Started off in a studio flat in Whitechapel. Matthew came over in 2005 as lead web developer.
Early Growth lessons
Don’t overextend – scale with your growth not before. Until you need the four man tent, stay with the two man tent.
Make sure revenue sources scale with increased usage, e.g. Google ads, user subscriptions
Involve users in your web application’s story, make sure that they know about your start up, know that you’re growing, and make them ivolved in the growth. If it’s a social network, growth should be a selfish aim, recruiting new users should improve their own experience. be as open as you can afford to be, it will pay off.
Late 2004, Audioscrobbler and Last.fm has separate sites. But site news was front and centre, and they talked candidly. Put development up there as part of the product. Started off with a donation model. Had server failures, followed by a flurry of people giving money to buy new servers. As soon as you pitch it as a feature set, it doesn’t really give you the opportunity to do that.
Openness and growth
In 2004, had a cool service but needed data. Rather than try to do it themselves, they created the Audioscrobbler Protocol 1.0, so that any developer could create an app that sent track data to their server. Once you had a profile, you should be able to access it via webservices For developers wanting to make a plug-in, they have to see a ’round trip’ and see their results helping users. Immediately had Winamp, iTunes, WMp, Amarok, plug-in. And dozens more.
Promote a community around your application. They had forums and news on the front page. People to hear bad news than no news. So they want to know that a disaster is happening rather than be cut out of the loop. They will be more likely to tolerate your growing pains then.
Showed their submission graphs to the users to that people could see what what happening. Users appreciated information. Appreciated transparency. When things went bad, they would parody their graphs.
People found the 404s funny, and took advice to go make a cup of tea literally and Flickr’d it.
Anil joined when they had moved to the new offices, 8 months ago. Process, growing a team, how do you protect product development when you’re rapidly growing a team. What happens when your product grows up, and your community grows. Company has doubled in size to 40 employees over last 6 – 8 months. What works when you are 6 hackers sitting round a table, doesn’t work when you’re at 40.
Need ways to co-operate. People trump processes, which become lightweight and get out of the way and let you get on with your work. If you show up, it looks like chaos, but have their own lightweight processes.
Got to start picking up the right tools – simplest tools that do the job, and customise them.
Four or five teams, need to radiate information across different channels in the company.
Use IRC. There are bots which feed information into the IRC channel, as well as having people having conversations.
– 15 million unique users a month
– plan for going global
– 10 languages in December
– need operations to support that – native speakers to track customer feedback and communications.
– affiliations should enhance the user experience, concrete end-user features when you’re talking to affiliates.
– harness critical mass
– embed your service in others, let people take your stuff and use it elsewhere
– open yourself up as a platform
Events feature, includes user generated content, affiliate, and is self-moderating.
Album art quilt, charts, etc. Allow users to export their identity, from your platform to others.
Have always been a platform, but needs to think about your service as a platform, have APIs, let people run with it, because that drives growth.
Last.tv, Snapp Radio,
Openness is a key to web platform.
Every page still says ‘powered by Audioscrobbler’, because the scrobbler information built everything else they can do with the platform. But scrobbling data is attention data. Normally when an app tracks your actions and sends it off to a server is called ‘spyware’ but think of Audioscrobbler as ‘myware: spying on yourself’.
Attention economy alters traditional meaning of ‘active user’. So some users are very active, others are more passive, but the passive users aren’t ‘freeloading’ because you can still get attention data from them. It’s not all about tagging tracks or being an active participant.
Events screen very popular. It looks at where you are and takes the user profile, list of what you listen to, and creates an events calendar for you.
Does the same thing with journals, and can give a tailored list of recommendations.
Those are some things you can do with attention data.
Fred Wilson, Future of Media, Nov 05 monetising attention, said:
1. Microchunk it, reduce content to the simplest form
2. Free it, put it out there without walls
3. Syndicate it, let anyone take it and run with it
4. Monetise, put the monetisation and tracking systems into the microchunk
How to monetise attention data
– Powerplay is sponsored airtime. In music industry has ‘payola’ which is sponsored but opaque buying of airtime, but Last.fm have transparent sponsored music. Can target people depending on their profile.
– create new metrics for reporting, how may people list, love, skip, ban
– no more CPMs, the ‘scrobble’ is the attention unit.
Tag cloud moderation
As soon as you open up tag clouds, Paris Hilton tags, not flattering. Should be open, and the tags are valid, but people were waging hate campaigns against specific artists, co-ordinated campaigns, and decided that it was tag spam, and an actual problem.
Can’t use blacklists – what you think about an artists counts, can’t censor people’s expression of their opinion. Attention data doesn’t like, so it knows what content can be listened to. Weight user tags by how much attention the user pays to the content, so if you listen to a song a lot, your tag is weighted more heavily. If you listen to Paris Hilton you have more of a say on what shows up on her tag cloud.
Attention data is a good filter for user generated content.
Future of last.fm
More: growth, streamable music, ambient findability – discover blogs and events more organically and contextually, more personalisation/things you can do with your data, need to get the data out in interesting ways
Less: fewer interfaces – need to streamline, reduce barriers to entry – cold start when you join where the service knows nothing about you so need to improve that, fewer gradients.
Questions from floor
What about spamming?
Rudimentary protection into the protocol, so for example you can’t listen to a 3 minute track 8 times in 5 minutes. Can identify zombie accounts, and take moves to try and counter attention spam.
We’re with Index, and it’s been very positive. Already had a product and platform before VCs came along. Hasn’t been a noticeable change in the way that the company operates or the atmosphere. Product hasn’t been influenced by them.
Should people have more control and ownership of their own attention data?
Looking at ways to do that, provide more ownership and control.
Are you going to open source your IRC hack?
Netcat, you should use it!