Monthly Archives: February 2007

Profiling Library Customers

Jim answers more questions from fellow Drillers

Want to see additional questions & answers from fellow Drillers?

Here’s the blog archive; the pre-blog email newsletter archives are here.

Q: I work for a local council in England and was recently asked to provide some local demographic profiles to help our local libraries market themselves more effectively and hopefully increase book loans.

A:  Hmmm…this is definitely the first time for this question!

Q: I’m no marketer, I usually muck about with crime, economic and census data, but this seemed instinctively wrong to me after reading how Tesco’s had used its clubcard data to understand it’s market.  When looking for some help, I’ve obviously found your website, bought the book, got the software.

A: I would have to agree with you on the Tesco point.  If you want people to “do something”, you look at behavior.  If you want people to take out more books, you look at book loan behavior.

Q: Firstly, I just wanted to thank you for making your ideas so accessible, actionable and easy to understand.  I’ve picked up some other marketing textbooks for help and they seem to mainly consist of dry schematic diagrams, and bland statements.  Great for a degree I’m sure, not so great for the rest of us who have three weeks to write a report on the subject!

A: Well, thanks for the kind words.  That was the intent of the book – to give people the “how to do” as opposed to the “what to do”.

Q: Secondly, have you any advice or experience of this model working in the non-profit sector or specifically in libraries.  For example, predicting life cycle / trigger points seems a little more complicated than the examples you use.  People don’t seem to stop using the service gradually, but stop abruptly and then start up again without no warning.  I’m also dealing with thousands of records.  My data seems a lot less clear cut than the examples you talk about.

A: I’d agree it’s not a “clear cut” situation, but not because of the models or the channel.  This kind of behavioral profiling as been used offline for decades, and it works in all kinds of situations.  For example, you mention crime data, so you have probably seen that the more Recently someone has committed a crime, the more likely they are to commit another.  Not that you can really take any action on that information – you can’t lock people up for being “likely to commit a crime” – but interesting just the same.

And I think you face a similar challenge.  You can run the scoring models and generally predict who is likely to slow or defect from their book loan behavior, but the question is, what do you “do” about that?  I have worked in other educational situations (likely to graduate, likely to contribute to the school after graduation) where the incentive is not straightforward but nonetheless you can create incentives to encourage people to continue their behavior.  I’m struggling a bit with how to create one in this situation, since the product itself is free.

But before we tackle that issue, I want to run through a bit of a “model” for this “business”.  It seems to me you have a market or segment shift going on.  If the primary reason people go to the library is to research a topic, clearly access to the Internet has suppressed the need for people to take books out on loan from the library.  For example, many of the trade journals that used to be hard to access or expensive to subscribe to are now available on the web.  So you have this “research” segment you have to deal with.

There no doubt is another segment, “core readers”, who simply for the love of reading visit the library to discover books and read them.  This segment is probably what I would call “good customers” because the library provides a service to them they cannot get elsewhere, the “value proposition” of the loan program matches their needs precisely.  This in contrast to the “research reader”, who now can do a lot of research from work or home on the web.

For research readers, a possible alternative would be providing access to internet terminals in the library.  But now we’re starting to encounter a different definition of “customer” and “loan a book”, right?  Let’s say, for a library, the “profit” in the venture is the “contribution to the community”, and this contribution was always measured in the past by “books out on loan”.  This metric has been the library’s KPI (Key Performance Indicator), if you will.

Let’s also say that many libraries have installed internet access terminals, as they have in the US.  Because of these terminals, you would expect that “books out on loan” would fall for the “research readers” segment, correct?  So to get a proper valuation of the contribution the library was making to the community, you would have to look at “books on loan + number of web terminal uses” to approximate the old metric “books out on loan”.

Follow?  So let’s say the real issue at hand here is the local government is trying to be “accountable” for what they spend on libraries, and they measure the “profit” of this spending by looking at books out on loan.  The library administrators are feeling some kind of pressure to “serve the community better” (increase “profit”) because books out on loan have fallen.  The problem is that “books out on loan” is no longer a viable metric – the “base” has changed, if you will.  The research segment is being served through a new method – web terminals – and this has been overlooked in terms of measuring the “contribution” the library is making.

In terms of tracking, if there is no login required to use a web terminal, someone in the library simply needs to count the number of people in a week that use the terminals x 52 weeks and use that as an approximation.  Better would be a system where in order to access a terminal, you have to enter your “library card number” or some other unique identifier tied to the person or household.

In this way, you find out more about your segments:

Researchers = only logs into computer
Core Readers = never log into computer
Multi = both logs into computer and takes books out on loan.

These “multi’s” would typically be the very best customers, since they are engaging in more than one library offering.

Above this, the library probably offers other types of services and special events.  The more of these services and events a customer engages in, the more valuable the customer is.  These are the customers the library should strive to keep active.  And like the example of books on loan, if the “value” of the library is only being evaluated on books on loan, attendance at these other services and events really should be included in the evaluation in some way.

