I learnt html in Colombia in 1999, London in 2000 & got a job the next day, became an IA, went to the US 2002, wrote a book, Worked on quite a few horrific SAP intranet-type enterprisy products.became a consultant and worked mostly on consumer products, went to live in Belgium, and live now in Colombia.Today, I will discuss:What I think IA is.My experience and tricks on doing categories, structure, wireframes etc.Not: UI, …
Go over each briefly.Business model: currently working on a product where we know the real value, but haven’t decided the business model yet.User experience is not just usability, it’s all your interactions, and it is closely related to Brand.Speed is part of ux. Speed, especially online, is crucial. If your product is fast, almost all other UX issues get much easier.Price is part of UX.
An example of how I work on products.This is all product design, and part of my work as an IA. But narrow IA is just a subset of this.
In fact, I’m building a House in Medellin, and I’ve been struck by how similar the process of IA, and the process of building products is to actual Architecture. There are the black and white plans, that never entirely represent reality. There are the discussions. Even the delays are similar.
Let’s narrow it down.There are a number of elements that I think are important. None of them are about user interface or usability. CategoriesStructureMeaningThe Practice of IA.
Let’s talk about categories.
Definitions:An organization scheme is a method by which we make categories. For example: we can organize things by Place. Then categories could be “Bogota”, or “University”, or “USA”, etc. Or we can organize them by time. Or by Topic. Or by Color. Or by Age. A popular organization scheme is by Activity. By Target Audience. By Price. Etc.A category is abstract, it can have a really long name for example, like “everyone who is in attendance here today”.A label is the word(s) for that category. In our case, it could be “attendees”, or “2010 attendees”, etc.To come up with categorization schemes and categories, we have to look at 3 things: users, content and context.Users: for WHO are we organizing this? How do they think about this? Research techniques: freelisting, cardsorts.Content: WHAT are we organizing? What are the properties of this stuff? Research: painstaikingly become familiar with it. Content inventory.Context: WHY are we organizing this? What are the limits, goals, etc? Research: talk to stakeholders. Gather so-called “requirements”.Then we have to come up, iteratively, with categories. Tips:Use categories and labels that are easy to understand and straightforward. Not fancy labels.Use a consistent organization scheme, but not too consistent. Action-based (using verbs), for example, doesn’t usually work all the way. HR example: promote, approve, but you end up with a lot of “manage my x”.Audience-based is also tempting, but also often breaks down. “small/medium businesses, consumers, enterprises”. But what am I? Do I get a different deal as consumer? Etc.Multiple users will want multiple ways of accessing content.You should balance your scheme: not too wide, not too deep.Examples.
Let’s do an exercise, let’s organize the software that you all use. How would we do that?Ask the right questions. What why who
I want to share how you can examine the categorization of existing sites.Let’s look at a real example of categories and organization systems. Large companies like Oracle are always good to look at for classification examples because they are so large and complex that they have a lot of organization challenges. Other good examples to study: ibm.com Amazon.com The us gov main site. Etc.ENGAGE AUDIENCEGood: lots of ways into the informationNote: acquisitions etc.Countries could be organized by continent?Self-assigning audience is always problematic.“I want to” is example of problematic action-based. You get “manage x”… etc.Products/services/solutions is a very typical problematic way to divide things. Is java a product or a solution??
It turns out that categories exhibit some interesting properties, if we look really closely at them. These come from cognitive science, George Lakoff mostly, but turn out to be very useful for us.First of all: we tend to think of categories as buckets, but in reality, their borders are not hard. They’re fuzzy. There are almost always things that are clearly part of the category, and other things that aren’t, but often there are also things that are kind of hard to tell whether they’re part of the category. They sit in the fuzzy border.Second, and related to the above: categories exhibit the property of centrality. What does that mean? That things can be more in the center of a category, or more towards the border. It’s not in/out, it’s how central you are. So:Most categories have fuzzy borders: some things are not clearly in or out.Most categories have centrality: some things are more in the middle than others.Question to ask: what is the best example of this category? (That goes in the middle.)Note: “chair” is a category. Almost all nouns are categories. Proper names aren’t.Examples: Chair. Blog. Mother. Fruit (guayabano), Country. Office Document. Cars (tuktuk?). Books. So, When you categorize, don’t be surprised by this.
