10 Hard Problems with Geospatial Data

Location intelligence promises that layering business data over a map reveals patterns, optimizes routes, and prices risk early. Practitioners know the hard truth, which is that geospatial data is messy, heavy, inaccurate, politically charged, and difficult to manage at scale.

Vendors gloss over the friction. Here are ten problems that recur in every serious geospatial program, each documented by a named executive or pioneer who confronted it in production.

1. The Maintenance Trap

Elon Musk dismissed reliance on High-Definition (HD) maps and Light Detection and Ranging (LiDAR) at Tesla Autonomy Day in 2019, arguing that “the maps are really brittle. The world changes, and if the map doesn’t update, the car can’t drive.” This is not specific to self-driving cars, because every map is a snapshot of the past from the moment of capture. Satellite revisit runs days to weeks, street-level data is months old when served, and centimeter-accurate HD maps for full autonomy demand a permanently funded capture—then-update maintenance loop. Stale data produces failures that look like human error, such as trucks routed down closed roads and flood risk priced from an old map. Crowdsourced updates concentrate where users are affluent, and models trained on that skewed coverage inherit bias.

2. The Accuracy Illusion

Brian McClendon, who led engineering for Google Maps and Google Earth, stated that “map data is deeply inaccurate” and that mapping the world accurately is very hard. Precision and accuracy differ. A coordinate with 12 decimal places can still sit 50 meters from the truth. Address geocoders interpolate, guessing that address 150 falls midway between 100 and 200 when it may hide behind address 105. Global Positioning System (GPS) receivers drift with atmospheric conditions, multipath reflections off buildings, receiver quality, and adversary interference. False confidence in the pin produces physical failures, from riders canceling when the driver stops on the wrong side of a complex to crews trenching into a gas line.

3. The Privacy Minefield

In 2018, the fitness application Strava released a global heatmap of user activity, and analyst Nathan Ruser observed that United States military bases were “clearly identifiable and mappable,” traced by soldiers jogging base perimeters in Afghanistan and Syria. Location data resists anonymization because it carries spatial uniqueness. A device that spends nights at Location A and days at Location B identifies its owner with trivial external data, and aggregation reveals sensitive places even when individuals stay hidden. The General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA) increasingly treat precise location history as sensitive personal data. A breach leaks the physical movements of customers, which makes casually collected tracking a toxic asset.

4. The Interoperability Headache

Dan Vesset, group vice president at the International Data Corporation (IDC), estimates that people spend 60 to 80 percent of their time trying to find data, and the silo problem is acute in geospatial work. Engineering holds Computer-Aided Design (CAD) files, web teams hold GeoJSON, analysts hold shapefiles, and database teams hold non-spatial tables. The shapefile, introduced in the early 1990s, remains the de facto exchange standard despite a 2 gigabyte size limit, 10 character field names, and a dependency on 3 to 7 companion files. Conversions between mathematical CAD geometry and point-based Geographic Information System (GIS) polygons lose information, and schema mismatches force computationally expensive spatial joins. Modern formats such as GeoParquet and FlatGeobuf fight inertia of millions of legacy datasets, and teams spend months on pipelines while the business opportunity expires.

5. Semantic Ambiguity

Tim Cook apologized in a 2012 open letter, writing that Apple “fell short on this commitment,” after Apple Maps launched with melted bridges and stores placed in rivers. The failures were semantic, since the system never understood that a store does not belong in a river. The Empire State Building could resolve to its front door, its parcel centroid, or its loading dock, and address standardization struggles with abbreviation variants and rural directions. The Modifiable Areal Unit Problem (MAUP) compounds the ambiguity, because a neighborhood looks wealthy by zip code and poor by census tract, and human judgment must decide which boundary reflects reality. A routing engine that delivers to an industrial park’s centroid costs the driver 15 minutes per stop, multiplied across a fleet.

