There seem to be two aspects to "leadership" – individual’s ‘skills’ and organisational ‘style’. ‘Training’ required to improve ‘leadership’ capability in an individual’s ‘skills’ will have a different content to ‘training’ to improve ‘leadership’ capability/attitude ‘style’ of an organisation. So the first need will be to establish what it is that is to be invested in and therefore what is to be measured.
The next step against which "development" (I interpret that as "improvement") can be *proved* means establishing PRECISELY what it is that you want to improve. To do this you will need to write it down, clarify it, so that there is no ambiguity and within this statement define the ‘qualitative’ improvement wanted. Every ‘improvement’ required has to be described in behavioural terms. This needs to be in terms of what it is that individuals have to *DO* better &/or differently, in order to demonstrate that "leadership effectiveness" has developed.
In terms of individual skill competency this should be based in the things that "leaders" *DO* – for example, as identified by Adair in his Action Centred Leadership model. The formulating of required ‘developments’ or improvements provides a basis for a 180-degree assessment instrument. The best people to decide what ‘leadership’ improvement is required is, of course, the ‘followers’. From this survey the gaps – i.e. required outcomes – would be determined.
Only by clarifying the required "outcomes" — in terms of "leadership" ‘skills’ &/or ‘style’ — from any ‘programme’ invested in, would you have a basis for rational validation as well as evaluating the ROI. By establishing PRECISELY what development can be attributed to the ‘programme’ you have formulated assessment criteria and parameters for judging the programme achievements and thereby identifying any other contributing events or separating out impacts from parallel activities.
Obviously an overview of a fairly complex analysis/assessment approach, but only by putting time and effort into establishing *PRECISELY* what you want in the way of *improvement*, capturing relevant data, then analysing that data – selecting and rejecting – the affecting/non-affecting data can you have a rational approach to determining the ROI.