MetaBrainz:Annual Report/2006
Introduction
2006 was the MetaBrainz Foundation's first interesting year of doing business and it is time to take a quantitative look at the business practices of the Foundation. The MetaBrainz Foundation was established in October of 2004 and it wasn't until 2006 that the foundation had significant income from sources other than donations. Even though the Foundation earned a modest $42,312.93 in 2006 (as compared to millions of dollars over at the Mozilla Foundation) the trends over the year provide a glance at things to come.
In 2006 MusicBrainz grew significantly as it moved to a new data center in San Luis Obispo and expanded its collection of servers from three to twelve. MusicBrainz also gained quite a bit of traction as it unveiled a batch of new features, and MusicIP as its new partner in the acoustic fingerprinting business. Traffic to the site continually increased and the database grew at a consistent rate throughout 2006, even as a rift in the MusicBrainz community appeared in the summer. Disputes between developers and the community caused a fair amount of upheaval that culminated in the dismissal of one of the developers. In the fall business returned to normal as the community renewed its focus on building a world class music encyclopedia and improving features on all levels of the project.
The finances of the Foundation received a huge boost when Google donated $15,000, which allowed for a small salary for the Executive Director. New business relationships with the British Broadcasting Company and the emergence of the tagger affiliate program in the latter half of 2006 provided validation that the MetaBrainz Foundation is on the right track to becoming a viable open business.
MetaBrainz as an open business
MusicBrainz and MetaBrainz are deeply committed to the openness of its data, software and finances. In 2006 it became apparent that MusicBrainz is starting to capitalize on its data and software assets and enabling others to capitalize on these open assets. With these accomplishments, it appears that the MetaBrainz Foundation and its supporters are lending credibility to its open business model.
The Open Business project defines an open business as:
.. entrepreneurial ideas which are built around openness, free services and free access.
The Foundation certainly fits the description of an open business, since its finances are open, the services free and the access to its data ensured by the Public Domain and Creative Commons licenses. In 2006 it became apparent not only that the MetaBrainz Foundation was beginning to capitalize on these assets, but third parties working with the Foundation were participating to earn more money than the Foundation itself.
In early August the MetaBrainz Foundation entered into its first affiliate agreement with Matthias Kunter, the author of the Magic MP3 tagger. In exchange for adding prominent links to his tagger site to MusicBrainz, Matthias agreed to pay the Foundation €250 a month or 10% of the registration income for this tagger, whichever amount is greater. Matthias' payment to the Foundation in December amounted to a respectable €563.29! Covering more than half of our monthly hosting costs, Matthias' payments have become our second largest continuous revenue source in 2006. Simple math tells us that Matthias is earning in excess of €5000 a month, which drastically exceeds the income of the Foundation itself.
A successful open business should foster the creation of other businesses to form an active ecosystem of companies that rely on each other. Matthias has proven that MusicBrainz is enabling others to create new businesses that were previously not possible. And now, at the beginning of 2007 two more authors have expressed interest in participating in the tagger affiliate program. This will allow the Foundation to set the goal to have the day-to-day costs of operating the service be covered by donations from the community and affiliate income.
Year in Review
2006 was a busy year for MusicBrainz.
Ramping up and moving to San Luis Obispo
At the beginning of 2006 MusicBrainz was running on three severely overloaded servers at the Community Colocation Project (CCCP) in Fremont, CA. At this point it had become clear that more machines were required in short order and that the CCCP would not be able to provide the services that a growing project like MusicBrainz required.
In January the search for more machines and a new hosting company were started. Progress came very quickly when Minor Ventures donated two Dell servers. In February the Foundation signed a contract to start hosting its servers in San Luis Obispo at the well renowned Digital West Networks. Renting 10Us of rack space MusicBrainz today has 6 computers in active service and one spare server to cover hardware failures.
SWSX brings about relationship with MusicIP
MusicIP, which was still Predixis at the start of 2006, agreed to provide MusicBrainz and its community free acoustic fingerprint services as part of their new MusicDNS service. MusicBrainz in turn integrated the MusicDNS services into its upcoming Picard music tagging application. In exchange for these services MusicIP started using MusicBrainz' live data-feed to integrate MusicBrainz data into the MusicDNS service. MusicIP customers can now choose MusicBrainz metadata as an optional add-on service for MusicDNS with MusicIP providing MetaBrainz with a 10% revenue share for the add-on service.
