We had a hack day at Last.fm to develop ideas for better web radio products. I got together with @tdhooper @marekventur to design an interface for creating customised radio stations.

The interface consisted of two input boxes, one to put in what you wanted to play using a mix of tags, artists, and user profiles, and one that would exclude things you don’t like (e.g. Coldplay, Adele and “industrial house” in my case!). To get a flavour of the music the station would play, we added a quilt of artists and tags, that you could fine tune by deleting tiles from the quilt. Once you were happy, you could then play, save or share the station.

We called it “Radiomongr”. Unfortunately the prototype isn’t available for public viewing, but hopefully it can be used as a model for a releasable product at some point.

User interface design

Play state

I’m impressed by the stand-out of Pinterest’s cursive “p”, Beats Aduio circle, and the clarity of thisismyjam.com branding.

As well as researching music brands, I also find it’s good to have a set of brands to study and admire just because they’re cool.

This is getting close to the final design. I have been consciously trying to make it more graphic, and less illustrative. I have a tendency to get fancy and use effects and gradients.

A note for the geeks, the nebula in the background was inspired by a certain sci-fi drama. It’s the Nova of Madagon.

I’ve decide to use stars as a metaphor for the most promising artists, and the constellations they form represent festivals. Star brightness = Current hype (how much an artist’s audience has grown over the last 150 days).

The data

@omar711, Last.fm’s data team wizard, Phd, genius and general all round nice guy, anylised the listening data of all artists playing festivals this year, and measured their audience numbers over a period to see who were growing the fastest; this produced a strange decimal number we call a “hype score”.

I needed to translate these hype score into a meaningful graphic. As the figures alone mean nothing, only relative scale from artist to artist would be meaningful. However I encountered a problem. The scale ran from 0.0013 to 0.095, and I couldn’t figure out a decent way to represent the scale graphically. Omar came to the rescue with a simle math lesson…

The difference between normal and skewed distribution.

Essentially I had a “skewed distribution”, which Omar said “could be solved simply by using square root or log”. He looked at me expectantly as I gawped back vacantly. A quick Excel lesson later, and I had a more usable set of figures.

Last.fm festivals infographic "star" scale

Festival star makers

Omar then looked at festivals over a number of years to see how many artists grew in audience number directly following (and likely as a result of) playing at the festival. This produced a score for each festival on how often they “grew” new stars. Again, this was a decimal that needed translating into something understandable. I decided to take the top 10 and give each a rating out of 10.

My initial idea for the infographic was based around a map, however when the data was ready it turned out the festivals at the top of the list were mostly in the UK and Belgium.

This was for the Camden Crawl flyer. Goes out to 25K people. And there’s no cmd-z!