Geographic Information Systems Stack Exchange is a question and answer site for cartographers, geographers and GIS professionals. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

A poorly designed map can not only look visually unappealing, but can convey the wrong message, which could lead to bad decisions being made.

I would like to ask people to post examples (that are in the public realm) of poorly designed maps, WITH justification on why it is bad design.

Although this 'question' does not have a clear answer, IMO it will proove a useful resource to see what merits bad design, so others can learn what NOT to do. I will let the votes choose the 'right' answer.

I would also like to see examples of bad design around web-mapping.

I would argue that although GIS Professionals generally know how to create a good looking map, I would say that they also have a tendency to over-complicate web-maps, by trying to re-create a GIS desktop application on the web: alt text Also slow to use, and a PITA to even get to the map.

I think now that normal people are more used to Google Maps style/simplicity, web-maps should follow a similar approach.

With the explosion of 'NeoGeography' particularly in the web realm, we now have a lot of non-GIS professionals creating maps for the web. A lot of these developers are often very good at user-interface design, but not trained in cartographic principles. IMO, with web maps, its all about combining the skills of both cartography and user interface design.


locked by PolyGeo Jun 2 '15 at 21:59

This question exists because it has historical significance, but it is not considered a good, on-topic question for this site, so please do not use it as evidence that you can ask similar questions here. This question and its answers are frozen and cannot be changed. More info: help center.

Good topic. Speaking of 'NeoGeography' equally as 'bad' is the improper use of some maps / services. News agencies love to show panning and zooming animations of Google Earth and like satellite imagery. Often, I can't even tell what geographical region they're showing; all i can see are strips of imagery, etc. – Jakub Oct 31 '10 at 18:55

11 Answers 11

There are a lot of great bad examples at For example this map of Europe (part of a world map that uses colors excessively and without obvious purpose):

alt text

Besides the coloring, the labeling on this map is a mess.

Great link. I'll be visiting that blog to make sure I never contribute any examples. :) – Tim Rourke Nov 1 '10 at 14:19

Overlaying a huge amount of data without taking care of the readability makes such pixel blots:

alt text

(some people pretend these "mashups" are the Cartography 2.0. Maybe not).

+1 = You see a lot of this. If you really need a lot of data to be displayed in G-Maps, make use of clustering your markers.… – Simon Nov 2 '10 at 0:27
Yes, there is an obvious need for map generalization in web mapping. Some solutions exist (see but are almost never used in web mapping. – julien Nov 2 '10 at 0:38

Another example of cartographic bad design from a federal French agency (detail):

enter image description here

I was going to ask for some commentary concerning why these are bad maps, but one only needs a single glance to see what the problems are. Outstanding examples! – whuber Dec 22 '11 at 19:05
Are you kidding? I love flouro green on a low-contrast yellow background! – naught101 May 22 '12 at 0:31
Don't miss the blue gradient :) And that just really shows that this is not designed map, but just stock vector, used as background for whatever purpose and looks like out of league. – zetah Dec 3 '12 at 0:40

Beat the NATO 'Map-Mania' (this is aimed for Children and Adults)

Takes a long to load up (uncompressed/un-optimized flash content)

Very few features (blue graduated background and some country/coastline) that's it!

Very poor world projection - find Hawaii...

too much to add to the list.

double dare you to try it:

alt text


I found a couple of examples of very poor choropleth maps posted on wikipedia depicting income inequality and poverty by nation. Although both metrics depict a continuous phenomenon, they chose seemingly random colors to depict different ranges along the continuum (not even diverging colors, random bright colors). I've reduced the size in the files I've uploaded compared to the originals (GINI, poverty), as I believe they have made me ill viewing the full size files (so beware!)

enter image description here

For a critique, it seems obvious they should have chosen a continuous color scheme (one that goes in a single shade from light to dark or vice versa) for each of the metrics since they represent continuous data (you should follow your own advice wikipedia!) Perhaps the scheme would be appropriate for nominal data, but even then I don't think such bright colors are a great choice (perhaps a few small multiple maps depicting certain categories would be easier to read). For a more palatable set of color choices, checking out the work of Cynthia Brewer and her Colorbrewer applet is a good start.

As a note on the original source, it says it is from the CIA world factbook, (which the data surely is) but I couldn't find any map this silly looking at the actual CIA world factbook website and can't trace it back through the Wikipedia commons for those files. It may be an original work, which I thought was against Wikipedia's policy. Maybe instead of complaining I should upload a new one!

