AI parsing saves millions by delivering better address verification and enrichment capability

Whether shipping a product overseas or offering an online subscription, cross-border businesses can collect up to millions of customers addresses every day. On a global scale, this means addresses with a range of formats are cycling through databases, some addresses more complex than others. Although these addresses may have region specific nuances, there are still ways to detect a pattern and reformat any input data into a workable format. However, the manual version of this standardization process is as laborious and unreliable as one would expect.

Parsing address data with machine learning

To improve data quality and increase confidence in results when validating addresses, our address verification solution uses artificial intelligence (AI) and rules-based parsing models to detect these patterns faster and more proficiently. Our enhanced AI Parsing engine enables greater address verification and uplift rates in existing and hard-to-address markets where input data is ambiguous or incomplete. AI Parser enables greater address match and uplift rates, on average delivering a 7.25% increase in capabilities across countries and territories.

In this post, we’ll share how the Loqate’s Verify tool delivers better address verification and enrichment capabilities globally due to the combination of our Global Reference Data and AI Parsing engine.

Why We Care About an Address

Addresses are a key data attribute that form a customer’s demographic or a business’ firmographic profile. In short, no customer or business record is complete without an address. No matter your industry and specific use case, quality address data plays a role in removing friction, helping prevent fraud, delivering superior customer experiences and enabling cross-border commerce. Of course, the reverse is also true and poses a tangible cost if and when not addressed.

No matter the type of business, it’s guaranteed you’ll have to collect, store, and manage addresses. And this data is tightly linked to the quality of your services, depth of your business insights, bottom-line revenue, and, most importantly, your users' satisfaction.

What is AI Parsing?

To comprehend the true value of AI parsing, we must first define the basic principles of parsing. Data parsing is the process of converting one type of data into another. It’s a form of data structuring that makes existing, often unstructured, data more comprehensible. The same concept holds true for address data parsing. At Loqate, customers often send a single block of address data that they’ve captured from the customer. Parsing is the process where we take that data and segment it out into the proper fields based on rules defined, coded, and frequently reviewed by our developers and data analysts.

For example, with a US address like 555 Greene Street, the parser will see the number 555 and determine that it belongs in the premise field. The data elements after that will be categorized as a street or even locality depending on what the technology thinks are the relevant fields.

However, all parsing methods are not created equally. The value of the results depends heavily on the quality of the formatting rules. While it may be easier to segment a UK or US address thanks to the clear guidance on how those addresses are structured, emerging markets with less standardized addressing systems, like India for example, are more difficult for analysts to navigate. And as we know, human interference leaves more room than welcome for human error.

Here is where AI and Machine Learning enter the equation. Instead of relying on the developer or an analyst to understand and code the rules themselves, address data now passes through our AI parser, which has been trained via machine learning to analyze and classify even the most complex data types. Meaning, with the evolving capabilities of AI, we are more likely than ever to put the right data elements into the right fields, and therefore we can then match it more successfully against any address we have on our in our data repository.

Why AI Parsing Matters to Address Validation

When evaluating the value of a potential address verification solution provider, we ask ourselves several questions. But for businesses looking to transact across borders, the main inquiries are: How comprehensive is the data? And; does this tool work internationally?

Our Address Verification solution gives businesses the capability of having and maintaining the highest quality location data possible, globally and at scale. AI Parsing, a key stage in the verification process, supercharges our ability to handle non-address elements in input data, delivering an incremental 19% address match and uplift rates in developing countries like Indonesia, and significant enhancements in established markets like Germany (+13.8%), USA (+8%) and in the UK (+3.9%). Cross-border commerce never looked so cost efficient and secure, from improving deliveries to helping prevent fraudulent transactions.

To understand the value AI Parsing delivers, you first need to breakdown each stage and component of the address verification process. The process has three main stages and requires an address reference database to verify against. To verify an address, it needs to be parsed, matched (corrected and enhanced) and formatted against a known source of truth, a trusted reference database. A performant verification engine will be able to take a full, unstructured or semi-structured address and return it as a valid, enriched and verified record. AI Parsing’s superior ability to standardize records, identify and remove non-address elements is foundationally critical to enabling the matching and formatting stages of the verification process.

To summarize, here’s how AI parsing helps delivers superior address verification capability:

• • Greater address match and uplift rates
  • Enhanced ability in handling non-address elements in input data
  • Continuous and automated evolution of parsing models

Quantifying the Value of AI Parsing

Let’s look at an eCommerce use case for a business that ships 10M packages cross-border a year to each the US and Germany (Amazon ships ~400M packages a month), taking into account the cost of $17.20 in the US and €14.69 in Germany for a failed delivery, AI parsing’s incremental benefit alone will save that company an additional $1.1M in the US and €1.4M in Germany by preventing more failed deliveries.

Now let’s consider an example in the fintech landscape. To stay agile and sustain growth, fintech's need to build a seamless and satisfying digital customer experience that meets consumer needs. This starts with reliable data and strong data management practices that help you connect with your customers and make informed decisions. Without a long-term investment in data quality, fintech's run the risk of breaking regulatory guidelines, such as Anti Money Laundering (AML) and General Data Protection Regulation (GDPR), that carry a penalty of two to four percent of the company’s revenue. Luckily our match rates are over 23% better than our competitors' and with AI Parsing, our customers can reach and verify addresses from the furthest corners of the world with absolute confidence.

Why Loqate is the best global address verification provider

Loqate gives businesses the capability of having and maintaining the highest quality location data possible through the depth and breadth of our premise level global reference data and superior AI powered address parsing engine, delivered via a single, flexible and performant API.

Our ability to connect people, places, and services through location data is at an all-time high and only growing. With better technology even more enriched reference data, we can equip our customers with the power to launch their business across all borders – especially in significant markets of APAC, South America and India.