The internet has evolved into a collection of resources intended to appeal to the large populous of users. Don’t get me wrong, this seemingly endless mass of information is a jovial tool, but it has also diluted the importance of localization. We often find our clients in the dilemma of providing a local service and also needing to have a national presence.
Take for example a medical practice wanting to serve local patients suffering from hair loss in metro code #636 -- Harlingen-Weslaco-Brownsville-McAllen Texas. The practice wants to appear to be local because this is to their advantage in the Harlingen, Texas area. In our example, the medical clinic earns 60% of their revenue from selling an organic topical cream to stimulate hair growth. E-commerce sales come from all across the United States.
The company is spending $48,000 per month in Google pay-per-click advertising. Selling both local and national on the same website presents problems because neither is best optimized for customer acquisition and conversion. This usually manifests in poorly performing PPC campaigns or a high cost per acquisition metric. Using different URLs is usually not an option due to cost and duplicative content/SEO concerns.
What now?
As an internet consulting firm, we often deal with these sorts of questions. Companies are looking for clientele on a global level and still want to accommodate local clients. Our solution is GeoIP or geolocation by internet protocol address (IP address). Using GeoIP, clients can reach out to prospects on a global level and still tailor services to prospects, members, customers or patients who are within driving distance.
There are a handful of solution providers offer geolocation by IP services. Most offer a hosted webservice on a subscription plan tied to a number of data queries (vendors like Quova, SiteBrand and ISAPILabs Country Locator). These services tend to be very expensive for high volume websites. With a hosted solution a XMLHTTPRequest object can be used to send IP address data to the vendor and get back location data. This all happens under the hood of a website, typically in the global.asax or tracking manager scripts (we're using ASP.NET here).
There are four credible sources to acquire raw data. MaxMind and IP2Location tend to be the largest players in the data market space. In our testing, HostIP (a free data provider) seems to be less reliable. Mining the likes of InterNic is very time intensive and is not practical. The commercial data providers have advanced techniques to scrub publicly available data and tend to be more accurate. We tend to deploy custom solutions using data provided by MaxMind. Data costs around $1k per year including updates.
How it works
When a person visits a website, code embedded in the website references a database and determines a users’ country, region, city, postal code, area code… Furthermore, GeoIP can provide information such as longitude/latitude, connection speed, ISP, company name, domain name, and whether the IP address is an anonymous proxy or satellite provider.
The notion behind GeoIP is simple. We deploy a data set consisting of:
- IP Numbers
- Longitude/latitude
- Country
- US/CA Regions
- City
- Sate
- US Zip Codes
- US Area Codes
- US Metro Codes
The dataset is imported into a series of SQL database tables in a database. The website converts a user’s IP address (accessible from most any website) to an integer using a proprietary formula. In our case we use a custom .NET class that we at D3 developed long ago. The integer is looked up in our database and the resulting recordset will contain Country along with additional information listed above. This .NET class is added to the sites tracking manager such that each website visitor will carry a session which contains the country code, state...
The website is augmented, most often using <if then> blocks our URL re-writing, to display dynamic information (i.e. DNIS [phone numbers], links, graphics, call-to-actions, order forms, different navigation, etc) based upon a region code or country. We recently used GeoIP to handle USA/Canadian customers. We had a .com and .ca website. Our solution redirects visitors to the appropriate site. When a user with a Canadian IP address clicks a link to our .com site or types it in directly into a browser we redirect the user to the .CA domain.
In our example above, the home page typifies a holistic medical practice inside metro code 636. Outside metro code 636, the website appears to be a national leader in holistic hair loss. Graphics, navigation and text flip depending on the geography. The solution is elegant and very powerful. The end result is serving two audiences on a single website. Hands down, the technique provides greater conversion and lower acquisition costs.
A deeper look
Let’s drill a bit deeper for those technical folks. In order to determine the location of a unique visitors IP address needs to be cross referenced in our database. The IP numbers supplied by our vendor have a beginning and ending IP range. This is done to keep the database optimized and under 6 million rows. When a unique visitor’s IP is being cross referenced, it is being checked by the database to see if it is within an ‘IP range’. The beginning IP number and ending IP number are set as whole numbers and not the common IP structure with octets (ie. 192.168.1.1) for performance reasons (i.e. integers are easier to lookup than nvarchars). We convert an IP into an integer using the following calculation:
IP address number = 16777216*w + 65536*x + 256*y + z (1)
where
IP Address = w.x.y.z
The reverse of this formula is
w = int ( ipnum / 16777216 ) % 256;
x = int ( ipnum / 65536 ) % 256;
y = int ( ipnum / 256 ) % 256;
z = int ( ipnum ) % 256;
Where % is the modulus operator.
Once the ip number has been converted to a whole number, it can be easily looked-up from the database using a stored procedure.
If you’d like to know more about how we use GeoIP technology, give me a call or send me an e-mail. I’d be happy to provide even greater insight. There are really interesting areas that need consideration. Things like:
- How do different search engine spiders respond to GeoIP?
- Does Google/Yahoo/MSN see this as a ‘black-hat’ technique?
- What to do about Canadian AOL user identification problems.
- Accuracy - 99.8% accurate on a country level, 90% accurate on a state level, and 83% accurate for the US within a 25 mile radius.
- IP version 6 addresses.
- How to handle anonymous proxies.
- Satellite provider identification issues (mostly Africa and the Middle East).
- Circumstances where there are no clear indications as to where the end users are located. For example, we may mark a corporate proxy located in London, GB as "EU" if we can see that the ultimate end users are connecting from various parts of Europe.
- What to do about changes - 1-2% of the GeoIP database changes each month.
|