Are the expectations of what sales analytics can do out of touch with the reality? How do you balance human intervention with striving to be "data driven"?

You need to be data driven but you can also get an excel sheet to tell you whatever you want. You must have human intervention because you know your book of business better than what simple math will do.

An example I've personally seen multiple times as a rep was:

Customer 1 went from $40k -> $400k in year 1. You as a human understand what drove that and why it happened but all an excel sheet see's is 10x growth. I know as a rep they are not happy with us and are more likely to downsell this year than to increase on the same trajectory. Does my CFO/CEO/CRO know that?

Even if you halve that same growth multiple (taking it to $2M the next year) in the forecast who's to say that's even possible. It's difficult to quantify with data that a company won't be expanding because they're having a bad experience with your product after scaling up in yr 1 and actually may reduce or that they've now maxed out their budgetary spend and can't increase or any number of 100 other factors (Competition, lack of progress in the product, Champion has left the business, etc)

This is why it must be both data and human intervention.

It always needs to be a mix between analytics and reality. For example, there are eight different ways to forecast sales. If you only look at one or two metrics, it will give you very limited results.
Combining business analytics with the team's input and their best guess will give you very realistic information. 

I agree with one of the other answers about too much focus on discrete data changes behavior, not always for the better.

There are measurements like Monthly billing increases that are not as arbitrary as most quotas set.

The answer to this question is not in the analytics themselves but in the data on which they depend.  If your sales team becomes aware of how you measure them it is very tempting to work the system ensuring their metrics are favourable. If you say you will measure on sales cycle time they will look to log their deals later. This can extend to the sales managers as well if they start to focus on the optics which is especially true when things turn a bit tricky. In fact, it is exactly these reason that makes it is so important to work hard on data quality and leverage everything to your advantage including good governance and, of course, AI. If the data is bad then you are better off soliciting opinion and using your judgement to decide what makes the most sense but this is far from ideal. One question to ask yourself is what individual is responsible for data governance and quality in my CRM system - if you cannot answer this questions then there is a good place to start. A finishing thought is to take a moment to imagine if you had perfect data...think about the power that would give you to make the right decisions and investments...this is why the data driven journey is worth it...because although perfect data might be a distant goal as you increase data quality so will you increase the power you have to make great decisions.