Academic references to outcomes theory
Don't silo a means under a single end where it contributes to more than one end.
In the real world, projects/strategies/activities are often not siloed under individual outcomes. Siloing means that a lower-level strategy is only focused on improving a single higher-level outcome. In fact, many lower-level strategies are focused on achieving a number of higher-level outcomes. In fact, such multi-outcome focused strategies are often sought out as the most useful strategies to pursue. However, siloing is very attractive to anyone who wants to put a dollar value on what is being spent in the pursuit of a single outcome. Where there is siloing, putting a dollar value on an outcome is easy, it is just a matter of adding up the cost of the various siloed strategies which are being used to pursue the single high-level outcome.
In cases where siloing represents the way in which the real external world works, there is nothing wrong with taking this approach. For instance, a siloing approach is sometimes used to show the amount of funding which is being spent on traffic safety enforcement by police. To the extent that policing functions can be clearly divided into those which are focused on traffic safety in contrast to those that are focused on crime, siloing is a legitimate approach in this case.
However in instances where there is no true siloing in reality, taking a siloed approach creates a problem. This is the fiction that there is more certainty about how much is being spent on pursuing a single outcome than there really is. In such cases, siloing can have unintended negative consequences. For instance, it can lead to the selection of strategies which are more mono-outcome focused in preference to other strategies which are multi-outcome focused but inadvertently less likely to be adopted with such siloed systems.
Don't siloize. If one is attempting to put a dollar value on the expenditure focused on each of a number of outcomes, this can be done by assigning a proportion of expenditure on a strategy to different outcomes. As with any such estimation, if this proportional assignment is not based on any coherent rationale, such an analytical approach results in rubbish in, rubbish out.
An example - research granting organization
A national research granting organization had over twenty different 'buckets' for funding. Researchers had to apply under one bucket and their applications went to a funding committee only looking at that funding bucket. This meant that the funding structure looked tidy and it was possible to report on the 'amount of dollars being spent on achieving each outcome'. In the case of some research projects, those that might have been naturally siloed, this approach would not cause any problems. However it worked against the selection and funding of those research projects which contributed to a number of outcomes. However, such 'synergistic' research projects are likely to be the kind of research projects that should have been being funded.
Another example - U.S PART agency accountability system
The U.S. PART system for government agencies reporting on their outcomes, at least during part of the first decade of the 2000s required a list of 'outcome' statements to be prepared by agencies. Because of the expectation that these should be 'outcomes-focused' these statements tended, in some instances, to be high-level rather than lower-level statements controllable by the agency. This meant that in some cases the statements lacked controllable indicators which could be used for direct accountability (this is an example of the Single Indicator List Problem).
At least some technical advice given in workshops on how the PART statements should be formulated by agencies, proposed that each separate PART statement should include an outcome (not-necessarily controllable) and an output (controllable) so that the statement could include both high-level and lower-level indicators. The problem with some interpretations of this proposed solution to the PART Single Indicator List Problem is that it could inadvertently lead to siloization. This would occur if those generating the statements did not feel relaxed about having the same output appearing in more than one statement even thought those different statements had different high-level outcomes in them.