The Breakdown: Announcing, Nimble: A Razorifish report on publishing in the digital age. Rachel provides a description of the report that she wrote for Razorfish’s Media & Entertainment practice, with support from research partner Semantic Universe.

Last week I mentioned being busy. One of the things that has been keeping me occupied for the past several months is writing and producing a report called Nimble. It’s aimed at content producers that are moving from traditional media distribution to digital, and finding themselves facing new challenges.

Most magazines, newspapers, TV shows, etc. have a website at this point, but it doesn’t mean that they’re making the most of the digital experiences that they’re creating for their audience. The report looks at three major areas of interest to content companies – how they attact and retain their audience, how they deliver content across new channels, platforms, and devices, and how they remain profitable in the new digital economy.

The key is: Content needs to be free. Not necessarily free-of-charge, but free to be accessed wherever and whenever the consumer wants it. And to truly be free, content needs to be “Nimble.” Content becomes nimble by being well-structured and having meaningful metadata.

The report discusses the types of structure that can set content free, and how this approach will change the role of the editor, the way content companies make money, the way they deliver content, and the way they attract an audience. It also includes information about emerging technologies and tools that can help digital content publishers move into this nimble world.

Read or download the entire report at http://nimble.razorfish.com and follow us on Twitter (@NimbleRF) for interesting developments and updates. I’ll be presenting the report at the Semantic Technology Conference on June 23rd, and we’ll be doing a lot more with this material in the coming months.

The Breakdown: Look no further to solve all of your iTunes song tagging problems. Robert Stribley tells us how his review of iTunes reveals some interesting principles about metadata and tagging, and may help you get to the music you love most.

I’m organized but not overly so. I mean, I kept my books in alphabetical order as a kid, but I didn’t think of it as an early indicator of the career path I’d take. But I do appreciate order, so when I import a CD into iTunes and iTunes assigns metadata to my lovely new tunes via its Gracenotes system (which compiles user-generated and submitted data), it’s a relief to me, when the provided data has been entered with, you know, some semblance of order. Often, it isn’t. In reviewing this data over time, I noticed users make certain mistakes consistently when tagging their music. These mistakes then, reveal principles, and though they happen to apply to music in my iPod, in practice, they also apply to tagging other digital files.

For your consideration then:

1. Some tags are more important than others

In iTunes, the artist tag proves more important than other tags (album and genre do also), and how it’s completed can affect iTunes’ ability to filter and present your music effectively. The wrong date tag may not matter. A typo in the artist, album, or song field however, may mean you can’t find your music. A typo in the artist field also affects more files than misspelling a single song title. The consequences of errors in completing some fields are simply far greater than others. Specifically, when fields are tied into key functionality, such as sorting and filtering, creating flawed metadata within those fields blunts that helpful functionality.

2. Tag key fields consistently

Some fields like the song name allow you to tag each file differently. The artist and album fields don’t. For example, you should complete the artist field with the artist primarily associated with the work, and use another category to add additional artists, which happen to contribute to a single tune. In other words, tagging Mos Def’s recent song “History” as “Mos Def with Talib Kweli” screws up a sort on “Mos Def” in your iPod because your iPod now thinks “Mos Def” and “Mos Def with Talib Kweli” are two different artists. A solution is to add “(with Talib Kweli)” after the song title. Otherwise, I end up with one orphaned song in my iPod: If I select to play all songs by “Mos Def,” that song doesn’t even show up.

Similarly, tagging one album “R.E.M.” and another “REM” creates two different bands. Be consistent. Simple things like the use of the word “The” (“Cinematic Orchestra” or “The Cinematic Orchestra?) or an ampersand (“Antony & the Johnsons” or “Antony and the Johnsons”?), can throw a wrench in the works. If alternate spellings exist for a particular piece of information, you should decide which will be the primary spelling and enter that consistently.