Aka the scary cat slide.Basic-levelness is another awesomely useful property of categories. It means that there are some categories that are more “basic” to our understanding, easier to understand, easier to parse cognitively, than others. These categories often have labels that are shorter in most language, and are easiest understood by children too.For example: “Cat” is a basic level category. It’s a short word, and every three year old can tell you what a cat is, and can categorize various types of cats together in the cat category. “Siamese cat” is a level below cat, it’s a specific type of cat, and it’s a category that is harder to grasp. You have to think a little more. “mamals” is a category that sits a few levels above “cat”, and again, it’s a longer word and it’s harder to understand.Other examples: furniture – table – coffeetable. Buildings – house – finca.How to recognize basic level categories?Ask kids.Count Google result counts. Mammals (27m) – cat (600m) – kitten (20m).Word is short.It’s in the middle of the hierarchy of concepts.Basic level categories are also easier to make a drawing of.When you’re creating categories, especially top-level categories, it’s often tempting to make them too generic. But it’s often better to keep them at the basic level, because it’s easier to interpret for the user.Expert users will have more knowledge of non-basic level categories, because they have learnt an agreed-upon system within their domain.
A compound category is a category that is made up of other categories. “Home and Garden”, for example. They are very, very useful, especially at the top of a taxonomy. You’ll see them used everywhere once you look for them.Why are they so useful? I actually figured it out only recently, I didn’t know the why, I had only observed that they were used a lot and worked well. We tend to make non—basic level toplevel categories, and this lets us use basic level wordsThere tends to be an overlap and expansion.The problem is that we tend to create fairly abstract high level categories, like “Consumer products”, or “Solutions”. These are used a lot inside enterprises. Everyone in the enterprise then learns that language. But your end users probably don’t. For example, let’s say we have a category where we have heaters and airco systems for homes, the ones for companies and factories are in another category. We could call this “Home Energy Systems”, not bad, but that’s a little abstract. If I’m looking for information about Airco’s, I might not see it. Better is “Home heating and cooling”. Even better is probably “Home Heating and Airco”.Amazon vs CostcoSo the trick is to combine BASIC LEVEL categories into COMPOUND categories.Another way to solve the problem of abstract, high-level categories is when most information falls within a well-understood category, but there are a few outliers which mean that you can’t use the well-understood category. Example from the us gov site: Food stamps and other nutrition assistance programs. The solution is to give the well known example, and add “& others”.
There is a type of category where it’s ok to use long labels, btw.Ad-hoc categories are categories that are created in a certain situation. Since they depend on that situation, they aren’t basic-level, and they often have long descriptions instead of short labels. For these, long labels are ok, in fact, it’s probably best to use a long descriptive label.“Everyone in attendance in this conference”.“The people who will be having drinks tonight.”
The everything else, or Other, category can easily be a sign of lack of imagination. The problem with it, of course, is that it doesn’t give an indication of what is in it, only that the stuff that’s in it is not in the other categories.But it can also be useful, and I’m actually a big fan: when there are a variety of things that just don’t fit anywhere, EE serves its purpose. The main guideline with EE is to place it deep in the hierarchy, not at the top. When you’ve drilled down and there are just no other options left, that’s when EE can be useful.
Look at examples to learn how to look at categories of other sites.Example: new gadget to put iphone in your car. Under cell phone accessories, or car tech?
In 1942, In "The Analytical Language of John Wilkins," Jorge Luis Borges (Argentina, 1899-1986) describes "a certain Chinese Encyclopedia," the Celestial Emporium of Benevolent Knowledge, in which it is written that animals are divided into
Dewey Decimal SystemMelvin Dewey, 19th century, religious family.Information, philosophy, Religion, Social Sciences, Language, Technology, Art, Literature, History Geography and biography.
