Don’t choose a visualization method in advance. The goal of a graphic is to invoke a content response, not a design response (i.e. wow, nice graph).

Organization charts are not informative — unless the lines (relationships) are annotated.

Every graphic should provide reasons to believe. Detail helps credibility and raises interest (as an extreme example consider aerial pictures). If you don’t show all the data you must be a “cherry picker”.

Letting people look at data themselves stimulates their interest and gets them thinking (and has more impact). More data points means more “hooks” for people to (potentially) be interested in.

When possible, bring something real to a presentation (for example a hospital bill when discussing medical expenses).

Annotate everything — even your text. Don’t use foot notes, create side notes.

It rarely makes sense to order items in tables alphabetically.

The Gil Sans font (a compact sans-serif font) works great for (time-)tables.

Don’t go for the lowest common denominator. Don’t be “flabby & generic”.

Make presentations that make proper use of the human eye-brain system (which has a bandwidth of ~10Mbs).

Tables outperform graphics up until ~500 data points.

Respect both the content and the audience. Why is it that an average PowerPoint table contains only 12 numbers, but tables in the sports section of a magazine can burden their audience with an average of ~200 numbers?

Distribute data in space (using multiple monitors, if must be) rather than not time (e.g. across several slides).

People should have their names on their work (as pride often results in better work).

Don’t separate information depending to its mode of production (e.g. in separate “multimedia” slide shows).

Be aware that design often mimics the hierarchy in an organization.

The original computer desktop at Xerox had icons for documents only (i.e. now trash icon etc).

Create “infographics” that you can leave around (maybe people will give them another look during a following, boring presentation).

“No matter how beautiful your interface is, it would be better if there was less of it” (though I don’t think he’d extend that all the way to command line interfaces).

“Flat” interfaces beat “hierarchical” interfaces in terms of usability.

Efforts to provide “personalized” interfaces are “pathetic” (shows that you don’t know how to present something, and try to offload the responsibility to the user).

Going beyond sparklines: Wavefields (?).

Instead of doing “presentations”, dump the data on the audience, let them read through it, and then have a “guided discussion” (i.e. answer questions).

“Kiwi” is a problematic term, as it can be used for:

• a bird
• a fruit
• people from New Zealand
• a name for various products or businesses

Moreover, the different uses of the term “Kiwi” are often closely related. For example, the Wikipedia article on “Kiwi (people)” contains the word “bird” (“The name derives ultimately from the flightless bird native [...]“). But this document shouldn’t appear (or score high) when we are looking for information about Kiwi birds.

Google

Google shows a suggestion to extend the query to “kiwi fruit” somewhere half way down in the list of results. The first hit is for some software called “Kiwi Syslog Daemon”. There is also a “definition” link that points to Answers.com where you get this:

Or did you mean: kiwifruit , kiwi (bird), Kiwi (Marine lingo), kiwi (people), Kiwi (Dragon Ball), Kiwi (shoe polish), Kiwi (horse), Kiwi (2002 Album by Dim Dim), Kiwi (Gaming Character)

That’s nice, but it’s just definitions, which doesn’t help us much with our web search (at least not directly).

Yahoo!

Yahoo! suggests (at the top of the results page):

Also try: killer kiwi, kiwi fruit, kiwi carpet cleaning, kiwi bird

No comment…

Live Search

Live Search suggests these “related searches”:

Related searches: Kiwi Fruit, Kiwi Bird, Kiwi Plants, Kiwi Casino, Killer Kiwi, Kiwi Bingo, Kiwi Photo Gallery

Also — for reasons beyond my comprehension — there are no less than three ads for buying “thong bikinis”…

Ask

Ask lists matching Wikipedia articles at the top (in a dropdown list). This isn’t terribly convenient, but in this case it does give a good idea of the different uses of the term. When you select an article the search is redone to match the selected article. Also, there are these suggestions:

Narrow Your Search: Eat a Kiwi, Kiwi Shoe Polish, Kiwi Fruit Facts, Kiwi Recipes, Kiwi Origin, Kiwi Bird Live, Kiwi Fruit Come
Expand Your Search: Kiwi Bird, Strawberries, New Zealand, Oranges, Pineapple, Grapes
Related Names: Peaches, Rhea

Clusty

Clusty is a “meta” search engine, i.e. it collects its results from other search engines, and reorganizes them a bit. This limits the possibilities as they must rely on the first n hits returned by the other search engines, but let’s see what they do:

New Zealand (82)
Fruit (11)
Game (8)
Blog (9)
Club (8)
Syslog Daemon (5)
Designs (7)
Cool Kiwi (6)
Holiday (4)
Specializing (4)

The indication of the number of matches is certainly useful. But why is “Kiwi” under “New Zealand”?