The landscape has changed, and it’s quite possible that the metrics have not kept pace.  Perhaps it is not your place to suggest this, but as the “evaluator”, I would certainly be curious about this metric “books on loan” and make sure it accurately reflects what people think it does and is serving the administrators in the way they think it is.

Now that we have a feeling for what the background might be, you still have the issue of what do you “do” about customers who appear to be defecting?  As I said, the behavioral models will give you this information, but then what can be done with this information, especially given what are probably fairly strict budget constraints?  It’s not like you can send them a discount on their next book loan!

In situations like this, I think the best alternative might be survey work.  That is, when the behavioral models identify customers who are likely to defect or have defected, the library simply asks them about it.  For example, the library conducts a telephone survey using a sample of these people – “We noticed you have not taken a book out on loan / used a terminal in the past 3 months, is there anything we have done to offend you?  Is there a particular kind of content you are now interested in that we do not provide?  What other events or services would you like to see us provide?”  etc.

And where possible, the library should respond and provide what these customers want.  It may not be able to keep these particular customers from defecting (too late), but over time the “mix” of content and services should improve in a way that attracts and retains high value library customers.

I humbly suggest that the above approach will be far more effective and less costly than a “CRM system”.  If the library doesn’t have one, what is really needed is a “tracking system” that simply keeps track of what resources of the library each customer is using.  This will make your models much more reflective of the true economic benefits provided by the library, and give you the customer samples you need to construct effective, segment-targeted retention programs.

This kind of work also provides very good tracking for “how we’re doing” as a library and can provide excellent justifications for budgets and new requests, because it is directly tied to what customers want – no guessing games.  Over time, budget should flow to the areas more desired by customers and away from areas that are perhaps less desirable or are “pet projects” of interested parties.

Q: Once again thanks, I can at least begin to show management that we should start doing this work in Access / Excel before spending on a CRM system, but any further advice would be appreciated.

A: Well, I don’t know much about how libraries really work, but I’ve given you my best guess as to what might be going on.  Hope that helps!  And do keep in touch on this as you get into the guts of it, should be a very interesting project!

Jim

Update:
Robert just checked in with new info on progress on this idea, see here.

“Scrap Learning”

New phrase I’d never heard before: “Scrap Learning” is training that is delivered to students that goes unused.  For example, employees sign up for training that has nothing to do with their job just to get out of working, or are forced into this situation by some misdirected “mandate”.  Or they are forced to take a course so far ahead of when they will need the knowledge they forget this knowledge by the time an on-the-job use opportunity rolls around. 

Apparently this is a pretty big problem in large companies and wastes millions of dollars in Training budgets every year.  Reducing / Eliminating Scrap Learning is one way to optimize Training budgets for maximum ROI; if you can get rid of Scrap Learning, you spend a lot less money and get virtually the same impact.  Kind of like segmentation in Database Marketing, right? 

Some estimates of Scrap Learning run as high as 50% of all Training delivered.  One of the easiest ways to reduce Scrap Learning is to simply trigger an e-mail to the employee’s supervisor and get confirmation the employee who signed themselves up really needs to take the course.  Hmmm…

Here’s the problem.  The success of many Training departments is measured by “volume” metrics like the total number of student hours consumed.  If that is your success metric, do you have any incentive to reduce Scrap Learning?

Just got back from the Training 2007 conference.  All sorts of stuff like the above is whizzing around in my head.  The “experts” say one way to drive Creativity is to Learn material outside your own knowledge domain, then try to connect this knowledge back to your own domain.  I have that going on in a big way…

Fear of Analytics is a huge problem in HR / Training; I’m beginning to think this is a pervasive problem across the entire enterprise.  We know this is a culture problem, but the question is, what is the right model for evaluating, addressing, and fixing this problem?  The issue has been addressed here and there for specific applications, but as far as I can tell, not in any kind of comprehensive way.  What can we do about Scrap Learning?  Scrap Marketing?  Scrap Sales?  Scrap Service?

And then, think about what this looks like from a cross-silo view, the Inter-departmental Scrap, the Scrap created because there are not clear metrics channels calibrated so the success metrics of one department do not conflict with the success metrics of another department.

Brother, that’s a lot of Scrap.  Are your metrics aligned with your mission?  Or are you incented to produce a lot of Scrap?

Déjà vu (All over Again)

The top issue in Training today is:

Accountability

Execs want to know what the “ROI of Training” is.  To find out what the ROI of Training is, one should create:

KPI’s – that’s Key Performance Indicators, in case you didn’t know.

To facilitate the use of KPI’s – to provide something to measure – one should design Training so rather than being Content-based, it is Performance-based.  In other words, the Training should be designed to have a measurable outcome.

Another way to say this is the Training should have a clearly defined Goal which directly addresses the “Gap” between actual performance and desired performance.

Geesh…and I thought Marketing was up the creek…these folks are just getting started.