The last time you used your zip code, you were most likely entering your address into a website to make a purchase, finding a store near your home or office, or filling out some other online form. You likely found the answer you were looking for and didn’t stop to think further about that five-digit code you’d just typed out.

However, lots of companies, marketers and data analysts spend hours looking at zip codes. They are deciding how to use data tied to those zip codes to understand trends, run their businesses, and find new ways to reach you – all using that same five-digit code.

Even though there are different place associations that probably mean more to you as an individual, such as a neighborhood, street, or the block you live on, the zip code is, in many organizations, the geographic unit of choice. It is used to make major decisions for marketing, opening or closing stores, providing services, and making decisions that can have a massive financial impact.

The problem is that zip codes are not a good representation of real human behavior, and when used in data analysis, often mask real, underlying insights, and may ultimately lead to bad outcomes. To understand why this is, we first need to understand a little more about the zip code itself.

Looking to learn more about Spatial Data Science? Check out our range of Spatial Data Science events!

The Zip Code: A Brief History

The predecessor to the zip code was the postal zone  which represented a post office department for a specific city. For example:

Mr. John Smith
3256 Epiphenomenal Avenue
Minneapolis 16  Minnesota

“16” represents the postal zone in a Minneapolis. But with more and more mail being sent, in 1963 the Postal Service decided to roll out the Zone Improvement Plan  which transformed addresses to look like the following:

Mr. John Smith
3256 Epiphenomenal Avenue
Minneapolis  MN 55416

The five digit code represents a part of the country (5_ _ _ _ )  a sectional center facility ( _ 5 4 _ _ )  and the associate post office or delivery area (_ _ _ 1 6).

The first digit for every zip code for the states in the contiguous United States

By 1967 ZIP codes were made mandatory for bulk mailers and continued to be adopted by almost anyone sending mail in the US. Over time, the ZIP+4 was added to add more granularity to the zip code to denote specific locations  even buildings for postal workers to deliver. The Postal Service even created a character, Mr. Zip, to promote the use of ZIP codes, who was featured on stamps, commercials  and songs.

ZIP codes themselves do not actually represent an area, rather a collection of routes:

Despite the geographic derivation of most ZIP Codes, the codes themselves do not represent geographic regions; in general, they correspond to address groups or delivery routes. As a consequence  ZIP Code “areas” can overlap  be subsets of each other, or be artificial constructs with no geographic area (such as 095 for mail to the Navy  which is not geographically fixed). In similar fashion, in areas without regular postal routes (rural route areas) or no mail delivery (undeveloped areas)  ZIP Codes are not assigned or are based on sparse delivery routes  and hence the boundary between ZIP Code areas is undefined.

The US Census provides data for ZIP Code Tabulation Areas, or geographic files:

ZIP Code Tabulation Areas (ZCTAs) are generalized areal representations of United States Postal Service (USPS) ZIP Code service areas.The USPS ZIP Codes identify the individual post office or metropolitan area delivery station associated with mailing addresses. USPS ZIP Codes are not areal features but a collection of mail delivery routes.

Here we find our first problem with ZIP Codes, that they do not represent an actual area on a map, but rather a collection of routes that help postal workers effectively deliver mail. They aren’t designed to measure sociodemographic trends  as a business would generally want to do. You can actually look up individual delivery routes  like the one below:

One zip code route in New York’s East Village

We are only scratching the surface of the issue here. Similar issues exist around the world, with postal codes representing strange boundaries

Postal Codes in London, Toronto, and Sydney.

Using ZIP codes for data analysis

Fast forward to today, where many companies can easily look into their database and find a dataset with a zip_code column in it, which allows them to group and aggregate data to see trends and business performance metrics. As stated earlier, the problem with ZIP Codes is that:

1. They don’t represent real boundaries  but rather routes
2. They don’t represent how humans behave

The later represents two specific issues in using spatial data: spatial scale of observations and spatial scale support (you can learn more about this in this lecture from UChicago’s Luc Anselin, here). The first is that humans don’t behave based on administrative units such as zip codes, or even census units. Their behavior is influenced much more by their neighbors, or areas such as a neighborhoods or high activity areas (such as central business districts). The second is that spatial data is provided a