Revisiting Pandora

I ran into Tom Conrad at Barcamp Block a few weeks ago, which reminded me to go check out Pandora again. I’d been an early adopter when they introduced it to the folks at the very first Barcamp, but accidentally stopped using it a while back when I changed out the computer in my office that I’d been running it on. I recently swapped in another system, and among other things I have it running Pandora again.

I like being able to launch a station with a single suggestion and get a few hours of “more stuff like that”, especially when it turns up something I haven’t heard before.

Here’s a sample “station” that I’m listening to at the moment, constructed starting with a track from Rick Braun, which has turned up selections from Larry Carlton, Jeff Golub, Joyce Cooling, and the Brecker Brothers on its own.

Pandora is now free

I spent a lot of time digging up new music a couple of months ago during the pre-launch period beta tests of the Pandora music service. I put together a list of interesting music that I found, and ended up purchasing a number of new albums, and put off signing up for their paid subscription service until I finished working through the new music. I thought the fee was OK ($12/quarter or $36/year) but I simply had too much other stuff to listen to, so it would have been wasted money until the backlog cleared a bit (all the CDs I found from listening to Pandora in the first place).

Given my experience (liked the service, liked the music, put off signing up temporarily when the fees started), and the opportunity for affiliate referral fees from Amazon and others, this move to a ad- and affiliate-supported service could end up generating more revenue in the end.

In addition to many new features (bookmarking, station editing, playlist improvements, etc.), Pandora v2.0 includes a free, ad-supported version. Listeners have the choice to subscribe and stay clear of ads, or use the free service, which will gradually incorporate advertising. What does this mean for you? You can now come back and listen to Pandora as much as you’d like for free–and all the stations you’ve created remain intact.

At a referral fee of 6% of sales, it would take around $50 of CD sales to directly replace the old subscription fee. However, many more users who would turn away if even a small payment were required might try using a free service. And Pandora is the sort of service that creates demand for new music that those “free” users might be happy to purchase from Amazon (or iTunes). I don’t know what their conversion rates look like, but if they look anything like my behavior, Pandora is far better off working on bringing in more music-loving users than trying to collect subscription fees.

See also:

Pandora launch, DRM, and media

Pandora officially launched this morning. It’s been over a week since I signed up for an account and I’m still using it. It’s specifically designed not to provide on-demand streams, but I’m getting the hang of steering Pandora into building playlists that have what I want, which is almost better than on-demand, since I don’t have to actually build the playlist myself.

What I really wish for is a sane way to make my personal digital media effectively (and legally) portable across my networked environment. Pandora will be providing $36/year streams of interesting-but-not-on-demand music, Rhapsody provides on-demand music subscriptions at $100/year, and iTunes Music Store provides downloadable purchases that may or may not work elsewhere on the network and won’t survive a computer transplant.

My general preference is to own the album. So I buy CDs, rip them onto the house server, then store the CD. This doesn’t work so well for iTMS. If I could get a reliable subscription service that provided the range of music that I’ve accumulated over the years and let me distribute content among the various client devices in our household, I’d be very interested.

With DRM and online distribution, I’m never sure I’m going to be able to put my music on some new device I get next year. Worse, I’m not even sure that the music I purchase will continue to work on the devices I already have. The short term future proofing is having a stack of physical CDs in a closet that can be re-ripped as needed. I also occasionally find myself looking to download a track that I have on vinyl LP that I haven’t ripped yet, since I’m generally unwilling to repurchase my entire collection on CD, (for those albums that are actually available in CD format). I’d be very happy for a subscription service that could effectively replace those albums.

Fred Wilson has a much nicer audio setup than most, and writes about his experience with iTunes and DRM.

We connected these servers to a multi-room audio system and we control them with a combination of crestron panels, java clients, and web browsers throughout our home

In the peer to peer world, with DRM working behind the scenes, we end up buying the music several times, and then can’t play it on every computer we own. That doesn’t make sense.

Peter Burrows wrote about his music purchasing experience with iTunes recently as well, and why he’s been using Rhapsody lately (reformatted computer, didn’t want to purchase music over again).

I had to wipe clean my PC and reinstall Windows upon my return (for totally unrelated reasons), but forgot to back-up my iTunes folder one last time before I did so. So when I got the PC back up and running and repopulated iTunes, I found that the album was no longer in my library. And since Apple only lets you download purchased music once, clicking on “Check for Purchased Music” didn’t do the trick, either.

…subscription services are another kind of user experience, that would appeal to many current customers and millions more.

At present, if you’re willing to live in an all-Apple or all-Microsoft universe, things can sort of work for now. I have a hard time accepting anyone’s DRM package as being the one true implementation, especially with some much interesting development going on around devices rather than desktops.