3. Avoid meaningless tags

Completing the genre field with tags like “other,” “unknown,” “unclassifiable” and, arguably, even “alternative” provides little if any utility. You may as well leave the field blank. “Unclassifiable” sounds cute to the fan, who doesn’t want to pigeonhole their band, but what does the “unclassifiable” genre sound like as a playlist? Probably a pretty icky mélange. “Unclassifiable” may reflect one’s personal appreciation, but effectively, it’s no better a category than “miscellaneous.”

“Alternative” seems useless for different reasons. It originally referred to bands, who signed with non-mainstream labels, so “alternative” doesn’t necessarily refer to a well-defined sound or genre.  “Alternative” doesn’t mean two songs sound even remotely alike. Furthermore, bands which were once “alternative” may now be mainstream (Hello, Snow Patrol!). Wolfmother sounds enough like AC/DC that you may as well label both “rock.” Or “metal.” Or something more helpful than “alternative.”

4. Combine redundant tags

This principle overlaps with the point about consistency, but we should highlight the importance of both combining and distinguishing between categories – something that may apply to the same file or song. (For a related discussion, see Rachel Lovinger on splitting and lumping, too.) For instance, when I searched on “electronic” within iTunes, I found I had songs tagged with the following genres: electronic, electronica, Electronica/Dance and Dance/Electronic, Electronica & Dance and Rock/Electronic. That doesn’t even include electro, which I’d allow a separate genre for or other genres like downtempo, dubstep or trip-hop which would often fall under Electronic.

What a mess. I selected all of these and replaced them with “Electronic.” Could I have distinguished between Dance and Electronic? Perhaps, though, if they truly share the Electronic genre, I’m happy to go with that. That does bring us to our next point, though.

5. Distinguish between different tags

As important as it is to group things consistently, it’s also important to allow their distinctions. Trip-hop, glitch hop, dubstep and electro, for example, are all sub-genres, which might appear under electronic. They are all, also, arguably distinguishable. If you’re familiar with these genres, you’ll provide more sorting utility by labeling music with them, rather than simply placing them solely under an amorphous tag like “Electronic/Dance.” We’re spoiled with a wealth of musical variety in the 21^st century. May as well help other folks discover these rich veins of music.

Now, if I could just get Apple to add a separate column for “tags” (or at least allow multiple genres), then I could place songs under more than one genre or subgenre (electronic and dubstep) and filter them in different ways (dubstep is also chill is also electronic). Then I could tag a tune like Burial’s “Shell of Light” with all of these. Creating a playlist just ain’t the same, and it’s more difficult to create playlists when songs aren’t tagged correctly.  Besides, I use playlists to create groupings, which ignore genre (workout, Summer Party, NYC, road trip, romantic, etc). Finally, allowing for additional, well-crafted tags would allow me to better create new playlists on the fly.

Well, back to my iTunes. I have housecleaning to do.

*Links provided to songs and bands on Lala.com, purely for your listening enjoyment

In the aftermath of the failed plot to blow up a Northwest airlines jet bound for Detroit, the finger-pointing over what President Obama called a “systemic failure” has centered on an inability to connect the dots. The language used in several commentaries I’ve read about the plot jumped out to me as oddly familiar. Reporters and columnists were speaking about a failure to rationalize databases, a failure to collate clues, a failure by the National Counterterrorism Center -the nerve center and “fusion center of all fusion centers“- to identify a dangerous set of data adding up to an extreme threat.

It’s not often that stock-in-trade content strategy work takes on an air of national importance, but the failures that occurred and the peril it permitted have roots in common challenges that content strategists try to solve from project to project: structuring and tagging data so that meaningful and useful information can be extracted and acted upon.

The Breakdown: Our own Rachel Lovinger gave a presentation at the MIMA Summit about what can happen when you don’t have a clearly defined content strategy. She showed several examples of common problems, and talked about content best practices that could have helped avoid these problems. The Summit will be posting video of all the presentations soon (including great keynote talks by Jackie Huba and Seth Godin), but for now, explore Rachel’s slides above.

