Skip to the main content

Click for the home page
FAST Consulting
FAST Consulting Home Web UIs Desktop UIs Services Websites About Our Team Amusements
FAST Consulting

Preface to the GUI Design Handbook
Checkbox to Cursor
Dialog Box to Drop-down List
Graph to Iconic Label
Keyboard Shortcuts to List Box
Menubar to Message Box
Online Help
Palette to Pushbutton
Radio Buttons to Status Bar
Table to Wizard

14 Steps of GUI Design
Color & Pattern
FAST Consulting

You are here: Home ~ Desktop UIs ~ Graph - Iconic Label


A graphical method for displaying numeric and quantitative information.

Good for:

Quickly showing relationships among data points (Fig. 1).

Fig. 1. A line graph.

Not good for:

Showing details like the data points themselves.

Design guidelines:

Following are the most common types of graphs. Make sure that you select the right graph type for the data and that your format is correct. For example, many applications mistakenly use an area-chart format for line charts or a histogram format for bar charts. (For more information, see Fowler and Stanwick, 1995, chapter 7, "Charts and Graphs.")

Changes over Time


Other names: Column

Good for comparing or ranking a small number of values (no more than 10 or 12).

Also useful when the data sets are so similar that they would overlap if shown as lines. By using a bar chart, you can visually separate the data sets.

The spacing between bars or sets of bars should be 1/2 the size of the bars.


Clustered bar, zero-line


Other names: Time series

Good for comparing one set of values to another. Also good for displaying trends.

Line graphs show interpolated points and slopes well.


In finance, high/low/close (in commodities field, also called "bar"); candle charts.


Frequency polygon

Other names: Bell curve, mistakenly—bell curves are smoothed normal distributions.

Good for showing counts—how many times something happened or how many times a number appeared. Shows frequency distributions (the count for each interval during which data were collected) as smoothed curves.


See Histogram.


Other names: Step

Good for comparing counts.

Shows frequency distributions as steps or bars.

Good when values fall into discrete sets and not good when they don't.

Note: Histogram bars always touch. Bars (or sets of bars) on bar charts do not touch.


Pyramid histogram


Other names: Scattergram, XY scatter

Good for spotting clusters or out-of-range points. Each data point is the intersection of two variables plotted against the two axes.


Bubble chart



Other names: Surface, component part, belt, mountain

Good for showing cumulative totals over time. Each data set is added to the data set below it, so that the top edge of the top set is the sum of the data at any point on the timeline.

Totals can be numbers or percentages.


Other names: Circle, cake, sector

Good for showing snapshots of proportional relationships, one snapshot per period of time. One pie is one whole (100 percent).

Bad for comparing two or more relationships. Most people find it hard to compare wedge-shaped areas from one pie chart to the next.

Segmented bar

Other names: Stacked bar chart, sliding multi-component bar chart, population pyramid, butterfly chart, subdivided bar chart

Good for showing proportional relationships (like pie charts) over time (like bar charts).

Use to compare parts of a whole¾ for example, how interest and principal equal total savings.

Do not include parts and the whole in the same bar. For example, don't stack interest, principal, and total savings on the same bar. The bar will be twice the height it should be.


Paired horizontal bar chart (deviation bar chart)

Color on graphs

Eight percent of men—one in twelve—have red-green color blindness. In other words, if you have 24 men in your office, two will have trouble separating red from green, either when the colors are next to one another or when the lights are dim. (Note that most individuals with color blindness see all colors of the spectrum, but simply can’t tell the difference between two of them. For this reason, "color confusion" has replaced "color blindness" as the term of choice.)

What does this mean to your GUI? It means that using red and green lines as the default colors in your graphs, for example, is a bad idea. Every twelfth male user won’t have a clue as to what the graph says. Using red and green borders to indicate which window has focus, as Motif does, is also a bad idea. Another bad idea is using red lettering on a black, brown, or green background, since all of these colors may blend into one another for users with color confusions.

However, say that you never use red and green in your graphs because, well, you don’t like green. Instead, you use blue and orange, or red and blue.