As part of the agreement MusicIP also sponsored Robert Kaye's time to allow him to work on the soon-to-be open sourced MusicDNS client as well as developing a number of new features in MusicBrainz. These new features would aid MusicBrainz in providing more scalable services in anticipation of upcoming growth. During this sponsored development time a new Lucene full text search, a new XML Web Service and the MusicDNS/PUID services were integrated into MusicBrainz. This new partnership was announced at South by Southwest (SXSW) music conference in Austin in March.
Improved web interface and a great dispute
The relatively quiet post SXSW period allowed MusicBrainz to clean up various loose ends left over from the hectic launch at SXSW. During this time Stefan Kestenholz worked on a significant update to the MusicBrainz web site greatly improving the standards compliance of the site. The overall look and feel of the site was improved and a number of new features that made life easier for MusicBrainz users were developed by Stefan.
In the weeks leading up to the release of the new site MusicBrainz was headed for its first serious community shake-up. After having become the de-facto server maintainer Stefan was continually involved in heated discussion between himself and members of the community. The overall communication process between the developers and the community broke down as tensions continually increased and tempers flared in all aspects surrounding MusicBrainz server development.
After weeks of unfruitful discussions on how to fix the broken development process and to resolve community tension, Robert Kaye took the unpopular step of removing Stefan as the de-facto server developer. After this event, various members of the MusicBrainz community got involved and helped out to extract the important lessons from the dispute and how to prevent these situations from escalating again. (The great dispute covers many aspects of online communities and social interactions that go beyond the scope of this annual report. If you'd like to find out more about this dispute, please read our "GreatDispute" wiki page).
Fall: Getting back to business
During the fall of 2006 the MusicBrainz community quieted down and focused on getting back to work. While the server development moves at a slower pace today, the community has been relatively conflict free since the late summer. In December one more server update was made to roll out a number of bug fixes and to smooth out some problems left over from the turbulent improved look and feel release.
In December John Buckman of Magnatune joined the MetaBrainz board of directors, replacing the perennially overloaded Joi Ito. Finally the year closed out with Santa Claus delivering a powerful new server from Sun Microsystems. (see section [#servers Servers and hosting] for more details on this)
Project finances
Profit and Loss
For 2006 MetaBrainz took in $41,312.93 and spent $34,973.98, for a total profit of $6,338.95. (non-profits can still earn a profit, but that profit must be reinvested into the company, not paid out to its officers and directors.) A detailed break down of where the income came from and where it went to is shown below:
| Income | |
|---|---|
| Direct donations | $ 15,740.00 |
| PayPal donations | $ 10,345.01 |
| Live data-feed licenses | $ 10,344.92 |
| Tagger affiliates | $ 2,864.06 |
| Data royalties | $ 1,500.00 |
| Bank interest | $ 218.14 |
| Amazon affiliate fees | $ 200.80 |
| CC data licenses | $ 100.00 |
| Total | $ 41,312.93 |
| Expenses | |
|---|---|
| Salaries | $ 18,573.37 |
| Hosting | $ 7,364.18 |
| Hardware | $ 3,421.49 |
| Office rent | $ 2,400.00 |
| Travel | $ 1,123.60 |
| PayPal fees | $ 790.24 |
| Marketing | $ 608.44 |
| Software | $ 180.45 |
| Entertainment (meals) | $ 160.80 |
| Gifts | $ 132.00 |
| Domain registrations | $ 122.71 |
| Supplies | $ 76.70 |
| Filing fees | $ 20.00 |
| Total: | $ 34,973.98 |
From Profit & Loss you can see a few interesting things:
- It cost $10,785.67 to run the MusicBrainz service last year -- this includes cost for hardware and hosting.
- PayPal donations, which represent end-user support for MusicBrainz, came to $10,345.01. The Foundation aims to have end-user donations pay for the actual cost of running the service, and considering that $740.00 of the direct donations can also be considered end-user donations, the end-users did in fact pay for slightly more than what it cost to host MusicBrainz last year.
- As it is with a lot of companies, salaries accounted for the largest slice of expenses. The Foundation paid $18,573.37 in salaries to Robert Kaye in the last year, which is far below a reasonable salary level for an Engineer/Executive Director in California.
- $11,944.92 was earned licensing the MusicBrainz data for commercial use, which represents 28.9% of the total income.