Is the writing on the wall for traditional and satellite radio? (image via Hyrck)

Pandora. It’s the most downloaded app for the iPhone. There are a whopping 27 million existing listeners. 10,000 new songs are added monthly. They have a growing roster of 70,000 artists. It’s the most popular radio station in virtually every major market across the nation. The average listening time is over 3 hours every day. So what’s Pandora’s secret?

It’s content strategy, of course. Pandora’s robust taxonomy and selective curatorial process are the two major building blocks to their service which now boasts adding 60,000 listeners a day, competing directly with Sirius XM and going head to head with broadcast radio.

Behold the Taxonomy

Begun in 2000, Pandora spent years building the Music Genome Project: a taxonomy-based music categorization platform. By identifying over 400 different characteristics of a piece of music, it categorizes and classifies all musical submissions. Your favorite songs might be tagged as having avant-garde leanings, a driving swing feel, a vamping harmony, a minor key tonality, emotional vocals, an outside piano solo or aggressive drumming. Give it thumbs up or thumbs down and Pandora begins algorithmically analyzing your likes and dislikes, serving you up engaging recommendations of music that have similar musical DNA. Pandora works as a personal music classifier instantly becoming your audio tastemaker. For good or bad.

Sifting Through the Submissions

Pandora has also honed its content curation process and this might be what makes it stand out from other online music services. According to founder Tim Westergren at a recent Town Hall at Razorfish in NYC, “The challenge isn’t [expanding] the size of the catalog, it’s not having the catalog get too big. Only 10% of the music submitted is actually ready for prime time.”

Their team of musicologists sifts through submissions to look for strictly high quality music that they deem worthy of inclusion in Pandora. Each song that makes the cut to be part of the Pandora library is evaluated for its musical characteristics. “We spend up to 30 minutes analyzing a song and assigning its unique descriptive characteristics,” says Westergren. It then lives in Pandora waiting to be served up to online fans.

Beyond the Browser

But will this marriage of musicology, taxonomy, content curation and user feedback aid in the demise of broadcast and satellite radio as we know it? The tide could be turning as Pandora offers an enticing alternative to the ‘pay for content’ model.

“Our goal is to have a billion listeners,” says Westergren. “This will be replacing the hours you spend on satellite radio. In some cases, we want all of your listening hours, whether it is an iPod, broadcast radio, or satellite. We think the most vulnerable is satellite radio right now. It’s tough competing with personalized free radio compared to paying $15 dollars a month. We’re pretty confident about that trade off.”

So how will Westergren increase adoption? It’s hard to say, but now with 1/3 of Pandora’s new listeners signing up every day through their mobile device it’s clear that he is strategizing beyond the browser and striving to integrate his product into users’ daily lives. But Pandora has its shortcomings, too.

Pandora vs. iTunes

Can a piece of art really be objectively judged solely by the sum of its parts? If your tastes are very specific, you will be interested in Pandora until it stops playing what you want—then you might be inclined to skip to the next song. A common pain point users have is the restriction of only 6 skips per hour. This is a minor inconvenience but a necessary limitation built into their current licensing model.

And how can Pandora effectively distinguish between artists who are identified for their original sound and less desirable ones that simply emulate that style? Pandora falls short as well if I want seamless uninterrupted listening on my morning subway commute to the office. The iPod is the logical choice, loaded with my favorite music, already with on-demand access. Will a user base armed with iPods continue to sacrifice control and reliability for free access to Pandora’s library? We’ll stay tuned to find out.

Don’t fall for this myth mis-categorization so quickly. (image via lairweb)

Sometimes people ask, “Can you teach me how to categorize things?” and I’m never quite sure how to approach that. How do you teach someone to do something that seems like second nature? I think everyone already has the basic, underlying skills required. At its core, categorization is about lumping similar things together into a group and splitting distinctly different things into separate groups. These are skills that nearly every child learns during early development.

You can see it as they learn to speak and make early mistakes. For example, when a young child first learns the word “bunny,” she quickly figures out that “bunny” represents multiple individual objects (unlike the word “mommy,” which she uses to refer to one unique person in her life). But for a brief time she will misapply the word bunny to anything white and fluffy. She is lumping in many things that aren’t bunnies, but through trial and error she will learn to split off the things that don’t qualify.

