Maps are one of the most natural ways to visualise data. We understand them almost instinctively. A shape represents a place, colour or shading adds another dimension, and patterns emerge because the data remains connected to geography, but maps also contain a visual assumption that is easy to overlook: physical area determines visual area.
When the subject is rainfall, bushfire extent, vegetation, geology or land use, that is exactly what we want. The physical size and shape of the territory are part of the phenomenon being studied. A large area of burned forest should occupy more space on the map than a small one. For many other kinds of data, however, geographic area has little relationship to the thing we actually care about.
Population, economic activity, customers, votes and political representation all belong to places, but they are rarely distributed in proportion to land area. A large rural territory might dominate a map simply because it is geographically large, while a tiny metropolitan area containing vastly more people or activity may be barely visible. The map can be perfectly accurate geographically while giving us a distorted impression of the data it represents.
One way of dealing with this problem is to deliberately distort the map. That may sound counterintuitive, but it is the basic idea behind a cartogram.
Changing the shape of the map
A cartogram is a map in which the size or shape of geographic areas is altered to represent some quantity other than physical land area. Instead of asking, “How large is this territory?”, the map might ask, “How many people live here?”, “How much economic activity occurs here?” or “How much representation does this area have?”
A population cartogram, for example, might enlarge a densely populated city and shrink a vast but sparsely populated rural region. The result no longer looks geographically correct in the conventional sense, but its distortion is deliberate. The amount of visual space given to each place now reflects population rather than land. This changes the purpose of the map. Physical geography is no longer treated as an unquestioned visual truth; it becomes one of several things we can choose to preserve or sacrifice depending on what we want the reader to understand.
A cartogram deliberately sacrifices some geographic accuracy in order to increase the reader's understanding of the data.
There are several ways to create one. Some cartograms stretch and compress familiar geographic shapes while trying to preserve their general relationships. Others go further and replace irregular territories with standardised shapes.
A Dorling cartogram replaces each region with a circle. The circle's area represents the quantity, while its position preserves an approximate sense of location. The example below uses a small illustrative dataset so the relationship between quantity and area is easy to see.
One particularly useful form is the tilegram.
From cartograms to tilegrams
A tilegram represents geographic units using equal-sized tiles, commonly squares or hexagons. Each tile represents one consistent unit of analysis: one electorate, a fixed number of people, a sales territory, a service area or some other discrete quantity. Rather than stretching geographic regions into unfamiliar shapes, the tilegram makes the abstraction explicit. Every tile is the same size, so every unit receives the same visual weight.
Hexagons are especially useful because they provide six possible neighbours and allow regions to form relatively natural clusters while still maintaining a regular grid. Approximate geographic relationships can be preserved without pretending that the resulting shapes represent precise geography, although hopefully the tilegram will still be roughly perceived as the geography it is representing.
We can think of tilegrams not so much as accurate geography, but as a diagram of geography.
The United States Electoral College offers a concrete example. Each state receives one electoral vote for each of its senators and representatives, while Washington, DC receives three. A tilegram can give each electoral vote one hexagon, so the number of tiles represents electoral weight rather than land area.
Australia makes the problem obvious
When we began building a Federal Election Explorer, it quickly became apparent that simply shading a conventional map of Australia by the winning party did not provide a particularly clear picture of what was happening nationally. The problem becomes especially obvious around the major metropolitan centres, where many electorates are compressed into a relatively small area, while enormous rural electorates dominate the map visually despite each returning exactly the same number of representatives: one.
It is partly why election coverage so often relies on other views alongside the map: a representation of the parliamentary chamber, state-level summaries, or close-ups of individual electorates. The geographic map remains valuable for showing where results occurred, but it is much less effective at showing how those results combine to form the Parliament.
That led us to use a hexagonal tilegram. The transformation is deliberately simple: one electorate, one representative, one hexagon. Every seat is given equal visual weight, while its position still preserves an approximate sense of geography. The effect is quite different. New South Wales occupies more space because it contains more seats, not because of its land area. Victoria occupies less because it has fewer. Tasmania becomes a small cluster, while Western Australia no longer dominates simply because it covers a vast part of the continent.
The visual area now represents parliamentary representation rather than physical geography. That single change makes the national result far easier to perceive — but it only solves the first of the design problems.
The cartogram is the canvas, not the finished visualisation
It would have been straightforward to colour each hexagon by the party that won the electorate and stop there, and that would already have produced a useful map. But we wanted the visualisation to reveal something more about how elections actually change.
Our starting hypothesis was that, although government may switch from one side to the other every few election cycles, large parts of the electoral landscape remain relatively stable. The decisive movement tends to happen at the margins: a relatively small number of seats change hands, sometimes by narrow margins, and those changes can have a disproportionate effect on the final composition of Parliament. We therefore decided that change itself needed to become part of the visual language.
