Why Are My Distance Results Clearly Wrong?

When GIS distance results are obviously implausible, start with three questions: what units are the coordinates in, which distance model is being used, and what geometric objects are being measured? Those checks explain most large discrepancies more reliably than changing the calculation at random.

Check the units before the algorithm

If the coordinates are longitude and latitude in degrees, an ordinary Cartesian distance between coordinate pairs is measured in degrees, not metres or kilometres.

A result such as 0.25 may therefore mean a quarter of a degree rather than 250 metres. The ground distance represented by a degree of longitude also changes with latitude, so multiplying by one global constant is not a robust correction.

If the data are projected, inspect the CRS unit. Many systems use metres, but some use feet or another linear unit.

A constant factor error across many features often points to units. If every result is roughly 1,000 times too large or too small, check metre/kilometre conversion. If the factor resembles 3.28084, inspect metres versus feet.

Distinguish planar from geodesic distance

A planar distance assumes the coordinate space is flat, which can be appropriate inside a projected CRS designed for the study area. A geodesic distance, by contrast, follows the reference ellipsoid and is usually more appropriate for long distances or geographic coordinates.

The difference may be small for short local measurements and important across countries or continents. If the error grows with extent, your distance model may not fit the scale of the problem.

Euclidean Distance vs Geodesic Distance explains the distinction in detail.

Make sure you are measuring the intended geometry

“Distance between two features” can mean several things.

For polygons, software may return the shortest edge-to-edge distance, which is zero when two polygons touch or overlap even if their centroids are kilometres apart.

For routes, a line's length is different from straight-line distance between its endpoints. For multipart features, the nearest pair of components may determine the result.

Before judging the number, state the intended geometry:

point to point;
centroid to centroid;
boundary to boundary;
point to polygon;
route length;
nearest component of multipart features.

A correct edge-to-edge result is wrong only if you intended a different geometric question.

Do not confuse distance with travel

The nearest place geometrically is not necessarily the easiest or fastest place to reach.

If you expect road distance or travel time but calculate Euclidean or geodesic distance, the result can look “too short” because it ignores the network, where rivers, road topology, direction restrictions, bridges, slopes and transport modes all change usable travel.

For that distinction, see Straight-Line Distance vs Network Distance.

Look at the pattern of the error

Pattern · More likely cause

Values are tiny decimals — Degrees interpreted as linear units
Nearly constant conversion factor — Unit mismatch
Error increases with long distances — Planar method over large extent
Overlapping polygons return zero — Edge-to-edge geometry semantics
Route distances are much shorter than expected — Straight-line method used instead of network distance
Only a few records are extreme — Bad coordinates or outliers
Distances cross the globe unexpectedly — Axis reversal or wrong CRS assignment

The table is a diagnostic aid. Verify with known pairs before applying a correction to the full dataset.

Check representative coordinates directly

Choose two features whose approximate distance you know independently. Inspect their coordinates and map locations.

If they plot correctly and the calculation is wrong, focus on units and distance method; if they plot incorrectly, solve the coordinate problem first. A sophisticated geodesic calculation cannot compensate for incorrect source coordinates.

Similarly, if one outlier coordinate is thousands of kilometres away, the distance calculation is accurately exposing a data-quality problem.

Verify the final method

A defensible distance workflow records:

source CRS;
target or analytical CRS if projected;
distance model—planar, geodesic or network;
geometry definition used for multipart or polygon features;
output unit;
representative validation examples.

For network analysis, also record the network source, travel mode and cost field because “distance” may refer to route length while the decision is actually about time or accessibility.

Diagnose the coordinate model before changing the formula

Clearly wrong distance results are rarely solved by trying more functions until one returns a plausible number, because a plausible number can still be methodologically wrong.

Instead, identify units → coordinate model → geometry → movement model. Once those agree with the question, the numerical result becomes interpretable rather than merely believable.