So do people lose the ability to do this? It’s a fundamental aspect of developmental psychology – I don’t think we’d be able to communicate effectively without this skill. My theory is that people just lack confidence that the categories they come up with will be effective and will make sense to others. My best piece of advice is this: You just need to find the right balance between lumping and splitting.

Too much lumping will result in categories that are vague and meaningless. Too much splitting will lead to fine distinctions that are confusing, and the number of categories will become unwieldy. There’s no “magic number,” but take a cue from the scale of the groups. Ideally each of the categories should contain a similar number of items. It doesn’t have to be exactly the same number, but you don’t want one category to have 5 items and another to have 500. If your organization scheme is out of balance like that, try splitting up the larger groups, and lumping some of the smaller, related groups together.

Netflix origami: extending the experience beyond the flick. (via)

Being able to predict human behavior is a real talent that deserves praise and rich rewards. Netflix agrees and since 2006 has been holding an ongoing competition to improve the accuracy of its movie recommendations to members, handing over one million dollars to the team who can deliver a 10% increase in the accuracy of its “world-class movie recommendation system” Cinematch. This, of course, will require a complex algorithm and an equally complex judging standard involving something called the RSME (root mean squared error) of a data set. Put simply, competitors are vying to predict how likely Customer Sarah is to give “Sleepless in Seattle” a 5-star rating.

Setting aside the quibble about the notion of an increase in accuracy (nearly accurate isn’t actually accurate is it?), competitors are closing in on the prize. As of the end of January, team “BellKor in BigChaos” is within reach of the seven-figure prize, posting a 9.63% improvement. To give Netflix credit, Cinematch isn’t bad. After I submitted ratings on a couple hundred movies, including a 5-star rating on The Coen brother’s movie “Fargo,” it told me that I’d also love their movies “Barton Fink” and “The Big Lebowski.” Amazing! It also told me I’d love several dozen other movies, some of which I certainly would (who wouldn’t love “Duck Soup” and “Vertigo”), others I would certainly not, in diminishing degrees. Netflix is confident enough about its predictive abilities that it offers a site area labeled “Movies You’ll Love.” Come-hither marketing language aside, I’m confident enough about my own predictive abilities that it would be 100% more accurate to label it “Movies You Might Love.”

I’m all for a service that could weed out the junk from a storehouse of 100,000 titles, but I wonder if so much effort and a complex RSME calculation is really needed. Wouldn’t it be simpler to rely on the more transparent workhorse trinity of taxonomy, tracking, and tagging? In short, if absolute accuracy is an impossible dream and being mostly on target is good enough, wouldn’t it make sense to just opt for a lo-fi method? This would require just marshalling the metadata (some are inherent and filter-like, such as title, director, actors), tracking the traffic (what does my history say that I’ve rented with the same director or actors?), and by all means incorporating the tagging, mine and yours (not only did I rent it, I liked it, and several thousand other people liked it, too). All in all, it would be a recommendation with a bit of the machine and a bit of the all-important human. This would quickly get me a reduced set of choices. I can take it from there. By going the lo-fi route, Netflix could redirect future prize money (not to disparage the brilliance of the competing teams) toward increasing the number of movies it allows users to stream, which, by the way, greatly reduces the risk and aggravation of making the wrong choice of movie upfront. Don’t like it? Stream another.

Netflix isn’t the only one trying to hone its ability to predict human behavior. To mention just a couple, last.fm has its own recommendation engine and so does iTunes. Like NetFlix, they’re both pretty good. At the moment, most of the public attention to recommendation engines is going to entertainment, but I’m wondering what Netflix’s ambitions have to say to the corporate model of document retrieval. If I’m faced with finding the gem in a pile of research and development documents, how can I find the equivalent of that perfect dark comedy with heaps of bloody carnage and wit, made by a Coen-esque brothers team and starring John Goodman? By limiting the number of relevant documents by a factor of ten or even a hundred through taxonomy, tracking, and tagging, I’m very likely to get that small set of worthy choices from which I can pluck the prizewinner.

