Why the Same Data Can Produce Several Defensible Maps

Give the same dataset to three competent cartographers and you should not expect three identical maps.

That can feel uncomfortable because we often want a map to behave like a calculation: identical input, identical correct output. But a map is not simply a table rendered geographically; someone has to decide what question matters, which geographic unit carries the comparison, which values deserve emphasis, how much context belongs on the page and which distinctions the audience needs to see.

More than one answer can be responsible.

The important boundary is not between one correct map and all other maps. It is between choices that can be defended against the purpose and evidence, and choices that cannot.

The dataset does not decide the question

Take a table of health facilities with coordinates and service attributes.

One map can show the location of facilities, another can show facilities per 100,000 people by district, and a third can show travel-time catchments. Other maps might show referral flows or highlight facilities offering a particular service.

These maps are not competing visualisations of one fixed question. They are different questions posed to overlapping evidence.

The data constrains what can be asked, but it does not choose what should be asked.

Even a much narrower dataset still leaves interpretive choices. A district-level unemployment rate might be mapped with five quantiles, five equal intervals, natural breaks or an unclassed continuous scale. Each representation can be calculated correctly. Each can make different geographic contrasts more visible.

That is not a flaw in cartography. It is the reason the method has to be disclosed and judged.

Geographic units create different stories

A map can change even when the statistic does not.

Aggregate the same phenomenon by municipality, district and region and the visible pattern can shift dramatically: small clusters disappear into larger areas, extremes are averaged away, and boundaries can divide one continuous phenomenon into separate administrative units or combine unlike communities into one value.

This is one reason “the map says” is often too strong a phrase. The map says something after the data has been assigned to a particular spatial framework.

A defensible choice of geography is therefore one that relates to the question. Administrative units may be appropriate when decisions are made through those units. A regular grid may be better when administrative area size would dominate perception. Points may be preferable when aggregation would imply a precision or uniformity that the underlying observations do not support.

The same data can support several of these choices without making them interchangeable.

Classification is an interpretation of difference

Classification makes the plurality especially visible.

Quantiles give each class a similar number of features and can make relative rank easy to see, while equal intervals preserve equal numeric ranges and can expose how much of the distribution sits in a narrow part of the scale. Natural breaks, meanwhile, adapt classes to the observed distribution and can make clusters within the data more legible.

None of these methods is inherently the truthful one.

The question is what relationship the classes are supposed to help the audience perceive. A classification becomes misleading when it is chosen because it produces a desired political or visual pattern rather than because it clarifies a relevant structure in the data.

The same principle applies to the number of classes, midpoint of a diverging scale and treatment of outliers. The map author is not merely formatting the values. They are deciding which differences become visible as categories.

Defensible does not mean arbitrary

Acknowledging legitimate choice does not mean “you can make the map say anything”.

A defensible map should survive at least three kinds of challenge.

First, the method should match the claim. If the map is about per-capita risk, raw counts are not an innocent alternative. If the phenomenon is sequential, a diverging palette with an arbitrary midpoint introduces a meaning that the data does not contain.

Second, reasonable alternatives should be considered. If a small change in class breaks completely reverses which places appear exceptional, that sensitivity is itself part of the story.

Third, important choices should be explainable. The author should be able to say why this geography, this denominator, this classification and this visual hierarchy were selected.

Defensibility is therefore not subjective preference. It is accountable judgement.

Different audiences can justify different maps

A technical analyst and a public audience may need different views of the same result.

The analyst may need uncertainty intervals, detailed categories and methodological context, while a public-information map may need fewer classes, clearer labels and a narrower set of visible fields. A decision-maker, meanwhile, may need a map focused on the areas where intervention is possible rather than a complete inventory of every variable.

Simplification is not automatically distortion. It becomes distortion when the simplification changes the claim without acknowledging the change.

This is why audience belongs inside cartographic judgement rather than after it. The question is not only “Which map is most accurate?” but “Accurate for what decision, at what level of detail, for which reader?”

Comparing alternatives is a form of quality assurance

One of the best ways to review a map is to make another plausible version.

Try a different classification. Swap counts for rates where appropriate. Change the geographic unit. Remove the basemap. Reduce the number of labels. Test a static version against the interactive one.

The purpose is not to find the single version that survives every alternative, but to discover which conclusions are stable and which depend strongly on one design choice.

If the same broad pattern remains visible across reasonable alternatives, confidence in the interpretation increases. If the story changes radically, the final map should not conceal that fragility.

A strong cartographic workflow therefore does not eliminate alternatives too early. It uses them to understand the space of honest representations before committing to the one best suited to the intended audience.

A map is a reasoned position, not a neutral printout

The fact that several maps can be defensible should make us more demanding, not less.

It means a map author cannot outsource responsibility to the dataset or the software defaults. Choosing quantiles because the GIS selected them automatically is still a choice. Publishing administrative totals because those were the fields already available is still a choice. Centering the map on one region and clipping another is still a choice.

The goal is not to remove judgement from mapping—that would remove much of what makes maps useful—but to make it visible enough that another person can understand why this map, among several possible honest maps, is the one being shown.