These tools are really only as good as the information they are provided or access, and without looking at failure rates/reasons in the model do seem fundamenta

These tools are really only as good as the information they are provided or access, and without looking at failure rates/reasons in the model do seem fundamentally flawed to me. As mentioned by others, the results provided do look very optimistic (at least from what I can tell based on the values it provides for things I am involved with) and with refinement once they get the datasets will probably go down significantly in most cases. Time will be the real judge. It will be interesting is to see what they predict and what actually happens in the next years. Hopefully they will provide this information (semi-)publically, but it seems unlikely.