Facets expose a range of data points for greater specificity. (via)

Recently a colleague asked for examples of faceted navigation that employed more than just lists of data points. She was, in particular, looking for unique or appealing interaction design using facets. There’s plenty of faceted navigation out there these days, but it is quite constrained to eCommerce sites and therein tied to data points for color, price, brand, and size.

Generally, the executions are straightforward, useful, and a bit humdrum. Other that the impressive MoodStream by the Getty Institute, the selection of sites building amazing interaction from facets is quite thin. Kayak put a stellar, if visually tame, implementation on their travel site. Etsy does nice things with interaction design, but does a terrible job connecting the facets. You can’t filter on multiple values, to the point where it can’t be considered faceted navigation. FoodPairing has a very intriguing graph display, but relationships are narrow and the interaction is generally constrained to the text links rather than the graph. What’s a facet lover to do?

Aside from the interaction design question, the investigation made me puzzle over why faceted navigation is being used almost exclusively for eCommerce. There is so much other content out there with multiple access points and relevance to more than a single idea. There are very few content items, such as a stock quote, that have only a single characteristic of interest. Why not use faceted navigation for manuals, instructions, news articles, and any other content you can think of? Do IAs, IxDs, and even CSs think it’s not relevant unless considering how to drill a consumer down to the right pair of women’s-brown-high heel-leather-Via Spiga-size 9? What if I’m looking for a webinar-mobile-managing employees-Laurence Peter-construction?

Is faceted navigation considered too complicated to construct? Too heavy a maintenance burden? Are there usability issues? Consumer/end user acceptance issues? Is it only appropriate when there is an enormous volume of content? Let’s talk about what makes us shy away from it and what makes us embrace it.

The unbearable Twitness of Diddy. (via)

Before I was a content strategist at Razorfish, I was a metadata librarian at People.com. And though I endured endless jokes about the taxonomies I created, I have to admit, I loved the uniqueness of the tags and the complexity of the celebrity relationships. It was the perfect place to create tags such as “messy breakup,” “bad hair,” or “Promises Treatment Centers.” It was the ultimate intersection of pop culture, creativity, SEO, and content strategy.

There was infinite technical and intellectual strategy work around how to deal with P. Diddy (who seems to be going by only Diddy now). As he continuously changed names, we had to create the ability for the content management system to understand aliases and synonyms (so the editors/taggers could search on the back-end for any of the names), and yet only display his current name to the user on the site. This was also helpful and relevant for celebrities with stage names, married names, etc.; it made sure that if you searched for “Diddy,” you got everything that only mentions Sean Combs – and a keyword search alone wouldn’t have been able to do that.

So now that we could pull all of the Diddy articles together regardless of his name, what if we wanted to dynamically pull in related content, such as a story about J.Lo.? (Why would Diddy be connected to J.Lo? Because they dated from 1998-2001.) Again, a keyword search wouldn’t know this; however, a semantic relationship tagging system would. So I started developing a framework for how celebrities are connected to each other: “has gone to rehab with”; “has had baby with”; “has been involved with” (this was my way around having to untag after breakups, as well as not discriminate between gender, length, or the seriousness of the relationship). These semantic connections now allowed Diddy to be connected to J.Lo, for Liv Tyler to be connected to Steven Tyler, and for Matt Damon to be connected to Ben Affleck.

And what is Diddy’s career, exactly? Musician? Producer? Fashion designer? We needed to assign each of these attributes equilaterally – particularly in an industry where almost everyone has a fashion line – without limiting each celebrity to just one parent class. (Will Smith is the probably the most complicated, in almost every category possible.) So yes, Paris, you too can be a musician as well as a fashion designer.

So though it is easy to dismiss the content surrounding celebrities as fluff, the complexity of their relationships and identities was the perfect place for me to be introduced to content strategy: there was a limitless amount of unique, overlapping content that needed to be classified, organized, and connected. Thank you, Hollywood.