Geospatial Data Science

Thanks to inexpensive storage and the democratization of platforms and tools that can process raw data at scale, companies now have more raw material (data) than ever. The scarce resource is no longer data, but the experience to weave the data into a tapestry of intelligence.

With a constantly growing mountain of data, companies lack the ability to aggregate, de-duplicate, normalize and refine their data into something actionable. We all look forward to the promise of Machine Learning and Artificial Intelligence as elements of the solution. However, more technology processing more data is already proving to fall short of the mark because, as any data science expert knows, if the data is lacking, the results will inconclusive at best and misleading at worst. Misleading results lead to poor conclusions which are often expensive to reverse.

The G73 Difference

G73, formerly Gadberry Group, has been a LI data provider for twenty years. We’ve honed skills and insights by working closely with our clients. We are expert practitioners of the leading GIS technologies, understanding their strengths, weaknesses, and bias.  As a long-time data product company, we have extensive relationships with leading data providers, and understand how their build philosophies affect their data products. Building this expertise took decades of careful investigation and testing. We realized that the learning curve for consumers of these tools and datasets was extremely high.  

Our philosophy is to help clients integrate data into, and unlock value from, their models, machine learning pipelines, visualization tools or executive summaries. We leverage our expertise to build our Data Products to be easily assimilated at the scale of enterprise while also being accessible and useful by the individual practitioner.


Addressing seems simple in concept, it is convoluted in practice. Some localities have radically different addressing schemes, often left up to the historical assignment of address formats in the area and the preferences of the local Postmaster.  Further, addresses appear to be permanent but often are not.

Our address Data Products are Change Aware. Each vintage is built from scratch, only leveraging the previous vintage for corroboration. This laborious build process allows us to maximize inclusion of added, removed or changed addresses while avoiding the inclusion of invalid records into the address universe. This continuous improvement cycle is a key differentiator between our address universe and similar files, which are often updated infrequently, or stale address records are never removed.

G73's "Address Confidence Score" approach provides a simple-to-use guideline to allow the data consumer the option to include or exclude addresses based on their use case. G73 utilizes a rigorous corroboration process, leveraging a multitude of data sets to generate the Address Confidence Score. One consumer might want knowledge of whether an address has ever existed and therefore accepts all Address Confidence Scores. Another might want only the most pristine, highly-corroborated Scores.

Slide the bar in the image to see a "before and after" scenario. 


Google Maps has created the perception that finding the location of an address is a readily-available, free service.  There are dozens of geolocation providers who offer free, online services to return a latitude and longitude for an address. When all you want is an answer in the same vicinity of the address, these services are useful. If, however, you need to geocode addresses at scale or you need high-precision, these services are either inadequate or unaffordable.

Our extensive experience in the art of geocoding allows us to leverage a host of tools and datasets to select the best-in-breed geolocation for addresses. In most cases our geolocations are created at the building rooftop level. We have the highest percentage of rooftop geolocations for US addresses of any of the commercial geolocation providers.

As with our "Address Confidence Score" G73 also provides a similar instrument for geolocations. Our "Geo-Precision Score" provides insight into the methodology used to generate a geocode for an address. Data consumers can then use that score to select or exclude records based on their use case and requirements.


Location Intelligence doesn't end with knowing where a thing is on Earth. A location without context is only interesting, it does not drive intelligence. Our extensive collection of boundaries provide the context necessary to correlate various points into meaningful collections.

Using G73's API platform, clients can provide an address, geolocation or area, request one or more boundary types in order to receive all the intersecting boundaries, along with the attributes for those boundaries.

Use cases include identifying the school district or neighborhood for an address, selecting all the parcels in a minimum bounding rectangle, determining what telecommunications Wire Center serves a newly constructed home or determining which Census block contains one or more locations.

Boundary selection allows for logical grouping of locations. These groupings give insight into density as well to describe locations using the attributes associated with their containing boundaries.


Identifying areas with the most desirable personas is one of the key outcomes of commercial geospatial analysis. Variables such as population, daytime population, household size, gender, income level and education level provide details that allow businesses to model those cohorts. Identifying population change in demographic tendencies over time are examples of how the application of Demographics can create a competitive advantage.

G73 demographics include traditional Census based products and MicroBuild®, built using individual and household sources.  Data packages range from a handful of the most commonly used variables, provided at micro-geographic levels, to bundles which include thousands of variables.


The insights from individual demographic variables can be powerful, but can also be difficult to manage. Often companies have deep understanding of the general demographics and tendencies of their valued cohorts but lack the time or inclination to map those to dozens of variable ranges.

G73 segmentation provides easy to consume and highly-actionable demographics by clusters, or "segments". Each segment definition comes with psychographic profiles of generalizations of age ranges, education, income, family units, spending patterns and other consumer propensities. These precomputed segments are assigned to geography, beginning at Zip+4 and Census block, providing the highest spatial resolution possible.

With G73 segmentation, identifying the most micro-areas with the most desirable cohorts is simple and straightforward. Segmentation can enrich cohort analysis without the steep learning curve of modeling and understanding dozens of demographic variables.