Visualizing Geographic Data: Maps That Tell the Story
most business maps lie. not on purpose, but because the defaults are wrong. the most common map in solopreneur dashboards is a US state choropleth shaded by total revenue. California is dark red. Wyoming is white. the reader concludes California is the best market. except California has 39 million people and Wyoming has 580,000. the chart is showing population, not revenue performance. that is the geographic visualization trap and almost everyone falls into it.
geographic data adds a real spatial dimension that other charts cannot match. customers cluster in cities. shipping costs follow distance. ad performance varies by region. when you have geographic data, a map almost always tells the story better than a table. but the same defaults that make line charts forgiving make maps treacherous because area, distance, and color all carry meaning the moment you draw them.
this guide is for solopreneurs and small-team analysts who need to map customer locations, sales by region, or operational coverage. by the end you will have a map-type decision table, the per-capita rule that prevents most lying maps, the right way to handle clusters and outliers, and a list of tools that produce publication-quality geographic charts in 2026.
what makes geographic data different
three things. the data is spatial, which means location carries information beyond the value. areas have different sizes, which means a choropleth map naturally over-emphasizes large regions. and projection choices distort the underlying geometry, which means the same data can look different depending on the map you draw it on.
unlike a bar chart where the visual encoding (length) is honest, a map adds noise from area, projection, and aggregation level. the analyst’s job is to manage these so the data is the signal, not the canvas.
Geographic data visualization in 2026 should default to per-capita or per-customer normalization for choropleth maps, point or bubble maps for raw counts, flow maps for movement, and small multiples for time-based comparison. Avoid the Mercator projection for global data because it inflates high-latitude countries. Always include a legend with units (per 1,000 people, per dollar, per customer). The most common geographic visualization mistake is shading a choropleth by total volume when the underlying regions vary in population by an order of magnitude, which produces a map that shows population, not the metric.
normalize before you draw, choose the projection that matches your purpose, and label units in plain language. those three habits prevent most geographic visualization mistakes.
the map type decision table
different geographic questions need different map types. picking the wrong type is the most common geographic mistake.
| data shape | best map type | when to avoid |
|---|---|---|
| value per region (state, country) | choropleth, normalized per-capita | with raw totals (population dominates) |
| individual events (orders, locations) | point map | with overlap (clusters become solid mass) |
| individual events with intensity | heatmap | with sparse data (creates fake density) |
| value per region with magnitude | proportional bubble map | with overlapping bubbles in dense areas |
| origin to destination | flow map (arrows or lines) | with too many flows (becomes spaghetti) |
| spatial coverage (service area) | filled polygon overlay | when boundaries are uncertain |
| comparing same metric across regions | small multiple choropleth | with one big map and small text |
| change over time | animated map or time-slice small multiples | with single static frame (loses motion) |
| local detail at zoom | zoomable interactive map | when print or static is required |
for solopreneur dashboards the three workhorses are: per-capita choropleth for performance comparison, point map for individual customer or order data, and bubble map when you need to show both location and magnitude.
the per-capita rule that prevents most lies
if you are shading regions by an absolute number, you are almost always showing population. revenue by state, orders by country, customers by city. all of them correlate with population first.
the fix is normalization. divide by the relevant denominator before you map.
| metric | divide by | why |
|---|---|---|
| total revenue | population (per 1,000) | shows market penetration, not size |
| total customers | population (per 1,000) | shows reach, not raw count |
| ad spend | revenue | shows ROAS by region |
| support tickets | active customers | shows quality issue, not popularity |
| visitor count | unique residents | shows reach efficiency |
the moment you normalize, the map shows performance instead of population. that is almost always the question you actually wanted to answer.
a useful sibling read is visualizing comparisons charts which covers cross-category comparison without the geographic complication.
projection traps and how to avoid them
map projections are the way 3D earth is flattened to 2D. every projection distorts something. the question is what you are willing to distort.
three common projections cover most use cases.
Mercator preserves angles and shapes locally, which is why it became the web map default. it badly inflates high-latitude regions. Greenland looks the size of Africa on Mercator. Africa is 14 times larger. for global data, Mercator is misleading.
Equal-area projections (Mollweide, Eckert IV) preserve area, which is what you want for global per-capita choropleth maps. countries near the equator and near the poles read the same.