In an attempt to measure the overall effectiveness of the Foundation in converting donations from its users into usable services, we would like to offer the following metrics: In 2006, the Foundation spent $10,758.67 on hosting and hardware costs and served out 614 million web hits and 352 million web service hits. Calculating a cost per hit, we find that we spent $17.56 per one million web hits and $30.58 per one million web service hits. While these numbers may not be significant today, MetaBrainz expect that these numbers will become more interesting in subsequent annual reports.
Balance Sheet
The balance sheet for 2006 shows that the MetaBrainz Foundation closed out the year with retained earnings of $7,984.50, net income of $6,338.95 and total cash assets of $14,423.45.
Donations
Donations per month to the Foundation were highly variable. The chart below shows the total donations per month for 2006 and only includes donations that were less than $1000:
| Top Donations | |
|---|---|
| Google Inc. | $ 15000.00 |
| Matthew Mullenweg | $ 1000.00 |
| Wieslaw Syposz | $ 750.00 |
| MusicIP Corp. | $ 500.00 |
| Cory Doctorow | $ 500.00 |
| Mark Bogani | $ 250.00 |
| Alexander Dupuy | $ 150.00 |
| Duane Harkness | $ 100.00 |
| Edward Rustin | $ 100.00 |
| Matt J McInvale | $ 51.23 |
Larger donations such as the $15,000 Google donation are not shown in order to highlight the variability of the rate of donations. The largest donations of 2006 (of 651 total) are shown on the right above. Aside from cash donations and in-kind donations of server hardware a number of people made in-kind chocolate donations: A total of 18 chocolate bars were donated from 4 different countries.
Database Statistics
| Database Statistics | |
|---|---|
| Artists | 283,856 |
| Releases (albums) | 450,136 |
| Tracks | 5,374,216 |
| Advanced relationship links | 371,376 |
| Disc IDs | 179,201 |
| PUIDs (MusicIP IDs) | 1,727,905 |
| TRM IDs (Relatable IDs) | 11,597,996 |
| Edits (changes to database) | 4,201,501 |
| Editors | 281,714 |
Over the course of the year we collected a number of statistics about MusicBrainz.
Database size
At the beginning of the year the MusicBrainz databased contained no PUIDs (MusicIP acoustic fingerprints). In March MusicBrainz launched its PUID support with 800 thousand PUIDs in the database, and nine months later the number of PUIDs has more than doubled to 1.7 million PUIDs. It is clear that the MusicBrainz community has embraced the MusicDNS PUIDs! MusicBrainz also crossed the threshold of a quarter million registered users and five million tracks cataloged and a stunning 4 million changes to the database since the project's inception.
Traffic
MusicBrainz served out 614 million web hits, 283 million RDF web service hits and 69 million XML web service hits in 2006. Throughout the year the MusicBrainz traffic generally increased with the exception of January when MusicBrainz was recovering from having the MusicBrainz Classic Tagger be named one of the top audio products of 2005 by Wired Magazine.
The new XML web service starts picking up much more traffic towards the end of the year, and in December MusicBrainz nearly served out as many XML web service hits as RDF web service hits. This clearly shows that developers who are using MusicBrainz prefer the new XML/REST web service over the aging RDF web service.
In preparing these statistics it has become clear that MusicBrainz needs better usage tracking software to get a more detailed insight into its traffic. Webalizer, which calculated the traffic for MusicBrainz in 2006, has a number of shortcomings including the lack of a good way of representing page views compared to raw hits. Unfortunately MusicBrainz does not have good page view statistics for 2006. However, in early 2007 MusicBrainz started using Google Analytics for its future traffic analysis in hopes of presenting more meaningful numbers in the 2007 annual report.
Top Editors and Voters
MusicBrainz depends on its lively community to enter and maintain this vast body of data. 2006 proved once again that the MusicBrainz community consists of an amazing collection of people. Mudcrow, our top editor and voter (peer reviewer) of 2006 made an astounding 100,998 changes to the database and voted on 38,347 edits from other editors! Thank you for all of your efforts mudcrow!
