn on geo-targeting and optimised to your needs (remember to specify only locations you wish to be matched for, or use excludes to make sure you do not come up for certain locations)

These measures are pretty straightforward; there is no secret here or magic short cut. Like most things in online marketing it is about systematically tackling the list of things that need doing. Start by drawing up a list of all relevant locations - priorities which ones merit a landing page, information architecture and ad groups, then set about setting these up. For less important/minor locations consider using more generic ad groups and landing pages with dynamic keyword insertion (DKI) in the ad copy to handle relevance.

We’ll keep you posted to any major changes to geo-targeting and localised search as the search engines roll them out.