Processing addresses automatically and correctly is a challenge. Whether it concerns billing addresses, delivery addresses, or contract partners: varying sequences, abbreviations, or omissions regularly require manual intervention to ensure a process can be completed successfully. We show how a clever combination of proven techniques and well-known services can eliminate this problem.
The Everyday Challenge
A specialist sits at their screen, frustrated: A major client has submitted numerous delivery addresses in an order. While entering them, one employee left out the city. Another employee mistakenly entered the street name twice in the address field. A third colleague preferred to enter the delivery address in a comment field and added extra instructions. The assumption behind this: the recipient will find and fix everything anyway. This is the daily grind in a world of divided labor.
That is why human creativity should not be wasted on monotonous tasks if these jobs can be handled through automation—at lower costs and with higher processing speeds.
For the recipient of such electronic documents and datasets, data entry is a nightmare. Address data appearing in different places within a document, often with varying levels of quality, frequently necessitates complete manual re-entry into downstream systems. But it doesn't have to be this way!
Heterogeneity in Address Data Is Manageable
With the help of a combination of intelligent text interpretation and a geocoding API such as Google Maps, Here, what3words, or Geoapify, you can work wonders. Using proven RegEx methods, an address can be broken down into its components: company, street, house number, city, postal code, country, etc.
If it turns out that parts of the full address are missing, they are automatically searched for and retrieved via an API call to one of the aforementioned geocoding services. The validation routine then inserts the missing parts, ensuring the address is complete and correct.
This also works across different countries, as RegEx patterns can be "injected" into the parsing algorithm—meaning they can be dynamically fed from outside the automation process. Furthermore, this provides an opportunity to log which addresses or documents cause the most errors, allowing for systematic improvements through consultation with the originators.
Free Text, Comments, and Remarks Can Be Read Automatically
We use a similar approach for extracting address information from unstructured fields like comments, free text, or remarks. This is a particular challenge, as no off-the-shelf service offers this capability.
Many business partners write address information into any available text field, assuming the recipient will recognize and use it thanks to human intelligence. However, that cannot be the goal. Therefore, our solution-oriented creativity was required.
Here, too, machine processing must step in to identify unstructured address data in various free-text fields and prepare it for further processing. To this end, we have developed a special algorithm that can be utilized in any business process automation. To verify the result, a geolocation API call can follow, ensuring accuracy.
Geocoding Offers More Than Just Address Validation
Geocoding services offer other interesting functions as well: an address can be transmitted as a whole to Google Maps, what3words, Geoapify, or Here. The response from the called service usually contains further information, such as latitude and longitude or the type of address.
This allows you to determine whether a location is a building, a street (including its characteristics), a restaurant, or a specific business type. This is relevant in cases where additional data, such as transport restrictions for trucks, is useful for route planning.
Conclusion
Accuracy and completeness are prerequisites for successful automation. Only reliable results are useful results. Otherwise, the manual effort required to correct faulty data outweighs the benefits of machine processing.
When implemented correctly, however, the desired quality of results can be achieved quite effectively at manageable costs.
