Top-down means the top is responsible for data quality's failure. Engineering can only build what the business anticipated to get, but in the end, human needs to be the one driving the mandate. When you are looking at enterprise data management, with many data format and global business, with many legacy systems in place, data management is not possible without a full commitment from the top. Heads of each business division has to be responsible for the shit their department produced. Whenever engineering shouts "data management" the business will just leave that to the engineering folks to build automated system, but who is actually looking after the data? Machines doesn't know what they are doing - someone is coding the expectation, that's all. But data management is dynamic. You can't just have an engineer shouts "this stinks". The business needs to own the shout and the mandate.