Starting from the end

Week before last, Will Thalheimer and I had another one of our ‘debates’, this time on the Kirkpatrick model (read the comments, too!).  We followed up last week with a live debate.  And in the course of it I said something that I want to reiterate and extend.

The reason I like the Kirkpatrick model is it emphasizes one thing that I see the industry failing to do.  Properly applied (see below), it starts with the measurable change you need to see in the organization, and you work backwards from there. You go back to the behavior change you need in the workplace to address that measure, and from there to the changes in training and/or resources to create that behavior change.  The important point is starting with a business metric.  No ‘we need a course on this’, but instead: “what business goal are we trying to impact”.

Note: the solution can just be a tool, it doesn’t have to always be learning.  For example, if what people need to access accurately are the specific product features of one of a multitude of solutions that are in rapid flux (financial packages, electronic hardware, …), trying to get it in the head accurately isn’t a good goal. Having people able to access the information ‘in the head’ is an exercise in futility, and you’re better off putting the information ‘in the world’.  (Which is why I want to change from Learning & Development to Performance & Development, it’s not about learning, it’s about doing!)

The problems with Kirkpatrick are several.  For one, even he admitted he numbered it wrong.  The starting point is numbered ‘four’, which misleads people.  So we get the phenomena that people do stage 1, sometimes stage 2, rarely do they get to stage 3, and stage 4 is almost non-existent, according to ATD research.  And stage 1, as Will rightly points out, is essentially worthless, because the correlation between what learners think of the learning and the actual impact is essentially zero!  Finally, too often Kirkpatrick is wrongly considered as only to evaluate training (even the language on the site, as the link above will show you, talks only about training). It should be about the impact of an intervention whatever the means (see above).  And the impact is what the Kirkpatrick model properly is about, as I opined in the blog debate.

So, in the live debate, I said I’d be happy for any other model that focused on working backwards. And was reminded that, well, I proposed just that a while ago!  The blog post is the short version, but I also wrote this rather longer and more rigorous paper (PDF), and I’m inclined think it’s one of my more important contributions to design (to date ;). It’s a fairly thorough look at the design process and where we go wrong (owing to our cognitive architecture), and a proposal for an alternative approach based upon sound principles.   I welcome your thoughts!

Link to original post