The desert island discs project is one of the highlights of my career. It does lots of things that I think a redesigned site should do, and working on it gave me the chance to work with some of the best content the BBC has (which lets face it, is pretty good :-)).
You can find out about the project here, but suffice to say it has been months in the cleaning, curating and building.
My particular role was to lead the information architecture for the BBC. As part of the external projects team (EPIC) for the BBC, I worked closely with magnetic north., a brilliant agency, and a brilliant team.[Props to @cazsotorrio,@raymosley , sanj and Adam Todd amongst others for excellent work!]
To start the project, I suggested we domain model the experience, which starts with considering ‘What is desert island discs?’.
The constituent parts of the desert island discs experience are (I believe):
1. The castaway – the guest being interviewed.
2. The presenter
3. The music
4. The artists who wrote and or performed the music.
5. The luxuries.
6. The book
7. The author of the chosen book.
These are the core concepts in the domain model – and each concept has a relationship to other concepts. By starting with ‘What things have we got?’ and ‘How are these things related to each other?’ we begin to articulate the data we have to gather and clean.
Having defined the concepts, we can then define the attributes for each concept. For example, a castaway can be a musician (which means they can have a musicbrainz identifier – useful for the rest of the BBC and elsewhere to identify them). A castaway can also have a wikipedia article about them – which can be easily used to identify that person, since it can be used to derive the dbpedia URL for that person. Having unambiguous identifiers which are webscale (ie used around the world, across the web) helps us syndicate this content if we choose to. It also helps us in the BBC, because these identifiers are used across the business. So, if the music team want to build deep links into the DID site from /music , they can, because they know exactly which artists are being interviewed or played on the programme.
Once we know what data we might want, we can also start to describe a full list of user stories (the things we want users to be able to do).
As part of this redesign, we were required to use much of the BBCs existing infrastructure. Since much of this is well built, it also encourages good practice. For example, we now have a page per episode of desert island discs. We also have a tracklist for each episode, which tells the audience which pieces of music were chosen. This tracklist data is then used on /music to populate parts of artist pages, which gives us many links into the desert island discs site, which benefits the DID site ‘Search Engine Optimisation’.
The data cleanup and assembly for DID took a long time, and also involved a ‘traditional’ IA piece of work – classification. Jo Attree (@romneymarsh) did a great job classifying the castaways and their luxuries. It’s perhaps worth pointing out that this isn’t a traditional library classification – it’s a classification to try to build a great user experience. So, at one point we had a very precise classification, which had several categories with only one castaway in them. Great, accurate, but not that great for the audience if they select a category to be shown only one result. This is what we might call ‘a cul-de-sac’ of an experience.
So, each of the categories on the site have many castaways in them by design. My rule of thumb is at least 5 castaways for each category – not a cul-de-sac to be found until you start really using the filters.
The results of this classification are that the related episodes part of a castaway page really works. The site data is stored in solr, and (while my technical knowledge isn’t expert) solr is matching the classification given to a castaway, and then prioritising those castaways who have audio you can listen to.
As part of the site redesign, you can also now use the URLs of the different pages to talk about the site. Castaway URLs (each castaway has a unique URL) and search URLs should deep link into interesting results (or maybe this one is more interesting) you can blog about!
I am also particularly pleased that this is a multiplatform site – one URL for one thing, delivered in a way that is appropriate for the device being used. So, each URL can be shared and browsed using a mobile phone or a PC.
I hope those of you with screenreaders find the site reasonable to use – if not let me know and Ill try and get it fixed.
Hopefully you find browsing the archive to be useable and fun, I hope we’ve done the great content justice.