Location Analytics at Scale: Esri on GeoAI and Spark Spatial Tools
In a COLLIDE 2024 session, Esri's Brady Cline argued that location is an underused ingredient in enterprise analytics. He outlined business questions spatial data can answer and showed a recorded demo of GeoAnalytics Engine analyzing millions of customer records inside Spark.
By The AI Enterprise Magazine Team4 min read
Key takeaways
- Location is an analytic layer. Cline said many business problems with no map still have a spatial component.
- Scale needs spatial tools in the data stack. Platforms such as Databricks lack native spatial joins, he said.
- ZIP codes are blunt. Geoenrichment gives a finer view of customers than a ZIP code.
- GeoAI adds context. He said general chatbots lack geographic understanding.
- Demo at scale. A recorded demo aggregated 15 million records to 3,000-plus H3 bins in seconds on a laptop.
At COLLIDE 2024, the COLLIDE Data & AI Conference in Atlanta in October 2024, Brady Cline, Sales Manager of Enterprise Analytics & AI at Esri, argued that location is an underused ingredient in enterprise analytics. In a session sponsored by Esri, Cline said many business problems that never appear on a map still have a spatial component. He then showed a recorded demo of Esri's GeoAnalytics Engine, a Spark-native tool for running spatial analysis inside existing data pipelines.
What is location analytics and why does it matter for business?
Cline described Esri as a privately held location intelligence company and probably the largest geospatial intelligence company in the world. He said Esri has supported about 350,000 organizations over a career of almost 55 years. He said the top 10 insurers use Esri technology, and that 80% of the Fortune 100 use Esri tools in some way. These figures were his own claims in the session.
His core argument was that "everything that we touch and operate with happens in space and time." Mapping, he said, is only part of the picture. The bigger opportunity is applying spatial analysis to problems in supply chain, risk management, corporate security, finance, public health and logistics, even when no map is involved.
Which business questions can location data answer?
Cline listed questions that location tools address:
- Where else can I find a customer like this one?
- Where should I open my next storefront, and how does a physical store affect digital sales?
- Which property should I buy, and what surrounds the parcel?
- What is the climate risk of a landholding, searched across the globe rather than one area?
On the store question, Cline said that in several examples he had seen digital orders rise after a physical store opened. He attributed that to customers feeling connected to a store they can return to.
How do you handle location data at scale?
The technical challenge, Cline said, is making sense of "trillions of points" that also turn into polygons and more complex geometries. A logistics company moving packages overnight, for instance, has to weigh huge numbers of location factors at once.
He said many organizations lack data scientists with spatial tools in the form they are used to. The traditional workaround is to have a GIS professional pick a subset of the data to study. Cline contrasted that with throwing all of the data at the question, which he said gives a much richer view. He said platforms such as Databricks do not inherently include spatial joins, so Esri brings those tools to them.
He also explained geoenrichment. Geocoding gives an address a point, which supports entity resolution. A ZIP code is a poor stand-in for a customer's surroundings, he said, because a single ZIP code holds many kinds of housing. His example was a special tax zone defined as within three miles of a Metro rail line, which raises questions such as whether the distance is walking, driving or straight-line.
What is GeoAI?
Cline defined GeoAI as "any artificial intelligence that uses geography to answer the question." He said he asked ChatGPT about a point he had defined in Arlington, Virginia, and got demographic results that read like a Wikipedia entry. The point was actually the center of the Pentagon courtyard, he said, and the tool lacked that geographic context.
He added that visualization matters because executives make decisions from it without seeing the analytics behind it, so a visual has to point to action. He also stressed sharing and collaboration, so that data does not end up in new silos.
What did the GeoAnalytics Engine demo show?
Cline named three tools: GeoAnalytics Engine, decision support dashboards and Business Analyst, which includes Esri's Tapestry data. He then played a recorded demo by Dennis Gerard, who is introduced in the video but whose title is not given.
Gerard worked in a notebook with two sets of Parquet tables, one for customers and one for stores. The task was to enable geometry on both, measure the distance between each customer and the store where they transacted, and aggregate the results to H3 bins.
According to the demo, the code read more than 19 million customer records. The interactive run covered 15 million records, which it aggregated to more than 3,000 bins in a couple of seconds. Gerard said it ran locally on his machine, not on a large Databricks Spark cluster. He said the results could be exported as GeoJSON, stored in cloud systems, or pushed into ArcGIS for dashboards, and he showed a Power BI store assessment as well.
COLLIDE 2027 returns to Atlanta
COLLIDE 2027, the largest AI conference in the Southern U.S., will bring 1,000+ enterprise data and AI leaders to the Sandy Springs Performing Arts Center in Atlanta on Wednesday, October 13, 2027. Details are on the COLLIDE event page, and registration is coming soon.
Frequently asked questions
What is location analytics?
Applying spatial analysis to business data, so questions about customers, stores, risk and logistics account for where things are.
What is GeoAI?
Cline defined it as any artificial intelligence that uses geography to answer the question.
What is Esri GeoAnalytics Engine?
A Spark-native tool, described in the session, for running spatial functions inside existing Spark data frames.
What is geoenrichment?
Adding location context to records, starting with geocoding an address to a point, to resolve entities and enrich customer data.
This article is based on remarks by Brady Cline, Sales Manager, Enterprise Analytics & AI, Esri at COLLIDE Data + AI Conference 2024 on October 10, 2024. More coverage: COLLIDE AI Conference.
Filed under: Data & AI Infrastructure, AI Use Cases & Applications
