Address verification is a vital tool for many successful businesses. It helps boost conversion rates, improve customer experience, reduce failed deliveries, and prevent fraud.
But as new markets emerge and the global commercial landscape evolves, our customers’ needs inevitably change. Which is why we’re constantly striving for ‘better’. Better quality levels, better services, better product, and better outcomes.
Our AI parsing capability already plays a key role in this relentless pursuit. It’s enabled us to provide customers with enhanced verification performance and even greater address match rates. And looking ahead to the future, it will be instrumental in helping us create our own data for hard-to-address areas and allowing our customers to enter new markets.
Our product manager, Gunaseelan Krishnamoorthy, explains how.
In order to validate a global address, we first need to identify the individual elements in that address, such as the house number, street, city and postcode. That process of identifying these elements is what we call parsing. It’s where we take a single block of unstructured address data and segment it into the proper fields based on a set of rules that are defined, coded and reviewed by our developers and data analysts.
As you can imagine though, the results depend heavily on the quality of the formatting rules. In countries like the UK where there’s clear guidance on addressing structure, it’s easier to segment the data. In emerging markets with less standardised addressing however, it’s much more of a challenge.
In addition, as time passes and new areas are added or new addressing formats appear in a country, this complex and ever-growing list of rules becomes harder to maintain.
That’s where AI and machine learning come in. Instead of relying on developers or analysts to understand and code the rules themselves, our AI parser has been trained via machine learning to analyse and classify even the most complex data types. This means we can put the right data elements into the right fields with even greater accuracy, allowing us to match data more successfully against any address we have in our data repository.
However, like any AI application, it’s only as good as the human input behind it. That’s why we have a team of data scientists who spend their time researching and studying global addresses to train our machine learning models.
Our data creation initiative aims to diversify our sources of address data and use proprietary methods to identify addresses that can be used to increase our coverage of valid addresses in target countries.
When we receive address data from external sources, we give it an address confidence score. Some sources are more credible than others and some have better structured data than others. Only those that meet a certain level of confidence can become candidates for inclusion in our Loqate dataset.
But relying solely on external data sources has its limitations, as there are some countries with poor data coverage. This might be because either there simply aren’t enough addresses for that country or the current usage of our customers in that country is low.
These are areas where it takes a lot longer to build up a rule-based knowledge base from scratch. AI parsing reduces the need to maintain a long list of complex parsing rules and helps build our learning and coverage in these countries more quickly and efficiently.
In essence, by applying AI parsing to the data creation source, we can correctly identify and find even the most complicated addresses, which can then be added to our global dataset.
We don’t just simply acquire datasets at Loqate. We are developing the capability to figure out addresses on our own and supplement datasets with data that we’ve created.
Historically, if an address was missing from a dataset, we are reliant on external timelines to research and update that for us. But today, with our AI-backed data creation process, we will soon be able to respond to customer enquiries for new addresses and new locations faster than ever before.
Our knowledge and experience with machine learning is only growing. We continue to experiment with new ways of increasing our address verification rates and expanding our ability to provide address suggestions with fewer keystrokes.
Currently, our work is focused on our AI parsing machine, but we’re looking at employing machine learning across even more areas of our product. This will allow us to enter new markets and offer address solutions in countries that no one else can, adding more value and creating new opportunities for our customers.
We’re a global company but we think locally. For a long time, address data has been sparse and hard to come by for many areas of the globe. But with our global data coverage, capability of our AI parsing engine, and in-house global addressing expertise, we’re changing that.
With address verification from Loqate, you can now reach beyond borders – even in emerging and hard-to-address markets.