Difference between revisions of "Metrics/License statistics"

From Creative Commons
Jump to: navigation, search
(Baseline numbers from specific collections)
(Undo revision 97326 by Zelsa (talk))
 
(38 intermediate revisions by 9 users not shown)
Line 1: Line 1:
 +
''See [[metrics]] for a broader discussion of CC adoption and impact.''
 +
 
=Caveats=
 
=Caveats=
  
Estimating license adoption is a ''very'' inexact science. There is no authoritative source and we neither control nor have inside knowledge of the construction and volatility of the most comprehensive sources -- web search engines. From 2003 through 2005 we relied primarily on Yahoo! link: queries (Google's link: operator obtains ''very'' incomplete results). In late 2005 Google added CC queries via their advanced search page and also via their API (Yahoo! also allows this, but in this case Google's results seem more comprehensive).
+
Estimating license adoption is a ''very'' inexact science. There is no authoritative source and we neither control nor have inside knowledge of the construction and volatility of the most comprehensive sources -- web search engines -- primarily via Yahoo! link: queries (Google's link: operator obtains ''very'' incomplete results).
 +
 
 +
'''NOTE:''' Currently the best analysis of the data similar to that discussed below (based on a snapshot  independently gathered in January, 2007 may be found in Giorgos Cheliotis' [http://hoikoinoi.wordpress.com/2007/07/02/cc-stats/ presentation on CC statistics] from June, 2007.
 +
 
 +
= Raw data =
 +
 
 +
Creative Commons has irregularly run programs that collect estimated total results from search engine <code>link:{license_uri}</code> queries and queries filtered by license property (Yahoo! and Google advanced [[CcSearch|search]] support filtering by license).
 +
 
 +
This data is more fully described at [[Metrics/Data Catalog]].
 +
 
 +
== Linkback data ==
 +
 
 +
{{Infobox|This data is in the [http://creativecommons.org/publicdomain/zero/1.0/ public domain]. To read more about [[data|open data]].
 +
}}
  
= Linkback queries and CC API queries =
+
You can download raw MySQL dumps that are generated nightly from http://labs.creativecommons.org/metrics/sql-dumps/ -- this includes all data gathered programmatically by CC to date.
  
== Data ==
+
Single day data is available in CSV format from http://labs.creativecommons.org/metrics/csv-dumps/.  Here is a guide to the columns in the file:
  
You can download raw mysql dumps that are generated nightly from http://labs.creativecommons.org/~paulproteus/sql-dumps/all.sql.gz . This includes the schema, which is necessary to run our stats code.
+
# internal ID number for this row (e.g., 5041)
 +
# License URI (e.g., http://creativecommons.org/licenses/by-nc/1.0/jp/)
 +
# search engine (e.g., All The Web)
 +
# number of hits (e.g., 4680)
 +
# date and time this linkback query run started (e.g., 2004-Apr-04 0:00:00)
 +
# short form of license jurisdiction (e.g., jp)
 +
# short form of license name (e.g., by-nc)
 +
# license version (e.g., 1)
 +
# long form of license jurisdiction (e.g., Japan)
 +
 
 +
'''WARNING:''' There are gaps in the data and results from any given method may be volatile to extremely volatile. Take the raw numbers with a huge grain of salt.
 +
 
 +
== Flickr data ==
 +
 
 +
Information generated from Flickr is also available in the database dump above or [http://labs.creativecommons.org/metrics/stats/flickr/ here] as one-day CSVs, like this one for [http://labs.creativecommons.org/metrics/stats/flickr/2008-06-23.csv.imported June 23, 2008].
 +
 
 +
Also see [http://creativecommons.org/weblog/entry/13588 Analysis of Flickr data] as of reaching 100m CC licensed images and [http://creativecommons.org/weblog/entry/20870 around 135m CC licensed images], including a [http://wiki.creativecommons.org/images/1/19/Cc-flickr-20100225.ods spreadsheet snapshot] for the latter.
  
