When you are working with a large decentralized organization or different regions geographically for example, you may want to standardize a decision-making process and criteria but still allow for flexibility in some areas. If your goal is to roll everything back up into one model you will develop your model differently than if you never integrate the models.
If the desired outcome is one integrated model, consider standardizing on common requirements but recognize that there might be unique missions from sub-organization to sub-organization. You can create a placeholder mission criterion at the parent that breaks out into a different number of unique children across the sub-organizations. As long as the scores for the child-level are normalized before they are brought into the large model and you have weighted the remaining common criteria, you can import all of the alternative scores back into one model.
If the desired outcome is not one integrated model, consider standardizing on common requirements and potentially allowing for flexibility. If there are unique requirements and you are not rolling the alternatives back into one model, you may also give the sub-organizations flexibility to weight the entire model or parts of it uniquely.