It doesn’t matter which colors you pick. Everyone without a color printer might as well be color-blind when he or she prints out that graph.

The solution

There are a variety of solutions to this problem. However, the bottom line is this: Use color as a secondary, not a primary, signal.

Use color to quickly show correlations between things (for example, all required fields have a blue border) or to indicate changes. For example, you could have a temperature gauge on which a virtual mercury bar changes color as the temperature gets higher. The height of the mercury would be the primary cue. The change in color would be the secondary cue.

Following this rule, then, you could revise a line graph as shown in Fig. 2: One solid line, one broken line. (Remember that black and white are colors too—on a white background, a black line has more contrast than any color and vice versa.)

Fig. 2
. The right way to do a line chart.

You can fix clusters of bars on bar charts by using various hatching and fill patterns. However, the results can become very busy (see Fig. 3, "Poor").

Fig. 3. The wrong and right ways to do a bar chart.

For a two-bar chart, the best solution is one solid black bar (or white bar, depending on the background color), and one empty bar. For charts with more than two clustered bars, use shades of gray (Fig. 3, "Good"). You can either use shades of gray or you can use colors with appropriate grayscale values. Every color has a grayscale value (its "darkness," so to speak) as well as a hue (the "redness" of red). If you use colors separated by 20 percent differences in grayscale, everyone will be able to tell the bars apart.

The best way to check for grayscale is to create a grayscale chart, and then put your colors on the chart and compare them to the grays in low light or by squinting. If you’ve matched your color to the grayscale chart accurately, the color seems to fade into the gray. For a very low-tech approach, simply pick a few likely colors, fill in your bars, and squint. If the colors stay separate, then you’ve picked the right colors.

How to create a gray-scale chart

1. Pick a program with a color or palette editor. Open the editor and either find or create a set of nine grays and one black separated by 10 percent differences in darkness. Use white for the background. The values for each gray are:


RGB Values

HSV Values

HEX Value


229, 229, 229

0, 0%, 10%



204, 204, 204

0, 0%, 20%



178, 178, 178

0, 0%, 30%



153, 153, 153

0, 0%, 40%



127, 127, 127

0, 0%, 50%



102, 102, 102

0, 0%, 60%



76, 76, 76

0, 0%, 70%



51, 51, 51

0, 0%, 80%



25, 25, 25

0, 0%, 90%


100% (black)

0, 0, 0

0, 0%, 100%


2. Draw a set of gray boxes on a white background, one color of gray per box, ranging from 10 percent to 100 percent (black).

3. Draw diamonds of all the colors you want to test.

4. Drag each color sample over the chart, squinting as you drag it. When the color and a gray box seem to match, you've found its gray-scale value.

5. Save the colors that are either 20 or 30 percent apart (separated by two or three boxes) and discard the rest.

Note: Some colors, because of their brightness, maintain high contrast no matter where you put them on the grayscale. However, check the size. Small areas of yellow disappear against white. Red, if used for small dots or thin lines, shrinks away to nothing against a dark background. At the other end of the spectrum, avoid thin blue lines and text. Because of the structure of the eye and blue’s wavelength, it is hard to focus on small blue objects.

Show underlying data points

Make sure that users can access the underlying data. Here are a few methods:

  • Let users switch between tabular and graph views with a pushbutton or tabbed dialog boxes (Fig. 4).

Fig. 4. Add a table/graph pushbutton.

  • Let users toggle data points on and off with a check box (Fig. 5).

Fig. 5. Toggle data points with a check box.

  • If the user clicks on a data point or bar, show the underlying value (Fig. 6).

Fig. 6. Show the value when the user clicks the data point.

Recommendations for titles

For marketing or sales-oriented graph applications, find a way to let users put a message or point of view in the title—let them emphasize the point of the data. For example, "Company Sales Trend" doesn't say as much as "Company Sales Up in Northwest" or "Sales Down in Southeast." Provide a default title that users can overwrite if they want.

Recommendations for labels

State the units of measurement. Include the units in the X and Y axis labels. For example, if the dependent variables are percentages, include "Percent" or "%" in the Y-axis label.