A party comfortably retaining a seat it was widely expected to win is not the same kind of event as a marginal electorate unexpectedly changing hands. Both contribute one seat to the parliamentary total, but they carry very different informational weight. That led us to introduce significance as a second visual dimension.
Each electorate still retains the colour of the party that won it, because the current result needs to remain immediately visible. But visual intensity is used to control emphasis. Predictable outcomes can recede into the background, while more consequential or surprising results draw the eye. This creates a useful distinction between the state of an electorate and the significance of what has just happened there.
That same principle applies well beyond election analysis. Analytical systems are often very good at showing what happened, but much less effective at helping us understand what actually deserves attention. Treating every event with equal visual emphasis can be almost as misleading as omitting information altogether.
Showing change without overloading colour
A change of political ownership creates another design problem.
Colour is already doing an important job: it tells us which party holds the seat. Asking it to also communicate whether the seat has changed hands would force two meanings into the same channel.
We solved this by marking "flipped seats" as outlined.
The fill continues to represent political ownership. The strength of the fill conveys significance. The outline identifies a transition from one party to another.
This becomes one of the recurring themes in the explorer. The design is not attempting to make one clever visual device explain everything. It is assigning different jobs to different visual properties.
Moving from seats to states
Individual electorates are only one level of the story. Australian elections are also interpreted regionally, and the state summaries provide an intermediate layer between the individual seat and the national result.
A small bar above each state shows the composition of its seats. The tilegram below preserves the individual results, while the bar makes the overall pattern immediately visible. This avoids forcing the reader to mentally count coloured hexagons to decide whether Queensland moved strongly towards one party or whether Western Australia remained relatively balanced. The useful part is that detail and aggregation remain in the same view. The individual observations have not disappeared simply because a summary is available.
The same pattern is then repeated nationally.
The hierarchy is quite deliberate: individual observations become regional aggregation, and regional aggregation becomes political consequence.
Numbers become meaningful when they have context
The national composition bar also contains one of the smallest but most important elements in the design: the majority marker. A party holding 94 seats is a number. Knowing that 76 seats are required for a majority changes what that number means. The reference line converts a count into an outcome. This is a general visualisation principle that extends well beyond election graphics. Revenue becomes more useful when shown against a target. Response times become more meaningful against an acceptable service level. A measured value becomes easier to interpret when the viewer can see the threshold, benchmark or expected range that gives it context.
Dashboards often devote a great deal of space to additional metrics while overlooking these simple reference points. In many cases, the benchmark is more informative than another number.
Geography still has a role
None of this means the conventional map should disappear. The geographic view continues to answer questions that the tilegram deliberately sacrifices. It shows the actual boundaries and locations of electorates, exposes regional clustering and makes neighbouring areas easier to understand. The two views are therefore complementary rather than competing.
The conventional map preserves physical geography. The tilegram preserves equal representation. One is better for understanding location. The other is better for understanding parliamentary structure. Providing both is more useful than trying to force either representation to answer every possible question. That is an important part of designing analytical interfaces more generally. Different views of the same data can be optimised for different tasks without making one of them the definitive view.
Designing the visual grammar
The finished election explorer looks fairly simple at first glance. There are coloured hexagons, some bars, labels and a timeline.
Underneath that simplicity, however, is a fairly disciplined visual grammar.
- Area represents parliamentary representation
- Position preserves approximate geography
- Colour represents political ownership
- Visual intensity represents significance
- An outline identifies a change of ownership
- The state bars provide regional aggregation
- The national bar shows parliamentary composition
- The reference marker shows the governing threshold; and
- The timeline introduces change through time.
The important point is that none of these choices is accidental, and none is there simply to make the visualisation look more interesting. Each visual property has a specific job, and each is chosen to convey meaning in a way that is both perceptually truthful and easy to understand.
That is the foundation of good visual design. It is not decoration applied after the analysis is complete. It is the process of deciding what deserves visual weight, how different kinds of information should be distinguished, and which visual encodings will let the viewer understand the result with as little effort and ambiguity as possible.
Good visualisation is often less about adding information than assigning each piece of information exactly one visual job.
The tilegram is the most distinctive element, but it is only one part of that system. Its purpose is to correct the imbalance created by geographic area so that representation is shown more truthfully. Everything else builds on that foundation.
The broader lesson is that effective visualisation comes from being deliberate about meaning. Sometimes preserving geography faithfully is the right choice. Sometimes simplifying or distorting it produces a more accurate understanding of the data. The design succeeds when the viewer can see the difference immediately, without having to work to decode what the visualisation is trying to say.