 
== Software ==
 
== Software ==
  
In order to use the above dumps, you must:
+
=== Data gathering ===
  
* Import the above dump into MySQL
+
The code used to gather the above data is available from the <code>stats</code> module from our [[Source Repository Information|subversion repository]].
* Download our stats module from [[Subversion]]
+
 
 +
If you want to run it yourself, here's what to do:
 +
 
 +
Check out the software:
 +
 
 +
* svn co http://code.creativecommons.org/svnroot/stats
 +
* cd stats
 +
 
 +
Configure database access:
 +
 
 +
* rename dbconfig_EXAMPLE.py to dbconfig.py with a MySQL database and password you want to use (note that if you want to use a non-MySQL database, most of the tools we use are actually database-agnostic and require only tiny changes)
 +
* mysql -h '''dbserver''' -u'''username''' -p '''databasename''' < create_tables.sql
 +
 
 +
Check for dependencies.  Note that the script expects Tor to be running on localhost!
 +
 
 +
* python sanity_check.py
 +
 
 +
Do a stats crawl!
 +
 
 +
* python link_counts.py log
 +
 
 +
=== Charting ===
 +
The <code>stats</code> module also contains some chart generating code. To run this code:
 +
* Import all.sql.gz into a MySQL database
 
* Configure the database in dbconfig.py
 
* Configure the database in dbconfig.py
 
* Run chart generation software from the stats/reports/ directory.
 
* Run chart generation software from the stats/reports/ directory.
 
+
* Warning: the output is raw and the program takes a long time to run
That's it!
 
  
 
=Baseline numbers from specific collections=
 
=Baseline numbers from specific collections=
Line 27: Line 80:
 
! Repository !! 2005-08 !! 2005-11 || 2005-12 !! 2006-01 || 2006-04 || 2006-05 || 2006-07 || 2006-09 || 2006-12 || 2007-03 || 2007-06 || 2008-01
 
! Repository !! 2005-08 !! 2005-11 || 2005-12 !! 2006-01 || 2006-04 || 2006-05 || 2006-07 || 2006-09 || 2006-12 || 2007-03 || 2007-06 || 2008-01
 
|-
 
|-
! [http://flickr.com/creativecommons Flickr] ([http://www.swivel.com/graphs/show/16858232 graph]) (photos)
+
! [http://flickr.com/creativecommons Flickr] (photos)
 
| 4.1m ||  || 7.1m ||      || 10.8m || 12.7m ||      || 19.7m || 25.5m || 32.5m || 38.7m || 57.9m
 
| 4.1m ||  || 7.1m ||      || 10.8m || 12.7m ||      || 19.7m || 25.5m || 32.5m || 38.7m || 57.9m
 
|-
 
|-
! [http://www.soundclick.com/business/license_list.cfm Soundclick] ([http://www.swivel.com/graphs/show/16860809 graph]) (audio)
+
! [http://www.soundclick.com/business/license_list.cfm Soundclick] (audio)
 
| 159k ||  ||      || 200k || 220k  ||      || 249k  ||      || 294k || 324k || 372k || 430k
 
| 159k ||  ||      || 200k || 220k  ||      || 249k  ||      || 294k || 324k || 372k || 430k
 
|-
 
|-
! [http://revver.com Revver] ([http://www.swivel.com/graphs/show/16860811 graph]) (video)
+
! [http://revver.com Revver]<sup>*</sup> (video)
 
| na  || 0 ||      ||      || 19k  ||      ||      ||      || 119k || 214k || 296k || 417k
 
| na  || 0 ||      ||      || 19k  ||      ||      ||      || 119k || 214k || 296k || 417k
 
|}
 
|}
  
Also see [http://www.jamendo.com/en/?p=stats Jamendo stats] and [http://magnatune.com/info/stats/ Magnatune stats].
+
Also see [http://wayback.archive.org/web/*/http://www.jamendo.com/en/?p=stats old Jamendo stats] and [http://magnatune.com/info/stats/ Magnatune stats].
 +
 
 +
<sup>*</sup> Revver is an overestimate, probably total number of uploads to date, some of which may have been removed or never published.
  
