A Map Can Be Technically Correct and Still Be Misleading

Some mapping errors are satisfying to diagnose: the coordinate reference system is wrong, a join dropped half the rows, distances were calculated in degrees or the geometry is invalid. Fix the technical mistake and the map improves.

The more difficult maps are the ones where nothing is obviously broken.

Every feature is in the right place, the projection is defensible, the data values match the table and the legend is accurate. The map renders exactly as specified—and the audience is still being led towards a conclusion the evidence does not support.

That is not a software failure. It is a failure of the claim.

Correct execution can answer the wrong question

Imagine a map of health facilities where every facility has been geocoded correctly, the point symbols are legible and the basemap is appropriate. The title says Access to Healthcare.

But the map only shows facility locations.

It says nothing about capacity, opening hours, affordability, road conditions, staffing, service type or whether people can actually reach and use those facilities. The map answers a narrower question—where are the recorded facilities?—while the title invites the reader to infer a broader one—who has access to care?

Nothing in the rendering pipeline can repair that mismatch. A technically flawless point map remains an incomplete answer to an accessibility question.

This pattern appears everywhere. Raw population counts are mapped as though they showed individual risk. Administrative totals are compared without accounting for different population sizes. A current basemap sits beneath decade-old thematic data and makes the whole composition feel current. A set of reported incidents is presented as though reporting coverage were uniform.

The operations can all be correct. The map can still overstate what has been established.

Cartography participates in the argument

It is tempting to think of analytical choices as substantive and visual choices as presentation. Maps make that separation difficult.

Class breaks decide which places appear exceptional. A diverging colour scheme implies that its midpoint matters. Symbol size controls which locations dominate the page. The initial map extent determines what the reader encounters before they make any deliberate choice. Labels tell the audience which places deserve names. A detailed, familiar basemap can make a weak thematic layer feel grounded in authority.

None of these choices changes the underlying values, yet all of them can change the interpretation.

This does not mean every design choice is manipulative. It means design is part of the reasoning the reader receives. The map's argument is not only in the data table; it is in the relationship between data, transformation and emphasis.

Omission is often more consequential than error

A map necessarily leaves things out. The ethical question is not whether it simplifies, but whether the simplification hides something that materially changes the claim.

Suppose districts with no observations are left unfilled. If the legend does not distinguish zero observations from no data, many readers will interpret the blank areas as absence. An uncertain location shown as an exact point acquires a precision the evidence never had. A derived indicator published without its denominator can look like a direct observation. A boundary shown without a reference date can appear timeless.

These are not necessarily false statements; they are missing distinctions.

The danger is that maps are exceptionally good at making distinctions disappear. Once several analytical decisions have been compressed into colour and shape, the reader may see only the apparent pattern.

Technical QA validates only part of the chain

A conventional QA pass might ask:

Did the join use the intended key?
Did every feature render?
Are the class breaks correct?
Are units consistent?
Does the popup show the value stored in the dataset?

Those checks are essential, but they validate the implementation of a specification, not the validity of the specification itself.

A second review has to ask different questions:

What claim does this map invite?
Does the dataset actually support that claim?
What alternative explanation remains plausible?
Which visual choice most strongly shapes the conclusion?
What important limitation has become invisible?

This is harder because there is no universal validator for it. The answer depends on purpose, audience and domain context.

Write the claim before finalising the map

One useful discipline is to write the intended claim in a single sentence before the map is finished.

Not the title, but the claim.

For example:

Reported road incidents were more concentrated in these districts during the study period.

That sentence immediately forces questions that a vague title such as Road Safety by District can avoid. Are the values counts or rates? Does reporting vary? Are the districts comparable? Which time period? Does “concentrated” refer to totals, density or risk?

Now examine each major cartographic choice against the claim. The denominator, geography, classification, colours, labels, basemap and annotations should either help establish the claim or help qualify it. If a choice makes the claim feel stronger than the evidence allows, it deserves revision.

This is not a demand to turn every public map into a methodology paper. A concise source note, date, legend and limitation can carry a great deal. The point is that the qualification must survive the visual compression.

Trustworthy maps have more than one kind of correctness

Technical correctness is non-negotiable because an invalid geometry or wrong CRS can destroy an analysis before interpretation even begins. But a trustworthy map also needs analytical appropriateness and communicative honesty.

Analytical appropriateness asks whether the chosen data and transformation answer the intended question. Communicative honesty asks whether the design lets the audience see the strength and limits of that answer rather than only its most persuasive form.

Software can help with both. It can surface metadata, preserve provenance, make alternative classifications easier to test and keep limitations close to the published result. It cannot decide, on its own, what the audience is entitled to believe.

A map is not trustworthy because every instruction was executed correctly. It is trustworthy when the instructions themselves were worthy of execution.