Categories have a tendency to become invisible infrastructure, that becomes embedded/hardcoded in technology, and affects different groups.Categorization is boring.Becomes invisible/assumed. Hard to see.And embedded in databases, forms, etc., hard to change. “Invisible infrastructure”.But can have lots of effects. For example, if you’re categorized as “dependent” you pay less taxes. If your land is categorized as “protected” you can’t build a house. Especially in domains that affect a lot of people, like government.-----Geoffrey C. Bower & Susan Leigh Star researched the ICD, a long lived, global taxonomy, from a social science point of view.Published in 1999, Sorting Things Out describes the social aspects of classification. The authors played “IA archeologists” and dissected various large taxonomies. One of them is the ICD: the International Classification of Diseases: a classification of statistically relevant diseases.
Hispanics can be of any race (so you can choose Black and Hispanic). The Some Other Race category was introduced in the census 2000 questionaires, not originally part of the standard taxonomy.In 1990, “Other Race” was added, but the biggest change is that people can now choose more than one race. In the 1990 census, half a million people ignored the instructions and checked more than one box. Something had to be done. Imagine being a kid with parents of mixed race.Causes lots of lobbying & marches in Washington.These have been copied to Brazil.
Whenever users have to self-categorize, …
There’s much more to say about categories, and it’s a fascinating subject, but let’s move on to structure. Structure is what binds categories together.
The easiest way to give some structure to categories is to put them in a list. Civilization was built on lists. The ten commandments. Lists of cattle that I own, of people in my kingdom. Whenever you have less than, say, 100 items just put them in a list. I often see a lot of taxonomy being created around a fairly short list. If you don’t have a lot of items, there’s not much use for categories and taxonomy.One List is better than multiple lists, it is easier to get attention to it. Improve the one list, don’t add lists. When you have a list, you have order. Order is very important: the first items on a list get a lot more attention. By Newness is very useful, it focuses attention. Alphabetical is useful for known-item-searching mostly. By importance is very useful. The most successful lists (Facebook stream, Google results) use a lot of smarts in ordering, and use importance, which is hard but by definition useful.A sidenote on paging: paging isn’t good. People do scroll. When I’m making new sites these days, I avoid paging, I use a lot of endless scroll. Google images is a good example of endless scroll.
More is a way to prioritize lists. I like it a lot. It says that you’ve done the hard work for the user and thought about what’s important and what’s less important. It gives the user less options to worry about.
At some point, simple lists are not enough anymore, and we need to add sublists. For example, if you’re organizing countries, states and cities. Lists become trees, or taxonomies. Taxonomies are incredibly useful. A LOT of stuff in life quite naturally fits into taxonomies.But taxonomies also have their problems. One of the things everyone who tries to create taxonomies quickly runs in to is polyhierarchy. This is the problem of certain things belonging in two different places. If we put them in branch A, people looking for them in branch B won’t find them. But if we put them in both places, then it’s confusing: where am I when I’m looking at this?The easy solution is to keep everything living only in 1 place in the taxonomy, and use crosslinks. “See also”. The second problem with taxonomies is that there’s only 1 path to a certain item. Let’s say we’re classifying wine. Do we classify country > color, or color > country? Or color > country > cost?we need to move on to use a different type of structure, faceted classification.
Faceted browsing is an great way to explore large amounts of information, and I think it’s still underused. Every information architect should understand it intimately. One way to explain it: it’s just advanced search without the dropdowns. Another way to look at it: it’s letting the user find the information without worrying about which aspect of it they’ll want to search for first.This example is from epinions.com. The work was done by Peter Merholz as the main IA in 2001, and he introduced faceted classification to a lot of IA’s. It has stood the test of time very well, unlike tagclouds.Note how it shows you available options (and doesn’t show you non available options), and adjusts as you browse down. Also notice how every user creates their own unique taxonomy tree.To create a faceted classification system, the first thing is to figure out what facets are useful. Most things have a lot of facets, and some facets will be more useful to users than others.For example, let’s say we want to create a faceted classification of your photo library. Think of all the photos you have. What would some facets be?ASK AUDIENCE.So how would we determine which of these facets are most important? A good research method to figure out which facets are most important is the “10 questions game”. It works like this. ASK AUDIENCE.Now, let’s think about some facets for other topics. Who wants to suggest a topic?http://flamenco.berkeley.edu/Epinions.com
Faceted classification of Colombian food?
