[Web4lib] An Analysis Of Open Source ILS Market Penetration

Tue Oct 16 18:32:10 EDT 2007

Thanks for the question about my Koha numbers. As I pointed out several 
times in various posts, the numbers I have seem to underestimate Koha's 
impact. There are lots of issues and I apologize for the length of this post.

Marshall's database is the database for such information and it is a 
resource for the whole field. Any time I find an error or problem, as 
Marshall will, I am sure, attest, I send him a note and he corrects changes 
and errors quickly. If there are Koha libraries not listed, I did not find 
them. It would be useful if folks helped him update the dataset so when I 
do the fuller analysis, we have a good count. This is a case here ESR's 
notion of "many eyes" is operative.

For my analysis, I took only US public libraries. That strategy has several 
advantages among them that the data are better because there are 
national-level data available on all US public libraries so it is a good 
place to start a systematic means of measuring these libraries and their 
impact. As I noted, I have within my ken the US academics as well as some 
Canadian public and academic libraries (subject to limitations in Canadian 
library data.) Public libraries are defined as NCES defines them.

I discussed the counts of lib-web-cats differing from my estimates in the 
LISNews post and will repeat it here with some additions. I used a 
restrictive method--call it the Missouri ('Show me') count to assign a 
library to an open-source ILS and lib-web-cats uses a different method--I 
guess we could call it a broader method. Here I will discuss the one I used 
first and then discuss lib-web-cats' method. Good analysis should subsume 
both and there is an easy way to handle it with the public library data 
that, alas, proved impossible for reasons that I believe can be remedied.

The first difference between lib-web-cats and my estimates is that I use 
the NCLIS/NCES data. NCES aggregates data on the system level. It does 
produce something called the "outlet file" which has basic information on 
each public library outlet but the only the number it reports is the square 
footage per outlet. If you want public library data, you get the 
system-level data; there are no national-level outlet data other than what 
I have described. lib-web-cats gives counts of outlets so 61 is what 
lib-web-cats counts as Koha libraries by the method used there. Table 2 on 
the LISNews post has my count of systems and outlets. For Koha, I have 15 
systems and 23 outlets.

Why the difference? Because, I checked each system and went to its Web page 
to see what ILS the library presented while lib-web-cats changes its count 
based on an announcement. If you check Howard County now on lib-web-cats, 
it will tell you that it is a Koha library but you will be lead to its 
legacy ILS. It is not running Koha now. OK, so Marshall and I have 
different ways of thinking about this and that is fine. Good analysis can 
handle it.

But first, let's follow up on my numbers. Here are the 15 systems I found 
when I checked were actually running Koha:

   newkey     LIBNAME 
                      state
 OH0016    BARBERTON PUBLIC 
LIBRARY                                           OH
2 PA0074    BENSON MEMORIAL LIBRARY, INC.                                   PA
3 PA0069    CAMBRIDGE SPRINGS PUB LIBRARY ASSOCIATION           PA
4 PA0420    COCHRANTON AREA PUBLIC LIBRARY                               PA
5 OH0067    DELTA PUBLIC 
LIBRARY                                                     OH
6 ME0100    HARTLAND PUBLIC 
LIBRARY                                              ME
7 PA0071    JAMES A STONE MEMORIAL LIBRARY                               PA
8 PA0072    LINESVILLE COMMUNITY PUBLIC LIBRARY                         PA
9 PA0070    MARGARET SHONTZ MEM LIBRARY                                   PA
10 PA0421  MEADVILLE PUBLIC 
LIBRARY                                             PA
11 OH0156  NELSONVILLE PUBLIC 
LIBRARY                                         OH
12 PA0073  SAEGERTOWN AREA LIBRARY                                          PA
13 PA0422  SPRINGBORO PUBLIC 
LIBRARY                                          PA
14 OH0209  STOW-MUNROE FALLS PUBLIC LIBRARY                           OH
15 IA0072   WEST LIBERTY PUBLIC 
LIBRARY                                         IA

I am a little uncertain how this will render in your browsers--if anything 
makes the case for monospaced fonts it is such as this. Or I could dump out 
a spreadsheet if anyone is interested.

You can see the library name and state on the right. The newkey is the key 
variable explained in copious documentation to all <a 
href="http://www.nclis.gov/statsurv/NCES/pldf3/index.html">PLDF3</a> data. 
If you want to check my work independently, use the newkey.

NCLIS/NCES does not list Crawford County Federated Library system but the 9 
libraries in Pennsylvania in the table are all listed as members of this 
system on its <a href="http://ccfls.org/">page</a>. The page only lists 
nine libraries, however, not 10. Which one are they missing? The answer to 
the question about this system, then, is that the consortium is 
disaggregated and each system is treated separately, so the totals I report 
reflect these libraries.

lib-web-cats lists 49 PL systems running Koha by my count, so let's look at 
them in detail.

Easy ones first: Casco (Maine) Public Library. lib-web-cats reports that it 
ran a "locally developed" system prior to Koha but it is still running it 
at this minute. If you look at the source of the Web page, you will find 
the delightfully named OPAC is SLOPAC. Certainly not an endorsement for 
that piece of software!

Another easy one: as mentioned, Howard County listed as running Koha but 
still running legacy software (go to "Traditional Catalog" on the 
Aquabrowser splash page).

More difficult: Smith Center PL, Smith Center, Kansas. It doesn't present 
an OPAC Web page that I can find. By the criteria I used, I excluded it but 
it is a dubious case. But, the secondary analysis would catch it. To that 
in a bit.

There are 31 public libraries in Indiana. It seems they are members of what 
lib-web-cats lists as the Indiana Shared Catalog. You are quickly led to 
something by that name with its legacy software still running.

Now, the way to handle this difference in definition analytically is to use 
both definitions: the "Missouri" method and a very broad definition that 
includes libraries that have committed to changing. That way we bracket the 
true answer. I started to do that but ran into a problem. Look at this page:
http://www.incolsa.net/services/liblime.html

The consoritium is INCOLSA and here it announces that it is partnering with 
LibLime to offer open-source solutions. But which ones? Note the list 
headed "Products include:" and we see that LibLime will support INCOLSA 
libraries planning on (?) moving either to Koha or Evergreen, among other 
products.

Evergreen? Yep, with a nice write up down the page.

So, which is it? Or which are these 31 libraries going to use?

This is where I was stymied the other day. There did not to be a good 
analytic remedy until something is resolved but I just had a thought: I 
will add a third line to the analysis saying just something like "open 
source" and give the totals for the three groups: Koha, Evergreen, and 
INCOLSA. As INCOLSA libraries adopt one or the other, I will add them to my 
Missouri count. This seems a good solution and I can add the Smith Center 
Library, too. Give me a day or two.

PINES has 47 library systems, not 46. It takes a bit of work and 
lib-web-cats has some minor errors (it list 50 because of double counting) 
that I shall presently send Marshall the list and it will be corrected in 
time. If  you want the gories, I got them but this is long enough. Where 
did you get 46?

----
And I see that Marshall will be broadening the analysis at lib-web-cats. 
Good. It is hard to imagine how it could be better but now we know.

Bob Molyneux
drdata at molyneux.com
XyWrite forever!
0x 09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0