Don't stack letters vertically. When the left margin is too narrow for the Y-axis label, the label should appear above the margin. If you cannot put the label there, don't stack the letters like this:


Most people read by recognizing the entire shape of the word, not individual letters. When you stack the label, you force the reader to puzzle out the word from the letters. You can, however, turn the label sideways without causing as much of a readability problem.

Make labels clear. Spell out all words. If space is very tight, abbreviate using only standard abbreviations or symbols. Check an abbreviation dictionary.

Try to avoid keys (legends) whenever possible. Instead, put explanations on the bars, lines, or data points themselves. For example, rather than creating a key to explain what a set of bars means, label the bars directly by adding numbers to the data points or to the tops of bars.

If you do use a key, try to put it inside the graph panel, in a spot where there are no data. If you put it outside the panel, the eye is drawn away from the data. Do not box the key because then the box draws attention to itself.

Usability tests:

Obviousness: Are the goals of the graph apparent? For example, if the graph is supposed to highlight out-of-range data points, can users spot them immediately? Is the title too generic—can the users recognize the use or contents of the graph from the title?

Affordances: Do the users recognize the graph type and, if so, does it help them understand the data more easily?

Heuristics: Do experts agree that you’ve formatted the data correctly? Check with people with expertise in statistics and mathematics.

Many industries and business domains have specialized types of graphs. As well as developing lists of subject-matter expert reviewers, it might help to collect standard reference works, even textbooks, in the domain for which you’re developing graphs. Expect expert users such as stockbrokers and doctors to be visually literate and to prefer windows full of complex graphs and charts.

In addition, different cultures have different levels of visual literacy. Unlike mass-audience U.S. readers, for example, Japanese readers expect and can understand highly complex pictures, charts, and graphs (Kohl et al. 1993, 63-73). If you expect to internationalize your applications, check all graphical and data-analysis requirements with your international experts and marketing departments.

Mechanical: Users often prefer to see preformatted graphs as their first experience with a graph program. Later, if they need to, they can fine-tune the display. Have you made it easy for the user to get an interesting graph the first time he or she uses the application (perhaps with a wizard, if the display or data are complex)?

See also:


For examples of successful and failed charts, graphs, and tables from a variety of businesses and situations, see Edward Tufte’s Envisioning Information (1990) and Visual Display of Quantitative Information (1983).

To avoid stepping into one of graphing’s many pitfalls, see Darrell Huff’s How to Lie with Statistics (1954), a slender book now (for good reason) in its fortieth edition.

Gene Zelazny’s Say It With Charts: The Executive's Guide to Successful Presentations in the 1990s, is designed for people doing business presentations. The book explains what all of the major chart types are, how to use the various types correctly, and how to make them easy to read. Zelazny then goes a step further, and shows you how to create sophisticated and elegant charts from the basics.

Go to top of page

Icon, Desktop

A picture used to indicate and start an application or an object (printer, wastebasket).

Good for:

Starting an application (Fig. 7).

application icon

Fig. 7. Application icon.

Using an interface object, such as printing a document by dragging it to a printer icon (Fig. 8).

object icon

Fig. 8. Object icon.

Design guidelines:

The desktop icon not only identifies your application but it has a marketing and corporate identification function as well. A successful icon:

  • Looks different from all other unrelated icons on the desktop.
  • Resembles your corporate logo (or at least uses the same colors, shapes, and typefaces).
  • Gives the users some idea of what it does or represents. For example, a word processing program's icon might include a pen, an accounting program might show a ledger, and so on.
  • Is recognizable when it is no larger than 16 pixels square.
  • Looks as good in black and white as it does in color.

Because desktop icons are a very important part of your company’s corporate identity, make sure that the same professional artists and designers who designed your logo also design your icon (or oversee its design). This is not a job for amateurs.

Usability tests:

A desktop icon must be discriminable from all other desktop icons and should, if possible, indicate the type of software it represents (network tools, word processor, etc.). Use a paper-and-pencil matching test.

See also:

Iconic Label.

Go to top of page

Iconic Label

A pictorial description of a tool or function that generally appears on toolbars and palettes.