6. The Coordinate Nightmare

The Earth is an oblate spheroid, and thousands of projections flatten it in different ways. Web maps use Web Mercator, GPS reports positions in the World Geodetic System 1984 (WGS84), and local governments use State Plane zones, each identified by a code in the European Petroleum Survey Group (EPSG) registry. Overlaying datasets without reprojecting shifts data by feet or drops it at coordinates (0, 0) in the Atlantic Ocean, the Null Island effect. Tectonic plates also move, so a coordinate measured in Australia in 1994 no longer marks the same physical spot, and modern datums track the epoch of measurement. The vertical axis is worse, since definitions of zero elevation disagree. A drone commanded to 100 meters against the wrong vertical datum can fly into a hillside, and 6 inches of elevation error flips a property between flood risk classes.

7. Volume & Velocity Overload

John Hanke, chief executive of Niantic, has said Niantic deliberately schedules smaller daily player counts for Pokémon GO because cellular networks struggle when every player transmits location data at once. Spatial queries are computationally expensive, since finding all points within 500 meters of a moving coordinate requires structures such as R-trees and quadtrees, and fleet-scale tracking needs specialized spatial databases and graphics processing unit (GPU) acceleration. Storage for imagery and position logs drains budgets, and a four hour processing delay makes real-time dispatch impossible.

8. Dependence on a Single Point of Failure

Most location businesses stand on the Global Navigation Satellite System (GNSS), such as GPS, a government-operated utility that delivers a weak signal from space, easily blocked in urban canyons, jammed by cheap devices, and spoofed by adversaries. Multipath reflections in downtown Manhattan convince a phone it sits a block from its true position. Platform dependence adds a second fragility. At Uber, McClendon championed deeper mapping investment, and the company spent roughly 500 million dollars building its own maps to reduce reliance on Google, a supplier that could throttle access, raise prices, or compete. One jammer in a delivery truck disrupts a tracking system, and a platform’s interface pricing dictates a dependent company’s margins.

9. Indoor Blindness

People spend most of their time indoors, where satellite positioning stops at the roofline. McClendon, discussing complex transit hubs such as Shinjuku Station, noted that a map attempting to answer every question for every person becomes effectively unreadable. Indoor positioning requires different technology, such as Wi-Fi fingerprinting, Bluetooth beacons, or Visual Positioning Systems (VPS) that recognize surroundings through a camera. Interiors also change constantly, walls move, and modeling floors, elevators, and stairwells in three dimensions costs far more than mapping outdoor space in two. Warehouses lose throughput when they cannot route forklifts, and hospitals lose critical minutes searching for mobile equipment.

10. The Skills Gap

Roger Tomlinson, the acknowledged father of GIS, recalled that “the early days of GIS were very lonely,” and the loneliness persists as a talent shortage. Traditional GIS analysts know projections, topology, and cartography, while data scientists know code and cloud architecture, and few professionals hold both. A data scientist who treats latitude and longitude as two ordinary numeric columns will misuse nearest-neighbor algorithms, because degrees of longitude shrink away from the equator and angles resist simple averaging. Companies that staff only one side of the divide get beautiful maps that cannot integrate with the cloud stack, or scalable pipelines that produce spatially wrong answers.

How to Manage the 10 Problems?

These ten problems are reasons to respect geospatial data, the only data type that ties digital records to physical reality. The pioneers quoted here responded by investing more, and the durable practices are consistent:

• Treat maps as living documents with funded update loops.
• Invest in spatial data engineering as heavily as in the data itself.
• Build systems that tolerate fuzzy locations, and design privacy in from the start.
• Diversify map providers behind an abstraction layer, and key records to persistent identifiers or grids such as Uber’s H3 and Google’s S2 instead of addresses.
• Adopt cloud-native formats, such as the Cloud Optimized GeoTIFF (COG) for imagery and GeoParquet for vectors, and hire for spatial indexing skill.
• Pilot positioning that fuses GNSS with inertial and visual sensors ahead of a contested-signal future.