How do you perform address validation?

Question

Is it even possible to perform address (physical, not e-mail) validation? It seems like the sheer number of address formats, even in the US alone, would make this a fairly difficult task. On the other hand, it seems like a task that would be necessary for several business requirements.

Answer 1

For US addresses you can require a valid state, and verify that the zip is valid. You could even check that the zip code is in the right state, but beyond that I don't think there are many tests you could run that wouldn't provide a lot of false negatives.

What are you trying to do -- prevent simple mistakes or enforcing some kind of identity check?

Answer 2

It is not so difficult. There are several services that sell zip code data, and verification services. Also US census data has "most" every street(and #) in their data, but it's a bit of a task to get it out and make it useful at first go.

Answer 3

For us-based address data my company has used GeoStan. It has bindings for C and Java (and we created a Perl binding). Note that it is a commercial product and isn't cheap. It is quite fast though (~300 addresses per second) and offers features like CASS certification (USPS bulk mail discount), DPV (Delivery point verification) flagging, and LON/LAT geocoding.

There is a Perl module Geo::PostalAddress, but it uses heuristics and doesn't have the other features mentioned for GeoStan.

Edit: some have mentioned 'doing it yourself', if you do decide to do this, a good source of information to start with is the US Census Tiger Data Set, which contains a lot of information about the US including address information.

Answer 4

There are companies that provide this service. Service bureaus that deal with mass mailing will scrub an entire mailing list to that it's in the proper format, which results in a discount on postage. The USPS sells databases of address information that can be used to develop custom solutions. They also have lists of approved vendors who provide this kind of software and service.

There are some (but not many) packages that have APIs for hooking address validation into your software.

However, you're right that its a pretty nasty problem.

http://www.usps.com/ncsc/ziplookup/vendorslicensees.htm

Answer 5

As mentioned there are many services out there, if you are looking to truly validate the entire address then I highly recommend going with a Web Service type service to ensure that changes can quickly be recognized by your application.

In addition to the services listed above, webservice.net has this US Address Validation service. http://www.webservicex.net/WCF/ServiceDetails.aspx?SID=24

Answer 6

We have had success with Perfect Address.

Their database has all the US street names and street number ranges. Also acts as a pretty decent parser for free-form address fields, if you are lucky enough to have that kind of data.

Answer 7

Validating it is a valid address is one thing.

But if you're trying to validate a given person lives at a given address, your only almost-guarantee would be a test mail to the address, and even that is not certain if the person is organised or knows somebody at that address.

Otherwise people could just specify an arbitrary random address which they know exists and it would mean nothing to you.

The best you can do for immediate results is request the user send a photographed / scanned copy of the head of their bank statement or some other proof-of-recent-residence, because at least then they have to work harder to forget it, and forging said things show up easily with a basic level of image forensic analysis.

Answer 8

I have used the services of http://www.melissadata.com Their "address object" works very well. Its pricey, yes. But when you consider costs of writing your own solutions, the cost of dirty data in your application, returned mailers - lost sales, and the like - the costs can be justified.

Answer 9

Here's a free and sort of "outside the box" way to do it. Not 100% perfect, but it should reject blatantly non-existent addresses.

Submit the entire address to Google's geocoding web service. This service attempts to return the exact coordinates of the location you feed it, i.e. latitude and longitude.

In my experience if the address is invalid you will get a result of 602 from the service. There's definitely a possibility of false positives or false negatives, but used in conjunction with other consistency checks it could be useful.

(Yahoo's geocoding web service, on the other hand, will return the coordinates of the center of the town if the town exists but the rest of the address is bogus. Potentially useful as long as you pay close attention to the "precision" field in the result).

Answer 10

There is no global solution. For any given country it is at best rather tricky.

In the UK, the PostOffice controlls postal addresses, and can provide (at a cost) address information for validation purposes.

Government agencies also keep an extensive list of addresses, and these are centrally collated in the NLPG (National Land and Property Gazetteer).

Actually validating against these lists is very difficult. Most people don't even know exactly how their address as it is held by the PostOffice. Some businesses don't even know what number they are on a particular street.

Your best bet is to approach a company that specialises in this kind of thing.

Answer 11

USPS has an address cleaner online, which someone has screen scraped into a poor man's webservice. However, if you're doing this often enough, it'd be a better idea to apply for a USPS account and call their own webservice.

Answer 12

I will refer you to my blog post - A lesson in address storage, I go into some of the techniques and algorithms used in the process of address validation. My key thought is "Don't be lazy with address storage, it will cause you nothing but headaches in the future!"

Also, there is another StackOverflow question that asks this question. Entitled How should international geographic addresses be stored in a relational database.

Answer 13

Yahoo has also a Placemaker API. It is good only for locations but it has an universal id for all world locations.

It look that there is no standard in ISO list.

Answer 14

Fed Ex has a free AddressValidation web service. The have sample .NET (C# and VB) examples but I hit a road block.

While the code works in the Colsole app example, if you simply change he application to a library or a winform app, I get the error

Unable to generate a temporary class (result=1). error CS0029: Cannot implicitly convert type 'AddressValidationWebServiceClient.AddressValidationWebReference.ParsedElement' to 'AddressValidationWebServiceClient.AddressValidationWebReference.ParsedElement[]'

Here's an explanation, but the fix is beyond me.

http://www.dreamincode.net/forums/showtopic74653.htm

I am sure that the fix is small, but it's probably beyond me, so I am not sure where to go from here.