A High-Level Look at an ILS Migration

My library recently performed that most miraculous of feats—a full transition from one integrated library system to another, specifically Innovative’s Millennium to the open source Koha (supported by ByWater Solutions). We were prompted to migrate by Millennium’s approaching end-of-life and a desire to move to a more open system where we feel in greater control of our data. I’m sure many librarians have been through ILS migrations, and plenty has been written about them, but as this was my first I wanted to reflect upon the process. If you’re considering changing your ILS, or if you work in another area of librarianship & wonder how a migration looks from the systems end, I hope this post holds some value for you.

Challenges

No migration is without its problems. For starters, certain pieces of data in our old ILS weren’t accessible in any meaningful format. While Millennium has a robust “Create Lists” feature for querying & exporting different types of records (patron, bibliographic, vendor, etc.), it does not expose certain types of information. We couldn’t find a way to export detailed fines information, only a lump sum for each patron. To help with this post-migration, we saved an email listing of all itemized fines that we can refer to later. The email is saved as a shared Google Doc which allows circulation staff to comment on it as fines are resolved.

We also discovered that patron checkout history couldn’t be exported in bulk. While each patron can opt-in to a reading history & view it in the catalog, there’s no way for an administrator to download everyone’s history at once. As a solution, we kept our self-hosted Millennium instance running & can login to patrons’ accounts to retrieve their reading history upon request. Luckily, this feature wasn’t heavily used, so access to it hasn’t come up many times. We plan to keep our old, self-hosted ILS running for a year and then re-evaluate whether it’s prudent to shut it down, losing the data.

While some types of data simply couldn’t be exported, many more couldn’t emigrate in their exact same form. An ILS is a complicated piece of software, with many interdependent parts, and no two are going to represent concepts in the exact same way. To provide a concrete example: Millennium’s loan rules are based upon patron type & the item’s location, so a rule definition might resemble

  • a FACULTY patron can keep items from the MAIN SHELVES for four weeks & renew them once
  • a STUDENT patron can keep items from the MAIN SHELVES for two weeks & renew them two times

Koha, however, uses patron category & item type to determine loan rules, eschewing location as the pivotal attribute of an item. Neither implementation is wrong in any way; they both make sense, but are suited to slightly different situations. This difference necessitated completely reevaluating our item types, which didn’t previously affect loan rules. We had many, many item types because they were meant to represent the different media in our collection, not act as a hook for particular ILS functionality. Under the new system, our Associate Director of Libraries put copious work into reconfiguring & simplifying our types such that they would be compatible with our loan rules. This was a time-consuming process & it’s just one example of how a straightforward migration from one system to the next was impossible.

While some data couldn’t be exported, and others needed extensive rethinking in the new ILS, there was also information that could only be migrated after much massaging. Our patron records were a good example: under Millennium, users logged in on an insecure HTTP page with their barcode & last name. Yikes. I know, I felt terrible about it, but integration with our campus authentication & upgrading to HTTPS were both additional costs that we couldn’t afford. Now, under Koha, we can use the campus CAS (a central authentication system) & HTTPS (yay!), but wait…we don’t have the usernames for any of our patrons. So I spent a while writing Python scripts to parse our patron data, attempting to extract usernames from institutional email addresses. A system administrator also helped use unique identifying information (like phone number) to find potential patron matches in another campus database.

A more amusing example of weird Millennium data was active holds, which are stored in a single field on item records & looks like this:

P#=12312312,H#=1331,I#=999909,NNB=12/12/2016,DP=09/01/2016

Can you tell what’s going on here? With a little poking around in the system, it became apparent that letters like “NNB” stood for “date not needed by” & that other fields were identifiers connecting to patron & item records. So, once again, I wrote scripts to extract meaningful details from this silly format.

I won’t lie, the data munging was some of the most enjoyable work of the migration. Maybe I’m weird, but it was both challenging & interesting as we were suddenly forced to dive deeper into our old system and understand more of its hideous internal organs, just as we were leaving it behind. The problem-solving & sleuthing were fun & distracted me from some of the more frustrating challenges detailed above.

Finally, while we had a migration server where we tested our data & staff played around for almost a month’s time, when it came to the final leap things didn’t quite work as expected. The CAS integration, which I had so anticipated, didn’t work immediately. We started bumping into errors we hadn’t seen on the migration server. Much of this is inevitable; it’s simply unrealistic to create a perfect replica of our live catalog. We cannot, for instance, host the migration server on the exact same domain, and while that seems like a trivial difference it does affect a few things. Luckily, we had few summer classes so there was time to suffer a few setbacks & now that our fall semester is about to begin, we’re in great shape.

Difference & Repetition

Koha is primarily used by public libraries, and as such we’ve run into a few areas where common academic library functions aren’t implemented in a familiar way or are unavailable. Often, it’s that our perspective is so heavily rooted in Millennium that we need to think differently to achieve the same effect in Koha. But sometimes it’s clear that what’s a concern to us isn’t to other libraries.

