Small teams with skilled people, the startup model, works best even in large companies. Closer to the customer and more cohesive unit makes for a better project, that is why innovation happens at smaller companies or small research groups in companies or universities.
n(n-1)/2 is the formula to calculate the number of communication channels on a project where n=the number of team members/stakeholders on a project.
On a team of 1, it is all you, you make sure it works.
On a team of 2-3, it is just that few, you make sure you and others are doing their part.
On a team of 10, 45 communication connections, you might not have the same level of care because that isn't my role or someone else has got it. Beyond 7-8 team members communication becomes humanely impossible.
On a team of 100, you have so little power/responsibility that you feel out of place calling attention to issues, you don't know of everyone on the team even.
Metrics are a good thing, to use as an input, so is customer feedback, product goals/focus, usability, employee satisfaction and many other things that aren't always measured like technical debt, product quality, security, simplicity of processes/production, research and development and more.
If you measure the wrong things, it can give you the wrong idea that you are moving in the right direction while possibly overlooking aspects not captured in metrics.
If you only looked at revenue and employee head count for instance, growth might be deceiving. Profit and revenue per employee better, but also employee satisfaction, customer satisfaction, the market, product quality, timing, security, long term direction etc etc get overlooked in the metrics department mainly because domain knowledge has less power than oversight in most mature organizations.
This reminds me of bystander effect [0]. The demarkation of responsibility in large enough teams could invoke delays just in finding who is responsible for what.
Since then, however, there has been the constant desire to deskill workers. That is, they want explicit operating/work instructions that mean even a trained monkey could do the job. This is useful for some things (got a broken down Jeep? Pull out the manual and even your Lt can fix it!), but for development, acquisitions, and sustainment this is actually a horrible idea.
The infection of deskilling spread to the office. They've removed the office managers (among others) and left the work (critical, but time consuming and secondary to the mission) with the highly educated and trained staff. This diminishes their ability to focus on real work (see [0] from yesterday). From there, they continued to try to document precisely how engineering work is done. Believing that the process of the work is the same as the work. So if only my engineers knew how to properly fill out SF-1234 it wouldn't matter how educated they are, magically the work would be done.
Of course, we all know this is bullshit. Knowledge work is called that for a reason. The capacity for work is based on the knowledge of the workers. You cannot deskill computer science or aerospace engineering. You can deskill aspects of it, or really the workflow, but not the science and engineering work itself. I can eliminate or largely reduce the need for a classical configuration manager by establishing a (automated) peer review and version control workflow. But the creative act producing content that enters the workflow will always require skilled, competent, knowledgeable engineers and scientists.
[0] https://news.ycombinator.com/item?id=18157885
EDIT: I will not change the above, but I will make a note. After re-reading the Wikipedia page on Scientific Management [1] I see that some people consider Lean and others to also be extensions of it. So read the above as describing Fordism and Taylorism, not scientific management in general. Any management largely based on, or influenced by, statistics, models, and experimentantion could be argued to be "scientific management", that doesn't mean it's bad. It's when it's taken to an extreme. Like, in Taylor's case, where workers are treated with contempt. The purpose of the management being to make them fully expendable. They had no unique knowledge or skills that could really contribute to the business beyond their physical presence and ability to follow directions. His goal was to disempower workers, whereas other examples (Lean particularly) work to empower workers.
[1] https://en.wikipedia.org/wiki/Scientific_management