Good for:

Identifying tools that require a mouse or pen to be effective. Drawing tools—paintbrushes, erasers, and so on—are typical examples (Fig. 9).

Fig. 9. Tool label.

Identifying mouse shortcuts—for example, Save, Cut, Copy, Paste. Use standard images for these types of options (Fig. 10).

Fig. 10. Mouse shortcut for Save.

Not good for:
  • Abstract functions for which it is difficult to find visual metaphors (for example, Sort). Use text labels instead.
  • Functions that do not require or benefit from mouse use.

Design guidelines:

Most development packages contain sets of iconic labels that you can use for free. Take advantage of the most popular ones: Save, Cut, Copy, Paste, New, Open. However, other icons may not be very recognizable or standard. If you use these less well-known icons, make sure that they have tooltips and that you test them for comprehensibility. (See "Usability Tests" below.)

Official sizes

The Windows 95 style guide suggests that you let users toggle the sizes of the toolbar buttons. These are the suggested sizes:

  • 24x22 pixels and 32x30 pixels for the buttons themselves
  • 16x16 and 24x24 pixels for the picture labels

The guide also contains a list of all Microsoft’s common toolbar images and the official names of their functions (Microsoft 1995, 176-178). Other platforms have similar suggested icon sizes and uses; most development kits come with a basic set of iconic toolbar buttons. Reuse whatever you can.

What really works

Mullet and Sano (1995, 201-202) report that the toolbar icons described as most useful typically correspond to either concrete attributes of visible objects (font attributes, paragraph alignments, etc.) or to concrete system objects (printers and folders).

They add that it is difficult to develop visual representations that distinguish between similar functions (Save and Save As) or between controls that should probably look similar but have different behaviors (print a document, fax a document).

As visual designers, they argue for putting abstract commands and activities on menus, where you can use words to describe them, and putting only concrete settings and tools on toolbars and palettes.

This is the ideal solution. However, most development companies don’t live in the ideal world but rather in the trade-off world. Although an idealist might find this solution messy, human-factors research indicates that buttons with both pictures and text labels are the best solution. See "Usability Tests" below.

How to design an iconic label

The first rule of icon design: Start designing on paper, not on the computer. Sketching on paper is faster. It’s also easier to throw away the bad ideas when they’re only on napkins and scrap paper.

Following are some hints for finding and developing the right images.

Show the object (when you can)

Showing the object is probably the most direct way of communicating an idea. For example, to represent the idea of printing a file, make an icon of a printer. Or to represent deleting a file, show a garbage can or wastebasket.

If necessary, turn verbs into nouns. It is much easier to represent an object (a noun) than it is to represent an action (a verb) through pictures. For instance, rather than trying to represent the action of printing a document, it is much easier to show the printer.

Use a visual analogy

If you can't show the object itself, try to use a visual analogy.


Have an image stand in for an idea. For instance, say that you need an icon for a maintenance program. To condense a complex idea into a simple image, use a picture of a maintenance tool¾ a wrench, for example (Fig. 11).

Fig. 11. Possible maintenance icon.


A litotes (pronounced LIE-ta-tees) is a way of describing something by stating the negative of its opposite. For example, when you say "not bad" for "good," you’re using a verbal litotes. A picture of a broken chain could be a visual litotes representing "freedom." Any image inside a circle with a slash through it is a litotes (Fig. 12).

Fig. 12. A litotes.


A synecdoche (pronounced si-NEK-da-kee) is a metaphor in which a part represents the whole. For example, a chili pepper could represent a Mexican restaurant (Fig. 13), the Golden Gate bridge could represent San Francisco, or the Eiffel Tower could represent Paris.

Fig. 13. A synecdoche.


Hyperbole uses exaggeration to convey an idea. For example, you could use a bomb to represent danger, a bulging garbage can to represent a thrown-away file, or a remarkably steep incline to represent a hill (Fig. 14).

Fig. 14
. A hyperbolic image.

Make sure that a thread runs through the buttons

Fig. 15. A thread runs through a successful toolbar's buttons.