More on Pandora from Tom Conrad, TechCrunch, and my earlier post on Pandora.

For reference: Cory Doctorow’s talk on why DRM is broken (originally presented at Microsoft Research, June 17, 2004)

Pandora

I’ve spent a few days now playing with the prerelease “friends and family” version of Pandora, the “music discovery service” demoed by Tom Conrad at Barcamp last weekend.

Summary
Some quirks, but overall really good, and easy to get going. Unlike some other services, I’ve been running it most of the time I’ve been at my desk for the past few days.

Discussion
My personal taste in music is simultaneously eclectic and encyclopedic in some areas, yet with odd gaps. Using Pandora, I’ve been able to think of one or two songs, albums, or artists that’s representative of what I want to hear, and it will come up with a fairly decent playlist of similar tunes.

Although I’m finding that I could have theoretically constructed the playlists by hand, it’s really easy to try dialing in a tune or two until Pandora starts queuing up something like what I had in mind. The music discovery part seems to work reasonably well too, it’s turned up a couple of new artists for me to check out later.

On the Pandora blog, there’s a post with assorted user feedback, feature wish list, etc. Here’s some of what I’d like to see:

Wishlist

  • Playlist history (maybe with timestamps, like some of the radio stations provide), so I can go back and see what was playing a while ago.
  • Playlist lookahead (so I can see what Pandora is queuing, to help decide if I want to skip ahead)
  • Some mechanism for requeuing a past song in the future. I understand that at the moment, Pandora can’t provide a “backwards” function in the playlist, in order to avoid becoming an audio-on-demand service. On the other hand, having a method for indicating “I really liked this song and would like to hear it again” (or “I stepped away and mostly missed this song”) could be useful for the playlist queuing function. This may be handled by the “Guide Us” input form, not sure.
  • Music parameter template – since Pandora is building the playlist based on similarity to the starting tunes, I’d like to be able to see how it’s characterized the starting point.
  • Control over the parameter variation over time — I’ve let Pandora run for several hours at a time, and at times I’d like it to have wider variation over some aspect than others. For example, vary tempos gradually over several songs, but leave instrumentation and vocals more similar. Or vary instrumentation, but leave the tempo, echo, and bass / drums similar.
  • Some kind of clustering of characteristics for a given artist or album might be helpful. I get the impression that if you start with an album or artist, the starting “genome” is an average or perhaps a median of the entire collection. I get reasonable songs for a while if I enter something like “U2″ or “Lou Reed”, or “Lenny Kravitz”, but if I start off a channel with a specific song I will get very different results (as expected), but which never turn up otherwise (not entirely expected, since these all span a wide range of “sounds”).
  • Similarly, I might never want some combinations of characteristics to turn up on a given channel. So a way to specify the ranges or variances for a given “genome” parameter on a given channel would be handy.
  • A “time period” selector or bias might be helpful. This might not work well since there’s a lot of re-released material.
  • A progress bar and track info would be nice. Duration, artist, release info, link to iTunes, Amazon, etc
  • A way to stream the Pandora audio into devices on the local network, i.e. Roku and similar network players
  • A way to queue local audio data into the Pandora playlist, since I may have selections unavailable to Pandora
  • A community track rating function and/or message board, for promoting interesting discoveries among site users, and perhaps as data for improving the playlist generating function
  • Maybe a blacklisting function? Since the playlist is selected automatically by similarity, there can be interesting juxtapositions from a human listener’s point of view. I like that, but it might not work for everyone.

More data points:

  • A sample channel built using “Steely Dan” comes up with a reasonable start, but repeats tracks fairly regularly within an hour or so
  • A sample channel built using “Pat Metheny Group” is also reasonable, but repeats within an hour or so rather than moving to other albums.

These last points are easily fixed by using the “Guide Us” input form to select some additional starting points, but the playlist queuing function could probably use more latitude. I know the tracks are in the system, because I can use them as starting points as well, I just can’t get from “here” to “there” yet.

In addition to building playlists of music I know reasonably well, Pandora is turning out to be quite good at turning out electronica, techno, and club mixes, where I can throw in a couple of starting tracks and get back similar ones. I’ve already turned up a few tracks that I have heard, but didn’t know the artist or title. Since there’s often no artist, or the track is actually a DJ remix, Pandora provides a great way to find things. As a sample: starting with Gus Gus, Dirty Vegas, and Chemical Brothers turns up lots of similar, but different tracks.

Other early reviewers have mentioned Last.fm and Audioscrobbler. I ‘ve poked at these a little bit, but they’re geared more toward the social end and seem to require more upfront investment of effort. I think Pandora could ultimately benefit from the social functions, but it takes nearly zero time and effort to put together a very listenable channel or two. I’d probably find last.fm easier to use with something like Pandora spliced in as a selection filter, in addition to or instead of the user tags there.