Go a little more abstract than faceted classification, and you get networks. We no longer organize things in lists, trees, or multiple trees, but we create a collection of nodes that have random connections between them. This way, we can express much more complicated relationships. For example, Peter is connected to Bogota.If we give those connections or relationships names, for example, Peter “has visited” Bogota, then we can express even more meaning. This is called a semantic network.A good example of a semantic network system are topicmaps. At topicmaps.org you can find much more information.The problem with networks is that they don’t really map very well to a good interface online. Faceted classification maps to a very easy to use UI, but networks don’t. I remember getting very excited about them when I learnt about topicmaps and the semantic web a few years ago, but this stuff never really seems to end in easy to use products or interfaces, so these days I am not very excited about them anymore.(The images are a social network on Livejournal from US users vs. Russia users.)
Metaphor is a way to think about something in terms of something else. For example, if I say “Let’s nail this spec down”, it’s a metaphor. It says that the spec is something that is moving, and that we have to keep it in one place. It says that, once it’s nailed, we can’t move it anymore without a lot of extra work. Etc. There’s a strong argument to be made that most of our thought works by metaphor. This is totally unconscious. For example, if I say, “let’s get this talk moving”. That uses a movement metaphor, which implies that there’s a destination to reach, a path to follow, and we have to get somewhere. If I say: let’s dive deeper into this. That uses a diving metaphor. It says there’s depth to this topic, etc. Online, we use a LOT of metaphor. For example, a website is a “place”. “I am going to”. “We have x visitors”. A “link” is a chain, that binds things together. Etc.Metaphor in a UI makes it easier for users to understand what’s happening without having to explain it to them.Let’s look at metaphors used in gmail.“Archive”. -> it’s not deleted, but it’s in a dusty place where we can find it with effort.Labels: we can attach multiple labels to an item. They help us to find them. I can write what I want on these labels.Starred: a star is shiny. Children get stars when they behave well. A star is something special. If I add a star to something, it is special. This is actually a metaphor that is quite generic. Outlook uses colored flags. Mail. We know physical mail. We can send it to someone. They have an “address”. Trash. When we throw something away, it goes to the trash. We don’t need to look there, normally. ASK AUDIENCE: what other metaphors?
Let’s talk about the practice of IA. Of all the IA documentation that I’ve created over the years, the only one that has survived until today is the wireframe. I don’t make wireframes anymore, I rarely make functional specifications. I almost never have to write research reports anymore, thank god. My IA practice basically consists of thinking (which is doing research, sketching and iterating), and talking (with wireframes). The wireframe isn’t the end result, the conversation is.I’ve become much better about treating wires, and any documentation, as throwaway products. Example: meeting where 2 weeks of wires where discarded within minutes. That’s fine.
They’re used by multiple groups of people, that otherwise don’t have good ways of communicating. The whole idea of the wireframe is that it can be used in discussions with clients, visual designers and programmers alike. It’s an object that bridges communities, in a way that no functional specification or page description diagram ever can.And yes, that comes with some ambiguity and sometimes confusion about who owns what, but that’s the whole point. Remove the ambiguity, and you remove the usefulness.Describe example and describe team: business, tech, design.When the wires aren’t detailed enough: solution is to increase communication, not detail.
Who here uses Visio for wireframes? If +30%, I think we have time for a few Visio tips.Use the drawing explorer window to browse through pages. Much faster than the dodgy bar at the bottom.CTRL-K to add a link to an object. Click on Sub-Adress BROWSE and choose one of your pages to link to. This way, you can export clickable prototypes with SAVE AS WebPage, if your audience likes demos. For sharing, I still use PDFs. Print-As PFD works for me, CutePDF is free.CTRL-G to group things is your friend. Group and then copy&paste.I often search for “icon x” in Google and then just copy that in my wires.Background pages are your friends to avoid repetition. File > Page Setup > Page Properties to change them.