For instance, bib records for serials with large numbers of issues is an ongoing struggle for us. We have many print periodicals where we have extensive holdings, including bound editions of past issues. The holdings display in the catalog is more oriented towards recent periodicals & displaying whether the latest few issues have arrived yet. That’s fine for materials like newspapers or popular magazines with few back issues, and I’ve seen a few public libraries using Koha that have minimalistic periodical records intended only to point the patron to a certain shelf. However, we have complex holdings like “issues 1 through 10 are bound together, issue 11 is missing, issues 12 through 18 are held in a separate location…” Parsing the catalog record to determine if we have a certain issue, and where it might be, is quite challenging.

Another example of the public versus academic functions: there’s no “recall” feature per se in Koha, wherein a faculty member could retrieve an item they want to place on course reserve from a student. Instead, we have tried to simulate this feature with a mixture of adjustments to our loan rules & internal reports which show the status of contested items. Recall isn’t a huge feature & isn’t used all the time, it’s not something we thought to research when selecting our new ILS, but it’s a great example of a minute difference that ended up creating a headache as we adapted to a new piece of software.

Moving from Millennium to Koha also meant we were shifting from a closed source system where we had to pay additional fees for limited API access to an open source system which boasts full read access to the database via its reporting feature. Koha’s open source nature has been perhaps the biggest boon for me during our migration. It’s very simple to look at the actual server-side code generating particular pages, or pull up specific rows in database tables, to see exactly what’s happening. In a black box ILS, everything we do is based on a vague adumbration of how we think the system operates. We can provide an input & record the output, but we’re never sure about edge cases or whether strange behavior is a bug or somehow intentional.

Koha has its share of bugs, I’ve discovered, but thankfully I’m able to jump right into the source code itself to determine what’s occurring. I’ve been able to diagnose problems by looking at open bug reports on Koha’s bugzilla tracker, pondering over perl code, and applying snippets of code from the Koha wiki or git repository. I’ve already submitted two bug patches, one of which has been pulled into the project. It’s empowering to be able to trace exactly what’s happening when troubleshooting & submit one’s own solution, or just a detailed bug report, for it. Whether or not a patch is the best way to fix an issue, being able to see precisely how the system works is deeply satisfying. It also makes it much easier to me to design JavaScript hacks that smooth over issues on the client side, be it in the staff-facing administrative functions or the public catalog.

What I Would Do Differently

Set clearer expectations.

We had Millennium for more than a decade. We invested substantial resources, both monetary & temporal, in customizing it to suit our tastes & unique collections. As we began testing the new ILS, the most common feedback from staff fell along the lines “this isn’t like it was in Millennium”. I think that would have been a less common observation, or perhaps phrased more productively, if I’d made it clear that a) it’ll take time to customize our new ILS to the degree of the old one, and b) not everything will be or needs to be the same.

Most of the customization decisions were made years ago & were never revisited. We need to return to the reason why things were set up a certain way, then determine if that reason is still legitimate, and finally find a way to achieve the best possible result in the new system. Instead, it’s felt like the process was framed more as “how do we simulate our old ILS in the new one” which sets us up for disappointment & failure from the start. I think there’s a feeling that a new system should automatically be better, and it’s true that we’re gaining several new & useful features, but we’re also losing substantial Millennium-specific customization. It’s important to realize that just because everything is not optimal out of the box doesn’t mean we cannot discover even better solutions if we approach our problems in a new light.

Encourage experimentation, deny expertise.

Because I’m the Systems Librarian, staff naturally turn to me with their systems questions. Here’s a secret: I know very little about the ILS. Like them, I’m still learning, and what’s more I’m often unfamiliar with the particular quarters of the system where they spend large amounts of time. I don’t know what it’s like to check in books & process holds all day, but our circulation staff do. It’s been tough at times when staff seek my guidance & I’m far from able to help them. Instead, we all need to approach the ongoing migration as an exploration. If we’re not sure how something works, the best way is to research & test, then test again. While Koha’s manual is long & quite detailed, it cannot (& arguably should not, lest it grow to unreasonable lengths) specify every edge case that can possibly occur. The only way to know is to test & document, which we should have emphasized & encouraged more towards the start of the process.

To be fair, many staff had reasonable expectations & performed a lot of experiments. Still, I did not do a great job of facilitating either of those as a leader. That’s truly my job as Systems Librarian during this process; I’m not here merely to mold our data so it fits perfectly in the new system, I’m here to oversee the entire transition as a process that involves data, workflows, staff, and technology.

Take more time.

Initially, the ILS migration was such an enormous amount of work that it was not clear where to start. It felt as if, for a few months before our on-site training, we did little but sit around & await a whirlwind of busyness. I wish we had a better sense of the work we could have front-loaded such that we could focus efforts on other tasks later on. For example, we ended up deleting thousands of patron, item, and bibliographic records in an effort to “clean house” & not spend effort migrating data that was unneeded in the first place. We should have attacked that much earlier, and it might have obviated the need for some work. For instance, if in the course of cleaning up Millennium we delete invalid MARC records or eliminate obscure item types, those represent fewer problems encountered later in the migration process.

Finished?

As we start our fall semester, I feel accomplished. We raced through this migration, beginning the initial stages only in April for a go-live date that would occur in June. I learned a lot & appreciated the challenge but also had one horrible epiphany: I’m still relatively young, and I hope to be in librarianship for a long time, so this is likely not the last ILS migration I’ll participate in. While that very thought gives me chills, I hope the lessons I’ve taken from this one will serve me well in the future.


Comments are closed.