I've currently found another example of essentially the same problem in this wired magazine article, What a Hundred Million Calls to 311 Reveal About New York. They also have omitted the numerical break points for the legend as well! – Andy W Oct 3 '11 at 0:20
Original maps and diagrams are not against Wikipedia policy, just like original (unplagiarised) text is not. The policy is No Original Research (i.e., data, conclusions, interpretations). – Max Oct 24 '14 at 7:44
I consider the making of a map or graph to be original research (without regard to Wikipedia's policy), but I understand the utility of being able to upload supplementary materials to aid understanding. – Andy W Oct 24 '14 at 11:44

Although this map is aesthetically pleasing (at least the small multiples on the left), I believe it is quite poor when all three of the layers are plopped on top of one another. For what I believe are better ways of displaying such multivariate data check out this other thread on the site - Effectively displaying demographic data on a printed map .

Map taken from A better look at the actual map (with the ability to zoom) is available here.

enter image description here

Another example of color blending like this is in Friendly (2007). To be more explicit about what is problematic about these color schemes are that they confound observations in the color scheme. That is, a polygon can have different attribute values, yet receive the same color! (see this presentation, An Empirical Study of Colour Use by Paul Murrell and Ross Ihaka for a more detailed description of this). The above citation of Friendly gives an example in the footnote where two different sets of the same attributes would map to the same color. This just extends in trying to discriminate between observations by color in the current map. You have to do some impossible mental mapping of colors to figure out what the original attribute values are, and it isn't quite as simple as the legend makes it seem.

Below I have tried to recreate what their legend would have looked liked. Although I ended up being unsucessful, I think it is still enlightening as to what the problem is (I suspect I was unsuccessful not only due to my initial colors being off, but also because the transparency overlay is likely not consistent within and between colors to achieve the above map, it is also possible the way my software handles transparency and colors is different than their application). To read the legend is as follows, the ugly colors in the upper right is the blended panel of all threes colors. The array of how the colors are organized are demonstrated in the neighboring panels below and to the left. Within swatches the yellow gradient increases, down the panels the blue gradient increases, and to the right of the panels the red gradient increases.

enter image description here

It is easy to see the contrast between items is greatly diminished when the colors are blended together. Although it may seem like I intentionally re-created a crappy example, in a bit of experimentation I was never able to reproduce the array of colors within their map, and all of the produced legends suffered from essentially the same problem (so if you let me know the magic colors to produce their map I would be glad to replicate them here).

As a note, I suspect I saw this map referenced somewhere else besides the original site (perhaps FlowingData), that probably also made a comment about how poor the color scheme is. If I come across the other source originally pointing me to this map I will reference it. – Andy W Mar 30 '11 at 17:53
+1 The closer you look, the more puzzling the map becomes. It took me a while to realize that red and yellow are representing similar things--educational attainment as measured in two ways--but in inverse senses, so that there are really only two factors being shown here, not three. – whuber Dec 22 '11 at 19:21
"a polygon can have different attribute values, yet receive the same color!" - I don't see how this can be true, considering they are using three primary (orthogonal) colours. Every colour should have a unique combination... – naught101 May 20 '12 at 23:23
@naught101, you are correct, I misinterpreted Friendly's original article, his legend that he produced confounded observations, not the actually amounts of RGB in the map (which is alittle naughty of him to produce an inaccurate legend), see footnote 14 on page 392. I will update my response in a bit when I get a chance, but note it largely doesn't change my point. Orthogonal in color space is not orthogonal in how we interpret different colors. In particular, when mixing an already saturated color it is very difficult to distinguish between saturation in another hue (let alone 2 hues!) – Andy W May 21 '12 at 12:20
Definitely agree on the point about interpretation of colors. Was just being nit-picky :) – naught101 May 22 '12 at 0:30

Not excactly an answer, but in the same vain as this thread. The book: How to Lie with Maps by Mark Monmonier is a fun read. Most of it is bad maps intentionally created to distort or hide data. Its so easy to manipulate maps to get your point across, the ideas in this book are good to keep in mind to make sure you don't really cross the line.

enter image description here


More 'lies': – radek Mar 31 '11 at 18:51
Mark Monmonier's how to lie book is certainly a classic and deserves to be widely read (and understood!). This answer however doesn't tell the reader what is wrong with the map depicted. The cutoff percentages are different, so what's wrong with that? – matt wilkie Sep 12 '11 at 19:12

I offer this one because it is typical of illustrations in a well-regarded textbook on cartography, so it's not a one-off bad map: it exemplifies what people are being taught in (some) universities.

enter image description here

The caption beneath it reads,

Map of residuals from regression. The geographic areas having Y values considerably under- or overpredicted relative to the regression line are mapped with identifiable area symbols. These sections of the study area need to be investigated more closely. Further data may be needed to determine why the dependent variables in these section behave as they do. Identifying deviate areas is a major application of residual mapping.

