BIBS: Usage Statistics

This page contains usage statistics about the system. These statistics augment the statistics given in the BIBS Report.

Most of these statistics were derived from analysis of the usage log recorded by the Real Media Server. The server writes a log record at the end of every session that includes information about the location of the player (IP Address), the file being played, the date and time when the file was played, the duration of the play in seconds, and the number of bytes delivered. BIBS records info about the class and lecture date in the file name so that we can determine which lecture is being played. This log is processed into a relational database on which further queries are run to compute the statistics reported below.

Many records show play durations of zero or only a few seconds. We interpret this to mean that either the session could not be started or that the user found the experience unusable. Consequently, we remove all entries where the play duration is shorter than 60 seconds before computing usage statistics. In the spring 2001 semester this removed more than 40% of the log entries. This failure rate is troubling but other researchers have found similar statistics when examining Real Networks logs.

Summary Usage By Semester

Semester	# Classes	# Plays	Plays/Month	Enrolled Students	Plays/Enrolled Student
SP 1999	3
FA 1999	7
SP 2000	11	43,569	10,892	2,914	15
FA 2000	15	67,642	16,911	4,193	16
SP 2001	15	76,046	19,012	4,728	16
FA 2001	15	84,073	21,018	4,972	17
SP 2002	15	176,736	44,184	3,714	48

The following notes explain some of the numbers in this table.

1. We do not have usage statistics for the Spring and Fall 1999 semesters.
2. The statistics for Fall 2001 and Spring 2002 were reported by IS&T and Educational Technology Services who took over operation of BIBS that semester.
3. During Spring 2002, approximately 18% of the plays (32,252) were made during finals. This confirms other data that shows students use the lecture webcasts primarily to study for examinations.
4. The enrollment number for Fall 2000 only includes fourteen classes. Econometrics 244 was not offered that semester. The lectures viewed by students were recorded in a previous semester.
5. A dramatic increase in viewing occurred in the Spring 2002 semester. The number of plays increased by a factor of two overall and by almost a factor of three for plays per enrolled student. Evidently the service is getting more popular and/or visible to the students.
6. Viewing in Fall 2000 increased by roughly 10% with only a 5% increase in enrollments. This result could indicate that more students are watching the lectures or that more non-students are watching the webcasts.

Plays of What Semester Material

We find that every semester lectures from previous semesters are played. Some are for classes being offered in the current semester and some are for classes not being offered in the current semester, typically graduate classes. For Spring 2001, 13% of the plays were for material from a previous semester broken down as follows. The Summer 2000 plays (SU00) are for CS61A lectures that were recorded that summer.

Semester	# Plays	Percent of Total Plays
SP 99	39	0%
FA99	1,000	1%
SP00	3,074	4%
SU00	231	0%
FA00	5,457	7%
SP01	66,209	87%

Spring 2002 Plays Per Class

Class	# Enrolled	# Plays	Plays/Enrolled Student	# Live Plays	% Live
Biol1B	467	30,468	65	x	x%
Chem1A	506	9,080	18	x	x%
CS3	165	2,788	17	x	x%
CS61A	311	11,165	36	x	x%
CS61B	251	11,137	44	x	x%
CS61C	258	11,094	43	x	x%
CS150	62	16,658	269	x	x%
CS162	251	16,372	65	x	x%
EE141	61	24,867	408	x	x%
EE213	5	2,232	446	x	x%
IAS 180	164	3947	24	x	x%
IDS110	388	10,978	28	x	x%
IDS114B	99	1,027	10	x	x%
ME219	31	9,826	317	x	x%
NS10	695	15,097	22	x	x%

The following notes explain some of the strange numbers in this table.

1. We use "plays/enrolled-student" to measure the popularity of the webcasts. Each class has between 30-45 lectures unless it is a seminar (e.g., IDS114B). Clearly, the "plays/enrolled-student" for CS150, EE141, EE213, and ME219 include too many plays for the number of students. These lectures are likely being played by many non-Berkeley students. Previous semesters, the non-Berkeley plays are estimated to be roughly 30% of all plays. The availability and popularity of these classes shows people want access to the Berkeley class lectures.
2. We publish the lectures for anyone on the Internet to access for free. An experiment should be done to assess the potential for charging access to the material.

Spring 2001 Plays Per Class

Class	# Enrolled	# Plays	Plays/Enrolled Student	# Live Plays	% Live
Biol1B	396	10,721	27	0	0%
Chem1A	385	8,031	21	269	3%
Classics 28	429	3,473	8	592	17%
CS3	287	2,633	9	144	5%
CS61A	446	8,134	18	1,051	8%
CS61B	335	8,035	24	737	11%
CS298-5	12	2,271	189	348	7%
EE241	26	1,092	42	23	2%
IDS110	462	5,995	13	675	11%
IDS114B	66	409	6	0	0%
MATH53	372	3,390	9	789	23%
ME219	50	1,786	36	85	5%
NS10	657	3,184	5	127	4%
PHYS8A	479	3,882	8	110	3%
PHYS8B	326	3,173	10	165	5%

The following notes explain some of the strange numbers in this table.

1. CS298-5 is the Berkeley MIG Seminar which has a large audience outside the campus. This semester there were 14 seminars each of which was played an average of 162 times. The actual statistics for the seminars are:
   • minimum number of plays=35
   • median number of plays=120
   • maximum number of plays=419 [Jim Barton's talk on the TiVo architecture]
2. IDS114B is also a seminar class, but it is less widely known since this is the first semester it has been offered. Nevertheless, each seminar was played on average of 29 times.
3. Plays per enrolled student is typically higher for graduate classes (e.g., EE241 and ME219) because students at other universities and people in industry often watch the lectures.
4. The high number of live plays for CS61A and CS61B, which are the first two classes in computer science, has been seen several semesters. This likely results from the familiarity of the students with computers and the desire to explore streaming media technology.
5. The high number of live plays for MA53 and CLAS28 is surprising since this is the first semester these classes have been offered on BIBS.