WebClust

WebClust is another “meta” search engine that tries to organize the output of other search engines better.

Shop (19)
Kiwi Fruit (19)
Amp (16)
Kiwi Bird (12)
Design (10)
Shipped,Price (10)
Kiwi Alpha (7)
Kiwi Music (5)
Pineapple Kiwi (5)
Version,Available (3)
Body,Bath (3)
Thousands (3)
Other (88)

Restricting to “Kiwi Bird” does get rid of some matches, but still includes “Kiwi Shoe Polish Reg. Size”…

Mooter

Mooter doesn’t return a list of results, but a graph that must be clicked on first:

fruit
men
book
kiwi
find
shop
accessories

Hmm…

Conclusions

1. Current automatic clustering methods seem to be unable to produce suggestions that make much sense, for the most part.
2. None of the search engines seems to be able to understand what a document is talking about, but instead rely on the occurrence of specific terms.
3. The best you can expect from current search engines is a suggestion to rephrase a query e.g. as “kiwi bird” — something most people with a bit of experience will do on their own.

Searching is more often than not an iterative process.

People rarely enter more than two or three search terms.

Larger search boxes encourage people to enter more search terms.

Most searches are done in order to a. find specific information, b. look for a page (navigation) or c. obtain a resource (download).

There seems to be some confusion between the search field and the browser’s address bar.

Experts start out searching similar to novices, but have more strategies to fall back to if the search is unsuccesful.

Page authors use different words when talking about the same thing. Making use of the anchor text people use to link to a page is a partial solution.

People are better at recognition than recall. They may have trouble formulating what they want, but can easily recognize and click on a link with an appropriate suggestion.

Web search is hard because it’s hard to know what people want and the data is heterogenous.

Searching can be frustrating because you don’t know if you didn’t find anything because there isn’t anything, or because the query wasn’t done properly.

There hasn’t been much innovation in the past years in search user interfaces. There is a lot of pressure to conform to what people are used to (i.e. Google).

[...] current ad hoc IR systems are not able to retrieve our example sentence when they are given the query ‘yeast cell cycle’. Instead, this could be achieved by realizing that ‘yeast’ is a synonym for S. cerevisiae, that ‘cell cycle’ is a Gene Ontology term, that the word ‘Cdc28′ refers to an S. cerevisiae protein and finally, by looking up the Gene Ontology terms that relate to Cdc28 to connect it to the yeast cell cycle. Although this will not be easy, we see this form of query expansion as the next logical step for ad hoc IR.

In the long run, simple style with quality construction and efficient performance wins. So a business that [creates] web sites for shipping, commerce, or information, would be wise to stick to the fundamentals. [...] Only for products whose goal is “entertainment, or style, or perhaps enhancement of a personal image” does fashion come into play. The design moust match the target audience and “it is probably necessary to have multiple versions of the design for different market segments” and “rapid changes in style and appearance” are required.

Three levels of design are distinguished:

1. visceral
2. behavorial
3. reflective

Visceral design:

People are often willing to overlook shortcomings if a product is pleasant to use.

The visceral level is the “look and feel” that makes people want to own your product, regardless of whether they need it or not.

Behavorial design:

At the behavorial level “appearance doesn’t really matter [...] performance does”. Design at this level means talking into consideration not only function, but also understandability and usability.

First, it is important to recognize that

tasks and activities are not well supported by isolated features

so check-list based development won’t do – you need to understand just how people will use a product.

There are two kinds of product development: evolution and innovation.

Asking potential customers for their views is not an effective way to innovate:

Focus groups, questionnaires, and surveys are poor tools for learning about behavior, for they are divorced from actual use. Most behavior is subconcious and what people actually do can be quite different from what they think they do. We humans like to think that we know why we act as we do, but we don’t, however much we like to explain our actions.

People have said they would really like some products that have failed in the marketplace. Similarly, they have said they were simply not interested in products that went on to become huge market successes.

Evolutionary development, too, is best done by watching people, as “people find it difficult to articulate their real problems”. Also, many people blame themselves for problems they have.