| Top Editors | ||||
|---|---|---|---|---|
| 1. mudcrow | 100,998 | 14. inhouseuk | 12,669 | |
| 2. drsaunde | 85,533 | 15. BeenInsane | 12,226 | |
| 3. zout | 64,737 | 16. cooperaa | 11,990 | |
| 4. Joan Whittaker | 49,531 | 17. dj_empirical | 11,952 | |
| 5. teleguise | 35,329 | 18. mo | 11,924 | |
| 6. ojnkpjg | 31,057 | 19. krazykiwi | 11,539 | |
| 7. dmppanda | 29,487 | 20. Kerensky97 | 11,166 | |
| 8. Schika | 29,194 | 21. lukz | 9,922 | |
| 9. MrH | 28,162 | 22. BrianG | 9,851 | |
| 10.keschte | 25,263 | 23. Dainis | 9,799 | |
| 11.Shepard | 20,942 | 24. mcymd | 9,399 | |
| 12.bench12345 | 13,284 | 25. chancey | 8,977 | |
| 13.natenobility | 12,827 | |||
| Top Voters | ||||
|---|---|---|---|---|
| 1. mudcrow | 38,347 | 14. krazykiwi | 9,235 | |
| 2. pankkake | 36,052 | 15. lukz | 8,749 | |
| 3. keschte | 23,541 | 16. nikki | 7,602 | |
| 4. teleguise | 22,271 | 17. cooperaa | 6,989 | |
| 5. inhouseuk | 15,104 | 18. closedmouth | 6,827 | |
| 6. Nyght | 12,297 | 19. 0moikane | 6,687 | |
| 7. rowaasr13 | 12,269 | 20. clutcher2 | 6,242 | |
| 8. drsaunde | 12,171 | 21. Shepard | 6,002 | |
| 9. zout | 10,937 | 22. battye | 5,784 | |
| 10. MrH | 10,649 | 23. dmppanda | 5,073 | |
| 11. ojnkpjg | 10,522 | 24. jellyfish | 4,924 | |
| 12. mo | 10,308 | 25. chancey | 4,770 | |
| 13. BrianG | 9,697 | |||
New users
MusicBrainz accepted new users at a fairly consistent rate over 2006, usually ranging between 7,000 and 9,000 new users per month.
Since MusicBrainz is heavily dependent on its user community it is also important that the community not grow too fast. A growth rate of 10,000 new users per month seems to work fairly well.
Servers and hosting
Over the course of the year, MusicBrainz grew from three machines, to a current complement of twelve machines. The cast of characters that served out over 600 million web hits includes:
- stimpy - the primary web server
- dexter - the primary database server
- blik - the search server
- scooby - the catch all services server: bug tracking, forums, mailing list, etc
- catbus - a replicated database slave handling web service database queries
- bender - our TRM acoustic fingerprint server
- lingling - a cold spare machine for hardware failures of other servers
- hefty - soon to be the UK mirror of MusicBrainz
- zim - semi-retired test and development machine
To bring our count up to twelve servers, Sun Microsystems donated a powerful database server and an anonymous donor donated two 1U, dual processor servers. All of these machines are still unnamed and waiting to be rotated into service. The Foundation is proud to be on first name basis with the machines that do all the work and fosters the ethical treatment of computers; the Foundation refuses to apply cold numbers as names of its machines!
Anticipating 2007
While 2006 presented a number of interesting challenges for the MetaBrainz Foundation, a lot of progress was made in scaling the service to handle more users and to round out the feature set of MusicBrainz. Looking forward to 2007, the Foundation will continue to take steps towards self sufficiency while adding one or more developers to paid staff. MusicBrainz will continue to develop its emerging Tagger Affiliate program and work to ensure that income from tagger affiliates and end-users covers the rising operational costs of the project.
The lack of active server development is the most significant challenge that the Foundation faces for 2007. To solve this problem more effort needs to be expended to sell more licenses for the MusicBrainz data and to forge more relationships with partners who are willing to support the development of new features in MusicBrainz. The overall goal for the Foundation in this respect is to hire at least one full time server developer in the course of the year and start working towards the implementation of a new database schema that drastically increases the information that MusicBrainz can collect.
Words of appreciation
The Foundation would like to thank all of its users, developers and supporters for supporting the project in 2006!
A special thanks also goes out to all the supporters behind the scenes that may not get much visibility on the MusicBrainz/MetaBrainz sites. The Foundation would like to thank Dan Appleman, our contracts pro-bono attorney with Heller Ehrman, LLP, Ed Cavazos, our pro-bono intellectual property attorney with Fish & Richardson, P.C., Matthew Dunn and his excellent team at MusicIP, Chris DiBona of Google, and Paul Lamere with Sun Research, and our board of directors: Dan Brickley, John Buckman, Cory Doctorow, Robert Kaye and Brian Zisk.