The problem is that this map and its interpretation are wrong in many important ways, starting from the very concept of its construction: a choropleth map of residuals is inferior to many other techniques available. The pattern of residuals mapped here not only is deceptive due to the poor cartography (and a truly awful legend), but in fact it is to be expected of residuals from a really good regression! In this fashion the book's author is creating a problem where none exists, developing unrealistic expectations, and recommending a potentially expensive and meaningless additional data collection effort.

I believe strongly in attributing one's sources, but perhaps I will be forgiven for not revealing the identity of the guilty party in this case. – whuber Dec 22 '11 at 19:01

A recent contender for the title.

I'll only describe my principal critiques on the fist map : "Population", seen at the default scale and position with Firefox (OSX).

  • Absolute quantitative data are classified and represented by color classes : huge information reduction. Approximatively 200 numerical values are reduced to five color values.
  • The classification method is peculiar : Indonesia and China belongs to the same class (234M / 1,338M).
  • The classification presented in legend does not reflect the real data values. Paraguay population is 6.4M, but the legend says its class begins at 9.8M (but the values are drawn from same source : World Bank).
  • The projection is inappropriate : it's a Mercator, look at the size of the Groënland for example. The size of the continental masses are grossly distorted relatively to the real surfaces proportions. No justification for this.
  • As a result of the previous remark, the scale presented is only true at a specific latitude (not precised), not for the whole map as implicitly implied.
  • Data granularity is not coherent with some displayed large map entities : Alaska is in the same class as U.S.A, Groënland as Denmark.
  • As a result of the previous remark, Groënland is falsely represented as belonging the minimal population class, which starts at 9.8M, but its population is only 57,000.
  • Legend colors are different than the map colors (verified with the photoshop eyedrop tool) (!).
  • At the default scale, the great lakes and inland seas are the same color as the countries, only a shade darker.
  • At the default scale, the map is cropped in the four directions.
  • Order of elements in the legend are reversed, compared to the recommended order (see for example T. A. Slocum, Thematic Cartography and Visualization, Prentice Hall, 1999 and B. D. Dent, Cartography: Thematic Map Design, McGraw-Hill, 1999), but it's the default setting for ESRI software.
  • Contour color is not enough contrasted relatively to some of the darker surface colors. For example, try to see the border between Turkey and Iran, or between some west-european countries.
  • Inexplicable lack of data : French Guyane.
  • Sorry, too tired to continue, it's quite painful.

My source for the discovery of this "pearl" is Maxime from the ForumSIG.

Please elaborate on your critique of the map. Perhaps we should not be worried about population growth, as about 1/3 of the Northern hemisphere (Russia) is decreasing in population growth! – Andy W Jan 4 '12 at 20:19
(+1) Good critique! The narrative is as confusing as the map, because it talks simultaneously about population and population density without clearly distinguishing them. This confusion may be at the root of many of the inappropriate map design choices: the map maker isn't really sure what they are showing or trying to say. – whuber Jan 4 '12 at 21:43

These are popping up ALL over FB, Gmail, and in general the internet. Everyone raves about them, but I think they're awful, especially considering they represent such simple data. Questions I have:
What does the darkness/density of color mean? More responses? Was that more responses per capita, per surveys distributed?
Are the white areas no response? Were they surveyed at all?
enter image description here Unfortunately the answers to those questions are available on the PhD student's website, but completely neglected when other websites shared the maps.

Example of how the maps are being shared:
Joshua Katz's description:


One of my personal favorites comes from p66 of the Eddington Transport Study (p20 of the linked PDF):

There are several immediate problems with this map:

  1. Another answer pointed this out for another map; this map overlays significant amounts of data such that it doesn't depict much more than a rough estimate of population density.

  2. Is rateable value by age depicted by the area of the pie chart, or the diameter? The legend says "rateable value by age", so is there a correlation between the area or diameter of the pie chart and the age distribution?

  3. In each pie chart, do the proportions represent the percentage of warehouses that were built in that time period, as of 2003?

Location of new warehouses with concentrations at motorway intersections


Not the answer you're looking for? Browse other questions tagged or ask your own question.