Pandora is still in limited trial mode, but apparently I can invite 25 people from my trial account. Let me know if you’re interested!

If you’ve read this far, you should definitely check it out…

Update 08-26-2005 13:56 – In an e-mail to the prelaunch users, Pandora founder Tim Westergren announced that the service will be launched to the public next week. $36 for a full year of service, new users get a “short period” free, plus some changes based on user feedback. More at TechCrunch.

Update 11-10-2005 11:55 – Pandora is now free

Notes from Barcamp


I spent the whole day this past Saturday at Barcamp, arriving at 10am and not making it all the way out the door until almost 1:30am. I didn’t know any of the organizers beforehand, so it was nice to be met by someone (who turned out to be Andy Smith) keeping an eye out for new arrivals.

I’ve been working with so many either geographically dispersed or very buttoned down corporate teams lately that it’s been a long time since I’d spent this much time in non-stop, face-to-face, somewhat random yet unpredictably creative conversations with a bunch of just-do-it tech hackers. Spent the entire time there going from one interesting presentation to another, with several interesting discussions thrown in along the way.

IMG_3927 The Scheduling Wall

All the talks were informal and held in very close quarters, scheduled by signing up on the scheduling wall. In many ways, the strength and weakness of the Barcamp presentations was the short lead time for the event and the relative intimacy of the discussion spaces. This format put an emphasis on less structured presentations, by people who were comfortable enough with their topic that the slides weren’t the primary content. At the same time, having everyone literally elbow-to-elbow, sitting on the floor, an armslength or two away from the speakers, tends to eliminate random anonymous verbal potshots, and encourages actual conversational questions. Some of the presentations rapidly turned into something like topical roundtable discussions, which would have been hard to set up as such. (The Industry Darlings talk is a good example.)

Among my notes:

  • Tom Conrad’s demo of the Pandora (formerly Savage Beast) music discovery and streaming service, which is based on the Music Genome Project. I’ve signed up for a trial account, which I’m listening to right now. It’s coming up with pretty good selections so far. More on this at TechCrunch. Looks like Scoble likes it too.
  • Chris Messina gave a demo of the Flock browser, which extends the Mozilla platform to provide a lot of blogger- and social-software-friendly hooks. (review by Roland Tanglao here)
  • Demo of Mobido, by Mike Prince. It’s a social photo sharing service for mobile camera phones, as well as e-mailed images. Among other things, it’s could be used for people to other people who attended an event during or after the fact if they’re using the service. The service also includes provisions for anonymizing contact info.
  • Demo of a personal phone management system (forgot the name) by Brad Templeton. The general idea is to allow people to advertise their availability for making or receiving calls, and having the system set up calls when both parties are actually available, rather than having them play phone tag. The system is built on Asterisk, and uses Caller ID to invoke its rules, which may have some future problems based on…
  • Jake Appelbaum’s demo of phone insecurity and other security hacking, in which he walked us through the social engineering and general weaknesses in most cell phone systems, Paris Hilton’s Sidekick, and US airline security, and also demonstrated the Asterisk hack for Caller ID spoofing, which makes your calls appear to come from anyone you choose.
  • Nicholas Chim demoed The Personal Bee, an aggregator for building your own version of something like Google News. It appears to scan a collection of feeds to assign weights to “interesting” keywords, which it uses to build the page. More on this at TechCrunch.
  • Riana Pfeffercorn’s on buying and writing ads for search engine, and Google Adsense specifically, with a few bonus tips on Yahoo Paid Inclusion by Beau Lebens
  • Caught pieces of discussions on making AJAX-y applications faster, KaPing Yee’s presentation on improving web security (anti-phishing), a discussion on how to visualize the effects of social behavior to help save the world (resource consumption)
  • Other conversations with Bill Lazar, Ross Mayfield, Kevin Burton, Brendon Wilson, Rashmi Sinha, Wolfgang Zeglovitz, and many others.
  • Ran into a couple of other Koreans there: Eugene Eric Kim, and Min Jung Kim. Pleased to meet you.

Although there’s a constant background question of “how do you make money doing this”, the basic feel of the weekend was about sharing interesting ideas and work in progress with other people interested in making something new and better. These days, that attitude may be a bit old school, but it could be the cure for what ails Silicon Valley. And there’s apparently interest in organizing similar events elsewhere.

Lots of appreciation goes to Andy, Chris, Eris, Ryan, Ross, and all the Barcamp organizers, sponsors, and other contributors!

Update 08-27-2005 00:05 – Barcamp – The Video