Repetition in documentation is a maintenance nightmare, because every time something changes, you have to update it in multiple places. So I’ve learnt to avoid repetition at all cost. Because if it becomes hard to maintain, then one of two things will happen: Your document will become out of date and teams will stop using it.Or worse: Teams will slow down the amount of changes they make to accommodate the limitations of your process.So the goal is: only specify each element once. The easiest way to do that in wireframes is to Only wireframe templates.Use modular system to specify repeated page elements.Pro tips:Find out who the audience is for the docs and talk to them before deciding on a format.
There’s a natural tendency to try to answer every possible question about the system in the spec. This is a mistake.Over time, I seem to have gravitated towards an average spec length of 50-ish pages. I suspect that part of this is that a spec needs to be maintainable and reviewable. Anything over 50 pages becomes a nightmare to maintain and/or review.I also remove stuff from a living spec as the project progresses. If the team knows what the spec tries to explain, then it doesn’t need to be in the spec anymore.Short specs come with a cost. There are things that aren’t explained in there. That scares people, but that is actually a good thing, depending on your environment. I suspect that long specs actually CAUSE teams to communicate less and therefore become more dysfunctional.Pro tips: Please, put teams in same room. Different rooms -> communication goes down. Different floors -> worse. Different buildings -> horrible.Use simple tools for your docs. This will encourage keeping them short. There’s a number of new ones right now, http://balsamiq.com/ and https://gomockingbird.com/ are pretty good, focused on keeping it simple and collaborating.
Finally, and on the same theme: don’t treat user research as a deliverable, if you can avoid it. And if you do, keep it short, with screenshots, user quotes, prioritized recommendations, keep it short.Bad way: set up research, deliver report.Good way: get everyone to observe users, discussion afterwards, no report.
Ask audience: who works in e-gov?Categorization (and metadata) creates a categorization burden. It’s important to be aware of this, and to look for ways to diminish this. The more people have to fill in, the more work it will be, and the more they will tend to create workarounds. You can oblige people to complete something (make it required), but you can’t oblige them to fill in the correct thing. Examples of workarounds: let someone else (the bank employee) fill it in for you. Leave things empty. Fill in made-up phone numbers or other made-up data. Examples:in the USA the Paperwork Reduction Act of 1995 was created to make agencies aware of the fact that filling in stuff is a burden. You may have seen this on immigration forms when flying to the US, it’s at the bottom.The Best guess example here is particularly elegant, from Google Health.
The more specific you make the meaning of your product, the more immediate value will provide, but the narrower its application will be. Let’s look at forms, as an example. You an offer a generic forms product, as Wufoo does, where people can easily create online forms. That’s fairly generic, there’s not much meaning in there. Or you can offer, as they also do, more specific forms, like a contact form, or a registration form, etc. Once you do that, you’ve added more semantics to your product, and it becomes more immediately useful for the people looking for a contact form. But it becomes less applicable for other people.You see this tradeoff often being made by what are called “architecture astronauts”. “Everything is a node”, they say, and you get so much power there. Yes, but things also become much more complex.
Everything I know about Information Architecture (mostly categorization) in 90 minutes
Everything I Know About
Information Architecture In Less
Than 90 Minutes
Peter Van Dijck
1. those that belong to the Emperor,
2. embalmed ones,
3. those that are trained,
4. suckling pigs,
6. fabulous ones,
7. stray dogs,
8. those included in the present classification,
9. those that tremble as if they were mad,
10. innumerable ones,
11. those drawn with a very fine camelhair brush,
13. those that have just broken a flower vase,
14. those that from a long way off look like flies.
• 210 Natural theology
• 220 Bible
• 230 Christian Theology
• 240 Christian moral & devotional theology
• 250 Christian orders & local churches
• 260 Christian social theology
• 270 Christian church history
• 280 Christian denominations & sects
• 290 Other and comparative religions
Categories become infrastructure that
affect different groups
Please choose your race (one or more):
- American Indian or Alaska Native
- Black or African American
- Native Hawaiian or Other Pacific Islander
- Some Other Race
Please choose your ethnicity (only one):
Please choose your race (only one):
- American Indian and Alaskan Native
- Asian and Pacific Islander
Please choose your ethnicity (only one):