Albers equal-area is the right choice for US-only choropleths. it preserves area within the continental US and produces the familiar shape.
if your dashboard is for a US audience and the data is US-only, use Albers. if your data is global and you are doing per-region comparisons, use an equal-area projection. if you are showing a navigable web map (Google Maps style) for individual locations, Mercator is acceptable because the user zooms.
most BI tools default to Mercator. switching to equal-area is usually one setting away. it is worth the 10 seconds.
handling clusters, outliers, and sparse data
three patterns cause map problems for solopreneurs.
cluster overlap
when you plot 500 customer addresses in a metro area on a point map, the points overlap into a dark mass. the map shows there are customers there but not how many. two fixes work.
cluster the points programmatically. most tools (Mapbox, Leaflet, Looker Studio) support marker clustering that aggregates nearby points into a numbered circle. the user zooms in to see individual points.
switch to a heatmap or bubble map at the metro-area level. the question changes from “where exactly” to “how dense”, which is usually the better question for a dashboard.
outliers that distort the legend
one customer in a remote region with 100x normal order volume will dominate the color scale. every other region looks the same shade because the legend is stretched. fix this by clipping the legend at the 95th or 99th percentile and labeling the outlier separately.
sparse data that looks like signal
if you are mapping a metric across 50 US states but only 8 states have customers, the empty 42 states still get colored white in most defaults. white reads as “low.” but no data and “low” are different. set null regions to a distinct hatch pattern or gray, with a legend entry that says “no data.”
for the underlying habit of distinguishing missing data from low values, see avoiding misleading charts which covers this pattern across chart types.
maps in dashboards vs presentations
a dashboard map can be interactive. a slide map cannot. design for the format.
the dashboard map can include zoom, pan, hover tooltips, layer toggles, and time sliders. the user has time to explore.
the slide map must read in three seconds. one map type, one normalized metric, one comparison. annotate the two or three regions that matter and gray out the rest. the data presentation for executives guide covers slide-specific design in detail.
the geographic dashboard checklist
before shipping a map to a dashboard, run this checklist.
- the metric is normalized by an appropriate denominator (per-capita, per-customer, per-dollar)
- the projection is appropriate (Albers for US, equal-area for global, Mercator only for navigable maps)
- the legend has explicit units (orders per 1,000 residents, not just orders)
- null regions are visually distinct from low-value regions
- outliers are clipped or labeled separately
- color scale is colorblind-friendly (avoid red-green; use ColorBrewer presets)
- annotation labels the 2-3 regions that matter for the dashboard’s question
- the chart title is the conclusion, not the description
a map that passes this checklist usually reads correctly without explanation. that is the bar.
tools for geographic visualization in 2026
geographic visualization tools have matured. five options cover the solopreneur use case.
| tool | best for | cost |
|---|---|---|
| Google Looker Studio | dashboards with simple choropleth and point maps | free |
| Tableau Public | publication-quality maps, advanced symbolization | free |
| Datawrapper | static maps for blog posts and reports | free up to limited features |
| Mapbox Studio | custom-styled interactive web maps | free up to 50k loads/month |
| Kepler.gl | quick exploration of large geo datasets | free; open source |
| QGIS | professional GIS analysis, custom projections | free; open source |
the recommendation for most solopreneurs is Looker Studio for dashboard maps and Datawrapper for the static map you publish in a blog post or send in a report. Tableau Public is worth learning if you regularly publish geographic insights externally because the styling options are far ahead of the alternatives.
for the spreadsheet side of geographic data prep, see Google Sheets QUERY function which covers aggregating raw transaction data into the per-region totals a map needs as input.
common geographic visualization mistakes
four mistakes show up in 80% of solopreneur maps.
absolute volume choropleths. the most common mistake. shows population, not performance. always normalize.
rainbow color scales. rainbow palettes (red to violet) imply ordinal hierarchy that is not in the data and read poorly when printed in grayscale. use sequential single-hue palettes (light blue to dark blue) for ordinal data and diverging palettes (blue-white-red) only for data with a meaningful midpoint.
unlabeled units. “0 to 100” on a legend without units forces the reader to guess. always include the unit. “0 to 100 orders per 1,000 residents” reads honestly.
incomplete legends. if the map has six color bands, the legend needs six labeled bands. missing the lowest or highest band makes the reader misread.
a sibling read on color choices specifically is dashboard color theory: a non-designer’s guide which covers palette selection in depth.
conclusion
geographic visualization adds the most signal of any chart type when the data has a real spatial dimension. it also lies the most often, because area, projection, and absolute-vs-normalized choices all encode meaning the reader does not consciously parse. the per-capita rule, the projection guidelines, and the cluster handling above cover most of what makes a map honest.
the next step this week is to take one map already on your dashboard. check whether the metric is normalized, whether the projection is appropriate, and whether null regions are visually distinct. if any of those fail, redraw the map. for chart-type-vs-data-shape decisions on non-geographic data, see visualizing time series data and the ASEAN market research data sources guide for the regional data sources that often power Asia-focused geographic dashboards.