In a gross oversimplification, they posit four levels:

~ level one – the reaction (the happy sheets)
~ level two – the learning that happens in the classroom
~ level three – the behaviour change that happens in the workplace
~ level four – the return on investment/expectations

The make a couple of interesting points.

First off, that with training you have to plan backwards from level four – ie you start with the business benefits and work backwards, rather than try to find them after-the-fact.

Secondly, you can never isolate the effects of training – there are so many other factors that will influence the results being measured. What you can do is build a case. For the training example, you start with the ROI, decide what the behaviours are that would lead to that ROI, then what would need to be demonstrated in the classroom, then how that training was received.

If you can show the first three levels, you can build a case for how much of the improvements in the fourth level have happened as a result of your intervention.

This doesn’t totally cross over here at the detail level, but I think some of it might transfer.

What is the ROI for the business of the project?
What behaviours would strongly lead to that?

If you can then measure those behaviours and show how much they have increased (along with subjective/anecotal evidence), you may be able to build a case to support the action you’ve been taking.

I don’t know if that makes sense, but the concepts certainly helped me to make measurement of ROI much more approachable.