 
=License property charts=
 
=License property charts=
Line 48: Line 103:
  
 
=Estimates over time=
 
=Estimates over time=
 +
 +
2010-06 -- 400+ million as of December, 2010
 +
 +
2008-07-01 -- 130 million total works estimated using [http://code.creativecommons.org/viewsvn/stats/ankit/cc_total_estimate_with_comments.py?view=log Ankit's implementation] of Giorgos' scaling algorithm.
 +
 +
2008-05-02 -- 67 million photos licensed at Flickr http://flickr.com/creativecommons
 +
 
2007-06-14 -- Multifaceted metrics presented at iSummit [http://creativecommons.org/weblog/entry/7551]
 
2007-06-14 -- Multifaceted metrics presented at iSummit [http://creativecommons.org/weblog/entry/7551]
  
Line 85: Line 147:
 
2003-12  
 
2003-12  
 
:1m
 
:1m
 +
 +
== Issues ==
 +
=== Fixed ===
 +
* Until 2008-07-01, the backlinks (e.g. [http://labs.creativecommons.org/metrics/csv-dumps/2004-04-01/00:00:00/linkbacks-daily-Yahoo.csv] and [http://labs.creativecommons.org/metrics/csv-dumps/2004-04-01/00:00:00/linkbacks-daily-Yahoo.csv] for 2004-04-01, Yahoo and Google respectively) between 2004-04-01 and 2005-06-20 were incorrectly labeled.
 +
** The problem was bad importing between data formats in 2005.
 +
** The issue was fully corrected by 2008-07-01.
 +
=== Confirmed ===
 +
* Google API queries are not working properly right now (as of 2008-06-25).
 +
  
 
[[Category:FAQ]]
 
[[Category:FAQ]]
 
[[Category:Metrics]]
 
[[Category:Metrics]]

Latest revision as of 15:46, 2 March 2014

See metrics for a broader discussion of CC adoption and impact.

Caveats

Estimating license adoption is a very inexact science. There is no authoritative source and we neither control nor have inside knowledge of the construction and volatility of the most comprehensive sources -- web search engines -- primarily via Yahoo! link: queries (Google's link: operator obtains very incomplete results).

NOTE: Currently the best analysis of the data similar to that discussed below (based on a snapshot independently gathered in January, 2007 may be found in Giorgos Cheliotis' presentation on CC statistics from June, 2007.

Raw data

Creative Commons has irregularly run programs that collect estimated total results from search engine link:{license_uri} queries and queries filtered by license property (Yahoo! and Google advanced search support filtering by license).

This data is more fully described at Metrics/Data Catalog.

Linkback data

This data is in the public domain. To read more about open data.

You can download raw MySQL dumps that are generated nightly from http://labs.creativecommons.org/metrics/sql-dumps/ -- this includes all data gathered programmatically by CC to date.

Single day data is available in CSV format from http://labs.creativecommons.org/metrics/csv-dumps/. Here is a guide to the columns in the file:

  1. internal ID number for this row (e.g., 5041)
  2. License URI (e.g., http://creativecommons.org/licenses/by-nc/1.0/jp/)
  3. search engine (e.g., All The Web)
  4. number of hits (e.g., 4680)
  5. date and time this linkback query run started (e.g., 2004-Apr-04 0:00:00)
  6. short form of license jurisdiction (e.g., jp)
  7. short form of license name (e.g., by-nc)
  8. license version (e.g., 1)
  9. long form of license jurisdiction (e.g., Japan)

WARNING: There are gaps in the data and results from any given method may be volatile to extremely volatile. Take the raw numbers with a huge grain of salt.

Flickr data

Information generated from Flickr is also available in the database dump above or here as one-day CSVs, like this one for June 23, 2008.

Also see Analysis of Flickr data as of reaching 100m CC licensed images and around 135m CC licensed images, including a spreadsheet snapshot for the latter.

Software

Data gathering

The code used to gather the above data is available from the stats module from our subversion repository.

If you want to run it yourself, here's what to do:

Check out the software:

Configure database access:

  • rename dbconfig_EXAMPLE.py to dbconfig.py with a MySQL database and password you want to use (note that if you want to use a non-MySQL database, most of the tools we use are actually database-agnostic and require only tiny changes)
  • mysql -h dbserver -uusername -p databasename < create_tables.sql

Check for dependencies. Note that the script expects Tor to be running on localhost!

