Address matching - It's not rocket science
Software developers new to the subject of address matching assume that it involves fuzzy string matching. It does, but that only accounts for about 10% of the problem for Irish addresses. The following are a sample of the challenges our team has overcome during the past 15 years.
Non standard addressing
Most countries have the concept of "vanity addressing" where an address presented differs from the postal address in order to make it appear to be in a more upmarket location. Due to the absence of an address standard in Ireland at least 50% of addresses entered by the public differ from their Postal Address. The graphic below shows typical examples of addresses encountered on the left versus the postal address on the right.
When initial attempts at address matching fail, developers usually look at examples that don't match and then come up with a series of rules to increase match rate. No matter how smart you make your algorithm, there is no foolproof method of increasing the match rate without increasing the error rate. There are always exceptions that catch you out. The solution? Unfortunately there is no short-cut, we painstakingly research each variation and create verified alias data that is incorporated into our address matching engine (see details here). This increases match rate without compromising error rate.
Postal Address Standard
Autoaddress are supportive of An Post's efforts to standardise addressing in Ireland by encouraging everyone to use their Postal Address. However your customers must be able to enter addresses in their preferred format if they do not want to use their Postal Address.
Any Eircode to address lookup service for quick address entry must allow a user to change the address to match their preferred format. Irish people have got quite used to putting 90210 into postcode fields and then entering whatever address they like. So what do you do? Do you reject addresses that don't match to Postal Addresses and lose business? Alternatively you could just let every change through, but then you have no validation.
Autoaddress have identified and solved a range of issues relating to differences between Postal Addresses and customer preferred format as follows:
- They're just different
At least 50% of address have extra/missing/different address lines to the Postal Address. We have the largest database of address aliases in Ireland in order to verify addresses in user's preferred format.
- Rural Townlands are Fuzzy
Townland boundaries are legally defined by Ordnance Survey Ireland. However local convention rarely takes these boundaries into account, so the first line of a rural address will often be different to the postal address. It may be a neighbouring townland, or it may be a completely different name. You cannot rely on the first line of a postal address for an eircode to match the first line of a users preferred format. Our solution is a combination of spatial logic and data insights from the 30,000 addresses we capture every day.
- Preferred Addresses
Autoaddress has analysed the address format used by the majority of addresses on each thoroughfare and townland in Ireland. Our solutions are Postal Address compliant by default, but you can toggle them to return user preferred format addresses instead.
Autoaddress has performed a number of Encoding Quality Assurance assessments for organisation that either built in-house solutions or used competitor solutions. The minimum error rate detected for unique addresses was 2.5%. The maximum was over 5%.
Our error rate is over 90% lower than the best we've measured, independently verified as part of our encoding of 46 million address records in Government databases. The statistical report is included below.
Your IT team is too busy to undertake this mammoth task, your address data is too important to append incorrect Eircodes. Choose Autoaddress.