Address capture technology has come a long way in recent times. I remember when the ONLY way to quickly capture an address was through a “reverse drill-down” methodology, but thanks to Google and the prevalence of address capture in online stores predictive “type ahead” address entry has now become commonplace (if you have an online store and still don’t have address capture you might want to consider it). Not only does it provide a better user experience, it also ensures accuracy of information being collected by the business.
Most of the solutions now on offer provide very sophisticated fuzzy matching logic to support predictive address capture. This assists in the presentation of relevant address suggestions as the user types, regardless of how the address is being entered. Unfortunately, even the most sophisticated logic is not without its limitations and solutions that rely solely on this simply create a large number of costly “exceptions” that must be handled manually.
So what causes these exceptions? And believe me – they do happen!
Dependency on reference datasets
The first issue with relying solely on predictive address capture logic (as good as it is) is that it will only return a matching address if it exists in the reference dataset. Most vendors of Australian address capture (including yours truly) reference the Australia Post Postal Address File (PAF) and/or the PSMA Geocoded National Address File (GNAF). Whilst these are very good, they do not offer 100% coverage and there is often a delay from when an address is created to when it appears in these datasets.
Further, with the significant growth in property development currently being experienced this can easily translate to a large number of exceptions (especially if your business is in an industry that targets new developments such as insurance and utilities).
Simply put, no level of sophistication in predictive address capture will find an address if it is not in the dataset being referenced.
Vanity or misinformation
The second issue that renders most address capture solutions ineffective is that many people don’t know where they really live. I know this might sound strange, but it’s a real issue caused by a number of factors. These include individuals providing localities that might border their own because these may be deemed more salubrious or inventing localities by adding “heights” or other labels that invoke visions of grandeur.
Of course, there are also those whose locality and even street names may have changed (due to boundary or other changes) and who genuinely don’t know they are providing incorrect details.
Another favourite is illegal subdivisions, but that’s a topic for a later date!
No matter the reason, predictive capture on its’ own cannot accommodate these errors, irrespective of how good it might be.
Just enough to be commercially viable
And now the final point that will probably see me extradited from the industry – most providers of address capture technologies are lazy! Most develop their solutions to the point where they are commercially viable – i.e. good enough that they can convince someone to buy it. What this means is they develop their solution to be compliant with the authoritative body rules (in this case Australia Post AMAS) and/or be suitably competitive with the majority of other market participants.
Now, just to make sure I get everyone offside I’ll openly state there is nothing wrong with this! This is typical for most software vendors. Technology providers need to be commercially viable (turn a profit) in order to continue their R&D and improve their offering to the market.
But where I do have a problem is vendors selling these technologies as the golden panacea of address capture and charging premium price.
The reality is, if you have a large enough budget to license an address dataset and access to a competent developer who will take the time to read and code against the relevant rules you can pretty easily develop an address capture tool.
However, whether it is choosing a vendor based on their predictive address capture prowess or being fortunate enough to afford having a go at building your own – it’s still not enough.
The complete solution
A truly complete address capture solution should offer advanced predictive capture logic, but combine this with proven real-time address cleaning capabilities.
I am not talking about taking the exceptions and passing them through a batch validation solution (although that is always an option). What I am talking about is being able to cleanse and format an entered address, even when it cannot be validated during the data capture process.
Fortunately, DataTools have been cleaning address data since long before reference datasets such as the PAF existed. As such we can provide address cleansing capabilities that are not dependant on reference data and can cleanse and format addresses even if a match cannot be found. We call it Advanced Address Repair (AAR)
To find out more about our AAR logic at http://kleber.datatools.com.au/method-description/?MethodName=DataTools.Repair.Address.AuPaf.RepairAddress or trial it for yourself at http://kleber.datatools.com.au/predictive-australian-address-capture/