  • python sanity_check.py

Do a stats crawl!

  • python link_counts.py log

Charting

The stats module also contains some chart generating code. To run this code:

  • Import all.sql.gz into a MySQL database
  • Configure the database in dbconfig.py
  • Run chart generation software from the stats/reports/ directory.
  • Warning: the output is raw and the program takes a long time to run

Baseline numbers from specific collections

We can also know the number of works licensed at various content curators. The largest of these based on recent (December 2006) for various formats may (there could easily be a larger CC-licensed video collection than Revver) be:

Licensed work counts at leading repositories
Repository 2005-08 2005-11 2005-12 2006-01 2006-04 2006-05 2006-07 2006-09 2006-12 2007-03 2007-06 2008-01
Flickr (photos) 4.1m 7.1m 10.8m 12.7m 19.7m 25.5m 32.5m 38.7m 57.9m
Soundclick (audio) 159k 200k 220k 249k 294k 324k 372k 430k
Revver* (video) na 0 19k 119k 214k 296k 417k

Also see old Jamendo stats and Magnatune stats.

* Revver is an overestimate, probably total number of uploads to date, some of which may have been removed or never published.

License property charts

These charts show a breakdown of the types of licenses deployed and the properties of deployed licenses, based on Yahoo! queries as of 2006-06-13. (As above the Google API is now superior for an aggregate count, but Yahoo link: searches are superior for measuring the relative deployment of specific licenses and thus specific license types.)

Distribution of licenses deployed. Those without 'by' (attribution) were not versioned past 1.0 (excepting public domain, which is not versioned).
Distribution of license properties across licenses deployed, e.g., 3% non-by are public domain and non-by 1.0 licenses.

Estimates over time

2010-06 -- 400+ million as of December, 2010

2008-07-01 -- 130 million total works estimated using Ankit's implementation of Giorgos' scaling algorithm.

2008-05-02 -- 67 million photos licensed at Flickr http://flickr.com/creativecommons

2007-06-14 -- Multifaceted metrics presented at iSummit [1]

2007-03-31 -- 33 million photos licensed at Flickr and growth over 1 year [2]

Based on a swivel.com user's data collection from http://flickr.com/creativecommons

2006-06-13 -- 140 million pages licensed [3]

Based on Google queries.

2005-12 -- 45 million pages licensed [4]

Based on Google queries.

2005-08-09 -- 53 million pages licensed [5]

Again based on Yahoo! queries, this number turned out to be overstated as Yahoo! tuned their results estimation after growing their index.

2005-06-13 -- CC search query breakdown[6]

Breakdown of search requests and desired license properties -- people searching for video want the least freedom.

2005-05-27 -- CC in Yahoo! Advanced Search[7]

Yahoo! queries say 16m pages linking to a CC license.

2005-03-23 -- Yahoo! Search for Creative Commons[8]

Close to 14m pages link to a CC license according to Yahoo! queries.

2005-03-07 -- CC search index breakdown[9]

Breakdown of (small) CC-nutch index -- audio publishers are most permissive, video publishers least.

2005-02-25 -- License Distribution [10]

Based on Yahoo! queries there are now 10m licensed documents. Pie chart of what those licenses are.

2005-02-18 -- How many pages link to a CC license? [11]

Based on Yahoo! queries, "well over 5m." At the end of 2003 it was 1m.

2004-09-17 -- Searching for Creative Commons on Yahoo![12]

4.7m pages link to CC licenses according to Yahoo! queries.

2003-12

1m

Issues

Fixed

  • Until 2008-07-01, the backlinks (e.g. [13] and [14] for 2004-04-01, Yahoo and Google respectively) between 2004-04-01 and 2005-06-20 were incorrectly labeled.
    • The problem was bad importing between data formats in 2005.
    • The issue was fully corrected by 2008-07-01.

Confirmed

  • Google API queries are not working properly right now (as of 2008-06-25).