When you're designing the pictures for toolbar buttons, simplify the process by finding a theme for the buttons (Fig. 15):

Give the buttons an intellectual theme. For example, for a recipe program, "sort" could be represented by a sieve; for a banking program, it could be represented by a coin dispenser.

For models of good thematic design, look at the symbol sets developed for the Olympic Games and the symbols developed by Xerox Corporation for copier and office equipment.

Give the buttons a visual theme¾ for example, use only geometric shapes or only rounded shapes, or use a common color palette. Macintosh buttons use similar colors by default, since only 34 colors of the entire 256 color palette are available in ResEdit. Microsoft Windows 3.x provides only 16 colors.

Reuse existing symbols

Many industries have their own iconographies¾ electricians have symbols for transistors and electrical lines, telecommunications workers have symbols for central offices and the public network (a cloud), network engineers have symbols for nodes, WANs, LANs, and so on. In the U.S., the American National Standards Institute publishes and/or redistributes national and international standards, including symbol sets. The symbol sets include everything from agricultural equipment and aircraft systems to electrical and mechanical systems to warning signs.

Helpful hint: Use old-fashioned images. Traditional, simpler, images sometimes work better than new ones. For example, a skeleton key—not a modern button lock, for example—is often used to represent locked or safeguarded information on Web sites. In the U.S., the most common image used for electronic mail is the rural mailbox, which is widely recognizable even though it is seldom seen in the cities where most users live (Horton 1994, 46-47).

If all else fails, find idiomatic images

Alan Cooper in About Face (1995) suggests that humans can learn and remember idioms ("beat around the bush" or "cool") very easily without relying on comparisons to known situations. Many idioms have no metaphoric meaning at all, he points out; the stories behind others were lost ages ago.

Although idioms must be learned, Cooper says, good ones only need to be learned once. It is quite easy to learn idioms like "politically correct" or "the lights are on but nobody’s home" after a single hearing. It is also easy to learn how a scrollbar or a resize button works, he says, and as neither exists in the real world, they are clearly not metaphoric. They are idiomatic.

How do you create an idiom, then? Branding—marketing—advertising, he suggests. "Synthesizing idioms is the essence of product branding, whereby a company takes a product or company name and imbues it with a desired meaning. Tylenol is, by itself, a meaningless word, an idiom, but the McNeil company has spent millions to make you associate that word with safe, simple, trustworthy pain relief" (Cooper 1995, 59-60).

In terms of a user interface, then, an idiom is any image or action that is striking enough to learn quickly and that has a reasonable affordance. (A scrollbar that closed documents, for example, wouldn’t make sense and would therefore be hard to learn.)

At the end, simplify the design

Once you have an icon or set of icons that everyone likes, simplify it (Fig. 16). See how many elements (colors, lines, shapes) you can remove without "breaking" the icon.

Fig. 16. A good design, and a too fancy design.

Usability tests:

The issues in testing iconic labels are:

  • Whether users understand the meanings of the pictures immediately (ease of learning)
  • Whether they remember the meanings readily, once learned (memorability)
  • Whether users can discriminate between similar pictograms or similar ideas
  • How long it takes them to become proficient and whether they graduate to doing complex or sophisticated tasks (Horton 1994, 302)

Ease of learning is especially important when you’re writing for casual or inexperienced users. Test using a paper and pencil matching test.

Memorability is most important when you’re writing for experienced users. If your user group will use the product daily, you can use "nonsense" pictures (a yellow triangle, a green square, etc.), provided that the icons are visually distinct from one another.

Some pictures may be hard to recognize no matter what you do to improve them. The solution is not to beat your collective heads against the wall, trying to find the best image. Rather, add tooltips. Many researchers have repeatedly found that images combined with text works better than images alone or words alone.

See also:

Icon, Desktop; Palette; Toolbar; Tooltip.

Go to top of page

FAST Consulting
FAST Consulting Home | Site Map | Accessibility | Web UIs | Desktop UIs | Services | Websites | About Our Team | Amusements
FAST Consulting
© 2008 FAST Consulting All FAST ConsultingRights Reserved Privacy Statement