Regarding usability:

[The product] should not require years of dedicated practice. New items appear every week, but who has time or energy to spend the time required to learn each one? Bad design is a frequent cause of error, often unfairly blamed on users

designing for the handicapped, the hard of hearing or seeing, or those less agile than average invariably makes an object better for everyone.

Why do so many designs fail?

• Engineers tend to focus upon technology, putting into a product whatever special features they themselves prefer.
• Designers like to make sophisticated use of images and metaphers that win prizes in design compititions but create products that inaccessible to users.
• Managers focus upon making sure that each division in a company receives the recognition its political power deserves.

Good behavorial design has to be a fundamental part of the design process from the very start; it cannot be adopted once the product has been completed.

Therefore it is

important to have a rapid, iterative development process. The prototype is tested with users right from the beginning.

Reflective design:

This is a more subtle level of design that not all people may be able to appreciate – much like good movies…

The best designs come from following a cohesive theme throughout, with a clear vision and focus. Usually, such designs are driven by the vision of one person.

If you want a successful product, test and revise. If you want a great product, one that can change the world, let it be done by someone with a clear vision. The latter presents more financial risk, but it is the only path to greatness.

Another issue is customization, which the author invariably fails to find useful:

The things I really want to customize [...] can’t be customized. [...]

If something is so complex that it requires the addition of multiple “preferences” it is probably too complex to use.

Proper customization comes through combining multiple simple pieces. I don’t customize my pen; I do customize how I use it.

The main purpose of visualization is to help people detect patterns – though the author cautions that people are quite capable of seeing patterns where there are none… The author also cautions against overuse of visual techniques:

Natural language is the most elaborate, complete and widely shared system of symbols that we have available. For this reason alone, it is only when there is a clear advantage that visual techniques are preferred.

Some practical advice extracted from the book.

• Don’t use grayscale values to color categories in a graph.
• Blue areas focus behind red areas, for most people. This can be used to create an illusion of depth. It can also be used to make text unreadable…
• Colors allow an additional three dimensions to be visualized (rgb!)
• Fast movements or sudden appearance of items in the visual periphery attract attention – which may or may not be intended.
• Multi-dimensional, discrete data can be encoded by spatial position of a “glyph”, color, shape, orientation, texture, motion or frequency (e.g. blinking).
• Items can be associated visually by making use of proximity, similarity (see previous point), symmetric or continuous arrangement, or enclosure (e.g. in windows or Venn-diagrams).
• Animation can be useful though “the kinds of animated critters that are starting to crawl and hop over web pages are often unnecessary and distracting. Just as elegance is a virtue in static diagrams, so it is in diagrams that use animation.”
• People are poor at remembering images, but good at recognizing previously seen images.
• Large collections of images can be scanned faster by flashing the images in rapid succession (up to 16 images/s) than by looking through thumbnails (requires refocussing, only 3-4 images/s).
• Display numbers in an object view rather than in tables, if there is a natural or metaphorical mapping to a physical object.
• For displaying 3D surfaces use a single light source that is located above (the brain assumes this!), at an infinite distance. Make shadows soft, and use textures.
• Using three dimensions may increase the maximum complexity in a tree or graph that can be understood.
• Most people get sick in virtual reality simulators when they move their head from side to side while moving.
• Flowcharts are less suitable than natural language or pseudocode for helping people understand processes, because they require the brain to perform an additional translation step.
• Diagrams on the other hand can help people understand relationships.
• Data input always has a speed-accuracy trade off. This means that applications that are tolerant of small errors allow people to work faster.

So what are the key ingredients of a good-looking website? Caudron suggests that the amount of graphics on the page should be strictly limited, perhaps to a single eye-catching image. “It’s not about getting as much stuff on the page as possible,” he says.

These days, enlightened web users want to see a “puritan” approach, Caudron adds. It’s about getting information across in the quickest, simplest way possible. For this reason, many commercial websites now follow a fairly regular set of rules. For example, westerners tend to look at the top-left corner of a page first, so that’s where the company logo should go. And most users also expect to see a search function in the top right.

Of course, says Caudron, the other golden rule is to make sure that your web pages load quickly, otherwise your customers might not stick around long enough to make that coveted first impression. “That can be the difference between big business and no business,” he says.

One of the interesting features of this search engine is that it attempts to cluster results by topic. This seems to work quite well for medical conditions (e.g. “stroke”), but is less suitable for distinguishing words used in different contexts (e.g. gene names from words). The ranking also needs some work, and a search for “uniprot” results in “Did you mean uniprost”…